Options, Perpetuals and Prediction Markets: size, traders, behavior, pain
How big each leverage product is on-chain and off-chain, who trades it, how long they hold, how often they get liquidated, what they complain about, and what happens when a trader is right on the thesis but wrong on the mechanics.
Data as of 9 Sep 2026 (BTC ≈ $78k)7 research threads · ~405 linked sources (primary and secondary) Includes an original Hyperliquid on-chain sample (686k fills, 420 accounts)
How to read this brief
Every figure carries its period and source. Confidence badges: High primary source, corroborated · Med single source or definitional spread · Low proxy or analyst estimate. Tags: derived computed from sourced inputs · est analyst estimate with stated logic · unverified reported but not confirmed — do not put on a slide.
Five measurement traps that explain most conflicting numbers in the press: (1) prediction-market "volume" is quoted as notional (contracts × $1) or dollar (contracts × price) and they differ 2–4×; (2) perps aggregators differ by venue scope — The Block tracks 15 CEXs (~$65T for 2025), Coinglass/CoinGecko count everything (~$86T); (3) Hyperliquid's own stats feed counts both sides of each trade, roughly 2× third-party figures; (4) DeFi options notional is gameable, so premium is the honest size metric; (5) "users" are addresses, not people.
Sources are linked inline as short tags and listed in §9 (six inline-cited URLs are not repeated in §9, and §9 also lists further reading not cited inline). Original computations are marked "own sample" and are reproducible from the saved scripts.
§1
Headline figures — the numbers for the deck
The figures below are the ones most likely to end up on a slide or in the memo. Each row links to the section with the underlying evidence.
Figure
Value
Period
Source
Conf.
Options
US listed options volume
15.2B contracts in 2025 (+24%); 2026 YTD ADV 70.5M (+23%), Cboe guides FY-2026 >18B (flat YTD ADV would give ≈ 17.7–17.8B)
≈ 400 : 1 central case (range 125–1,000 : 1 on 5–10M active; 1,900 : 1 if the 18.7M ever-traded pool is used) — 5–10M active US retail options traders (18.7M have ever traded) vs 10–40k professionals est
2024–26
§4.4
Med
Retail : pro volume (TradFi options)
≈ 1 : 1 on the initiating side; market makers on ~49% of all sides
≈ $1.8–2.5T notional in 2025 derived (Deribit $1.354T ÷ an assumed 54–75% share; the Deribit shares cited elsewhere in this brief, 49–80%, would give $1.7–2.8T); Deribit alone $1.35T options, $1.875T total; only 2–3% of crypto derivatives volume
Deribit ≈ 85% institutional by volume (CEO); whole market retail ≈ 15–35% est; headcount retail : pro ≈ 10–50 : 1 (order-of-magnitude; only the pro side ~5–15k is derived, no retail count exists) est
$2.5–2.7B notional / $72M premium per 30 days; ≈ 1–2% of whole-market crypto options volume (≈ 1–5% of Deribit alone) derived, ≈ 1% of OI; Derive is 85–90% of it with 24,604 lifetime traders
$6.4–7.9T in 2025 (aggregator-dependent); DEX share 3.4% (2024) → ~10–13% (2025) → 11–13% (2026 YTD), peak 21.6% in Nov 2025 (The Block basis: DEX volume against The Block's 15-venue CEX total; on the Coinglass all-venue basis the 2025 share is ≈ 7.5%)
~10–30M active crypto derivatives traders monthly est; on-chain 2–5M lifetime addresses, 0.3–1M monthly est; Hyperliquid 1.2–1.7M addresses, ~45–50k daily actives (Q1-26; feed ends 3 Apr 2026), top 100 addresses = 81% of volume
2026
§3, §5.1
Low
How long a perp position lives
Median 57 minutes; 50.7% closed within 1 h, 88.9% within 24 h, 1.8% open > 7 days (own sample, 80,501 Hyperliquid positions)
Mar–Sep 2026
§5.2
Med
How many get liquidated
53.6% of active accounts liquidated at least once in ~4.5 months (64% of the retail tier); 3.7% of all positions end in liquidation; 27% of random wallets lost >85% of capital within 30 days (a drawdown measure, not a liquidation count)
2024–26
§5.2
Med
Perp trader PnL
Only 14–25% of wallets profitable (three independent Hyperliquid studies); top 1% of accounts take 57% of all gains
2024–26
§5.2
High
Liquidation volume
~$150B reported in 2025 (true figure higher — Binance reports one order/second); 10–11 Oct 2025: $19.1B, 1.62M accounts, Coinglass' own estimate $30–40B; H1 2026 $73.4B
A 1× BTC long paid 24.2% of notional in funding in 2024 and 10.6% in 2025 (121% and 53% of margin at 5×); HYPE longs paid 22% in 2025; peaks of 500–5,000% APR on single hours
2024–26
§5.2
Med
Prediction markets
Volume
2025: Kalshi $23.8B + Polymarket ~$21B notional ($44B combined; the §2 KPI's $63B is all tracked incl. points-driven Opinion). 2026 YTD to 8 Sep: $277B notional for the big three (~$109B dollar basis unverified — not reproducible from the monthly dollar figures; components sum to $112B) ; Aug 2026 $45.3B, first combined decline since May 2026 on the notes' own series (The Block's "first drop in a year" rests on an April figure of $29.8B that is not reproducible — the components sum to $24.9B)
Polymarket 733k monthly actives (Mar-26) → 643k (Apr-26); Kalshi 1.2M active (Dec-25) + 3M new users in the World Cup; ~6–7M real people have ever traded one (components sum to ≈ 7.3M; Robinhood overlaps Kalshi) est
2025–26
§6.2
Med
Capital duration (open interest by market horizon)
79% of Polymarket open interest sits in markets resolving in <30 days, >40% in <5 days (not a trader holding-period measure); 85–95% of Kalshi volume resolves within days (sports + 15-minute crypto)
2026
§6.2
Med
Liquidation
None — fully collateralized YES/NO pairs at both venues. Analogs: one side of every binary expires at zero; 84% of Polymarket addresses are net losers; disputed markets total $972M of volume
—
§6.2
High
Bots, wash, arb vs real users
Wash ≈ 25% of Polymarket lifetime volume (Columbia), ~60% in Dec-2024 weeks, ~1% median in 2026 after fees; 12.6% of addresses = 81% of notional (2.3-day Apr-26 fill sample, legacy CTF exchange only); Kalshi retail takers have lost $583.5M to professional makers (Roosevelt; Kalshi disputes the study's trader classification)
2024–26
§6.3
Med
Trader PnL
84.1% of 2.5M Polymarket addresses in the red; only 2% ever made >$1k; <0.04% of addresses captured >70% of profits (that last figure is DeFi Oasis: 1.7M addresses, Dec-25 — a separate sample)
~90–200k people whose job is trading (narrow); ~35–90k of them trade derivatives est
2026
§3
Low
Retail traders
~700M–1B people hold equities; ~100–250M trade actively; ~25–60M trade derivatives in a year est
2025–26
§3
Med
Crypto traders
~740–775M owners; ~40–100M active traders; ~10–30M active derivatives traders monthly est
2025–26
§3
Low
§2
Market size overview — on-chain vs off-chain
One table across the three products. Perps dwarf everything by turnover; listed options dwarf everything by contract count and open interest; prediction markets are the smallest but the fastest-growing, and the only one of the three where the off-chain venue (Kalshi) overtook the on-chain one within a year.
Market
Latest annual volume
Latest month
Open interest
Largest venues
Sources
Listed options — TradFi (off-chain)
Global 88.7B contracts (2025, −50% on India curbs); US 15.2B contracts (2025, +24%); US premium ≈ $598B/month in Mar-2024 → ≈ $7T/yr run-rate then, volumes 25–50% higher now derived
US Aug-26: 1.44B contracts; global Jul-26: 9.9B options; US ADV record 72.8M (Q2-26)
Global ETD OI 1.67B contracts (Jul-26, +14% YoY); $5.7T notional expired on 20 Mar 2026 alone
India NSE (≈77% of global option contracts in 2024, before the SEBI curbs; APAC as a whole still 77% in H1-26 — share basis unverified) ; OCC/Cboe — SPX 5.1M/day (Q2-26 record quarterly ADV; 3.9M/day FY-25), SPY 42% of ETF options; 0DTE = 65–66% of SPX
≈ $1.8–2.5T notional (2025) derived (Deribit $1.354T ÷ an assumed 54–75% share; the brief's cited 49–80% shares would give $1.7–2.8T); Deribit 2025: $1.875T total, ≈ $1.35T options
Deribit record $266B (Oct-25); Jul-26 ≈ $50B options; CoinGlass H1-26 BTC+ETH options $865B across Deribit/Bybit/Binance/OKX
BTC options OI peak $108–120B (Oct-25), $63–65B (Q1-26); BTC options OI has exceeded futures OI since Jul-25
Deribit (49% of H1-26 volume, 75–80% of crypto-native CEX OI — ≈ 41–43% of all BTC options OI at $26.9B vs $63–65B, ~85% institutional) , Bybit 22%, Binance 13%, OKX 13%; IBIT options OI ≈ $27–33B; CME options OI $9B (Oct-25); Bullish >$6B
≈ $30B notional / ≈ $0.9B premium annualized from the current 30-day run-rate derived; lifetime DeFi options notional $75B
$2.5–2.7B notional / $72M premium (30d to 9 Sep 2026)
Derive ≈ $1B total OI at its Mar-26 peak (61% options); category TVL ≈ $103M ex-Derive (sum of the §4.3 table; ≈ $85M excluding Paradex's shared perps TVL)
Total futures OI ≈ $100B (Jun-26); record $235.9B (7 Oct 2025); BTC $29.5B + ETH $18.8B on 12 CEXs (8 Sep 2026)
Binance (27% broad / 51% of 15-venue set), OKX, Gate, Bybit, Bitget, MEXC, CME (4% of volume but 12% of OI — H1-26 third-party share; ≈ 10% of the 9-Sep OI column in §5.1), Coinbase Intl
$22–24B sector OI (9 Sep 2026, DefiLlama; Hyperliquid $14.7B ÷ ≈ 62% derived); Hyperliquid $14.7B (≈62% of DEX OI; 9.3% of all perp OI is a Jul-26 third-party share — ≈ 11–15% against the Sep-26 OI figures in this row)
Hyperliquid (40–44% of DEX volume), Aster, Lighter, edgeX, Variational, ApeX, Jupiter, dYdX, GMX — several with incentive-inflated prints
Kalshi $23.8B notional (2025, +1,108%); 2026 YTD $192B notional / ≈ $75B dollar unverified (to 8 Sep; the dollar figure is not reproducible from the monthly dollar figures)
Aug-26: Kalshi $37–40B notional ($11.4B dollar); Polymarket US $3.8B; Robinhood 13.6B event contracts in Q2-26
Kalshi ≈ $1.0B (peak ~$1.4–2B at the World Cup)
Kalshi (82% of big-three volume; distributed via Robinhood, Coinbase, PrizePicks, Clear Street), Polymarket US, DraftKings DKeX ($11B annualized), OG.com, FanDuel/CME; Betfair Exchange £84B matched in 2025 as the legacy reference (Covers article only, unverified; not covered by the sources linked in this row)
2025 crypto turnover in $T (CEX+DEX perps ≈ $92T; crypto options ≈ $2T; prediction markets ≈ $63B notional, all tracked incl. points-driven Opinion; $44B for Kalshi + Polymarket alone) . By open interest the picture inverts: BTC options OI ($63–120B) exceeds BTC futures OI.
On-chain share of turnover
11% · 1–2% · 12–20%
Perps (Aug-26, The Block) · crypto options (30d, DefiLlama vs whole market; ≈ 1–5% vs Deribit alone) · prediction markets (Aug-26, notional vs dollar basis).
Where on-chain is concentrated
1 venue each
Hyperliquid ≈ 62% of DEX perp OI; Derive ≈ 85–90% of DeFi options volume; Polymarket ≈ 75% of on-chain prediction dollar volume.
§3
Trader populations — professional, retail, crypto
No regulator or exchange publishes a global trader census, so every headcount below is a range built from registrations, filings, exchange disclosures and adoption studies. The ranges are wide on purpose; the anchors are the citable parts.
Bucket
Headcount range
Derivatives-trading subset
Conf.
Hard anchors (with period)
Professional / institutional traders
~90–200k people whose job is to trade (sell-side desks 30–60k, hedge funds 15–35k, buy-side dealing desks 10–25k, prop/market-making 10–25k, CTAs/commodity houses 10–25k, crypto-native 3–10k) est. ~1.5–3M if every licensed investment professional is counted.
~35–90k professional derivatives traders (options/futures market makers, rates/FX/equity-derivatives desks, vol funds, CTAs, crypto derivatives desks) est
Low
FINRA 639,723 registered reps (YE-2025); BLS 531,000 securities/commodities sales agents (2025); SEC Form PF 9,940 hedge funds / 1,834 advisers (Q3-25); IOSCO 2,709 qualifying hedge funds; NFA 1,164 CTAs, 36,859 associates (Jul-26); the 11 largest prop/MM firms employ ≈ 19–20k people in total (whole-firm headcounts across all asset classes, mostly per Wikipedia; the nine listed sum to ≈ 18.5–19k: Jane Street ~3,000–3,500, SIG 3,500+, Optiver 2,112, IMC 2,070+, Jump 2,000+, DRW 2,000, Citadel Securities 1,800, HRT ~1,000, Virtu 969); CFA charterholders 200k+ (upper bound of credentialed pros)
Crypto-native institutional
~3–10k trader headcount est; 900+ crypto hedge/VC funds tracked; 55% of traditional hedge funds now hold crypto, 67% of those via derivatives
Most of them — Deribit is ~85% institutional by volume
Low
Crypto Fund Research (Jun-26); PwC/AIMA 7th report (2025); Paradigm: 1,000+ institutional counterparties, 50+ market makers, 20+ prime dealers; Coinbase Deribit revenue booked as institutional
Retail investors (holders)
~700M–1B unique people hold equities worldwide est
—
Med
China 250.7M investor accounts (end-2025); India 131M unique investors / 237.7M demat accounts (2026); US 62% of adults own stock (Gallup 2025); Japan 37.4M individual accounts / 16.0M shareholders; Korea 14.6M; Germany 14.1M; Brazil 5.5M
Active retail traders
~100–250M trade at least monthly est (India's 19% active/demat ratio applied globally gives 130–190M)
~25–60M trade derivatives in a year; ~10–25M monthly est
Med / Low
India: 46M NSE-active clients (Aug-26) and 7.86M individual F&O traders in FY26 (down from 9.81M; 87.7% lost money); US: Schwab 39.9M active accounts and 11.6M trades/day, Robinhood 28.4M funded customers and 774M options contracts in Q2-26, Webull 5.13M funded / 213M contracts, IBKR 5.46M accounts — Robinhood + Webull alone ≈ 20% of all OCC options volume; retail FX/CFD ~10–15M (industry rule of thumb, unverified)
Crypto owners
~740–775M (Crypto.com 741M Dec-25 → 774M Jun-26; a16z ~716M in 2025; Triple-A 560M is a 2024 figure)
—
Med
All are on-chain/statistical inferences, not KYC counts — treat as ±15%. US: ~10% of adults used crypto in 2025 (Fed SHED); 21% own per NCA/Harris
Active crypto traders
~40–100M monthly est (a16z: 40–70M active users, 181M monthly active addresses; Binance 320M+ registered, OKX ~4% active/registered, Coinbase 8.2M monthly transacting users in Q1-26)
~10–30M active crypto derivatives traders monthly; ~30–60M per year est. On-chain perps: ~2–5M lifetime addresses, ~0.3–1M monthly est
Low
Perps are ~6.5× spot volume on CEXs (Q2-26: top-10 perp CEXs $12.7T vs spot $1.95T); no exchange discloses derivatives-user counts; Bitget: 200k "professional traders" and 1.1M copy-trading followers on a "120M" base; Hyperliquid 1.2M users (Apr-26) / 1.73M trading addresses (Mar-26), Lighter 188k accounts, edgeX 191k addresses
Sources for this table are listed under §9 "Trader populations". The single largest data gap for a TAM argument is that no exchange — Binance, Bybit, OKX, Bitget, Deribit — publishes how many of its users trade derivatives.
§4 · Options
Options — size on- and off-chain, venues, and who actually trades them
4.1 TradFi listed options (off-chain)
Global listed derivatives (FIA)
Total F&O
Options
Options YoY
Note
FY 2024
205.3B
177.1B
+64%
India's NSE/BSE peaked at 16B contracts/month (Oct-24); 84% of the world's equity option contracts traded in India
FY 2025
119.3B
88.7B
−50%
SEBI curbs cut India to 4B/month by Mar-25; North America +24% to 24.5B; end-2025 OI 1.52B contracts (+23%)
H1 2026
75.6B
56.5B
+41%
APAC still 77% of option contracts; North America 1.74B in June alone (narrower scope than the 24.5B/yr F&O total — consistent only with options-only; unverified)
Jan–Jul 2026
88.6B
66.3B
+40%
End-July OI 1.67B contracts (+14%)
US listed options (OCC / Cboe)
Contracts
YoY
Mix & detail
FY 2025
15.21B
+24.4%
Equity 8.27B / ETF 5.68B / index 1.26B; ADV 60.8M; record day 110M+ contracts on 10 Oct 2025; 21 days above 70M; SPX 3.9M/day, VIX 858k/day
Q2 2026
ADV 72.8M
+19%
Index +25%, ETF +27%, single-stock +6% YTD; SPX = 81% of index options, SPY = 42% of ETF options, NVDA and TSLA 9% each of single-stock; institutional blocks ≈ 10M contracts/day; 0DTE >20M/day across all US products, not SPX alone (+46%)
57% (Q3-25 quarterly avg) → 62% record (Aug-25, Cboe) → 59% (FY-25 annual avg) → 63% (Feb-26) → 65% (May-26) → 66.2% record (Jul-26) — mixed quarterly/annual/monthly windows, so the underlying rise is ≈ +4 pp from Aug-25, not +9 pp ; retail 47–60% of 0DTE flow, ~53% mid-2025; >95% of 0DTE trades are capped-risk
Premium / notional sizing
≈ $7T/yr
—
US option premium traded was $598B in Mar-2024 alone (FIA) — ≈ $7T/yr run-rate then, with volumes 25–50% higher since derived. A single March-2026 expiry carried $5.7T notional (index $4.1T). No official annual notional series exists.
4.2 Crypto options on centralized venues (off-chain)
Venue
Volume
Open interest
Users / mix
Note
Deribit
2024: $1.185T total, options $743B (+99%). 2025: $1.875T total, ≈ $1.354T options (BTC $1.131T, ETH $212B, USDC-linear only $10.9B); record month $266B (Oct-25). Jan–Jul 2026: $666B total, ≈ $493B options; Jul-26 ≈ $50B options — the market cooled hard in 2026.
Platform OI ATH >$48B (Nov-24) → "surpassed $60B" (2025); BTC options OI record $42.5B (May-25), $26.9B (Apr-26); record expiry $28.5B on 26 Dec 2025
Users not published. CEO: "approximately 85% of our trading volume comes from institutional investors"; ~15% retail. Block RFQ = 27.5% of volume (Jun-25).
Acquired by Coinbase for $2.9B ($700M cash + 11M COIN shares), closed 14 Aug 2025. Share of BTC+ETH option volume fell from 56% (Jan-26) to 42% (Jun-26) as Bybit/Binance/OKX grew.
Bybit · Binance · OKX
CoinGlass H1-26 BTC+ETH options: total $864.6B — Deribit 49.3%, Bybit 22.3%, Binance 13.4%, OKX 13.3%; Bybit leads ETH options (38% vs Deribit 29%)
Not disclosed per venue
No user or retail/institutional disclosure; Bybit positions itself as the retail options venue
Options are 2.4% of crypto derivatives volume in H1-26
IBIT options (BlackRock ETF, US-listed)
Record 1.58M call contracts in one day (19 Aug 2026); >1M/day three days running
$27.6B (Apr-26, first time above Deribit's $26.9B per Volmex) / ~$33B = 52% of BTC options OI (Jan-26, Checkonchain) — methodologies differ
Accessible to any US retail brokerage account; Deribit cannot serve US retail
The regulated venues (IBIT + CME + Bullish) now hold roughly half or more of BTC options OI (the brief's own components IBIT $27.6B + CME $9B + Bullish $6B ≈ $43B ≈ 66–68% of $63–65B; methodologies differ)
CME crypto options
Crypto suite (futures + options) ADV 278–280k contracts ≈ $12B/day in 2025; Q2-26 notional $459B; Aug-26 ADV 175k ≈ $12B/day
Crypto options OI record $9B (Oct-25); suite ADOI 216k contracts (Q2-26)
>1,000 large open-interest holders (institutional)
24/7 trading since 29 May 2026; options not broken out monthly
Bullish
>$9B option volume from Oct-25 launch to early 2026 (company-reported)
$3B (Jan-26) → >$6B (Apr-26); claims #2 among crypto-native exchanges by BTC options OI (behind Deribit; IBIT $27.6B and CME $9B are larger)
Institutional
Illustrates how fast regulated/institutional venues absorb hedging flow
Options vs perps, in crypto
Crypto options turned over ≈ $2T in 2025 against ≈ $92T of perps — 2–3% of derivatives volume, perps ≈ 35–50× options (Laevitas independently put options "just under 3%"). By open interest the ranking flips: total BTC options OI ($63–65B in Q1-26, peak $108–120B in Oct-25) has exceeded BTC futures OI since July 2025. In US equities, listed options out-trade futures more than 10:1 by contract count. The gap between crypto's 3% and TradFi's structure is the headroom argument.
DefiLlama, 9 Sep 2026 (ranking page vs API). Derive $2.25–2.27B of it.
30-day premium, all DeFi options
$72.5M
−33% w/w. Premium ≈ 2.9% of notional. DefiLlama defaults to premium because notional "is extremely easy to fake, so it does get faked".
Share of crypto options
≈ 1–2%
of whole-market volume (≈ 1–5% vs Deribit alone: $2.5–2.7B vs Deribit's $50B options in Jul-26; Deribit's $266B record is a total-platform month); ≈ 1% of OI (Derive ~$0.6B all-asset options OI at its Mar-26 peak vs $63–65B BTC options OI; ≈ 0.8% on a BTC-only basis). Derive's own peak claim: ~8% of Deribit's daily volume on 10 Mar 2026.
Protocol
Chain(s)
TVL
30d notional / premium
Users
Model & status
Derive (ex-Lyra)
Own OP-stack chain; collateral on Hyperliquid L1 (33% of TVL), Base, Arbitrum, Ethereum
$158M
$2.25B / $68.9M; cumulative $22.5B / $604M; >$1B/month in Feb-26; record week $294M
4.4 Institutional vs retail — headcount ratio and volume ratio
Market
Volume split (retail vs professional)
Headcount (retail : professional)
Conf.
US listed equity & index options
Retail ≈ 45–55% of contracts: Cboe "roughly half" (May-26); PFOF-routed retail = 46% of OCC-cleared contracts (Q4-25); "customer" origin (all non-MM, non-broker-dealer agency flow, institutional customers included, so an upper bound for retail) = 45.8% of sides (MEMX, Aug-25) ; academic 62% in 2019–21. Market makers sit on ≈ 49% of all sides, broker-dealer proprietary ≈ 5%. SPX 0DTE: retail 47–60%. Retail : pro ≈ 1 : 1 on the initiating side.
Retail: 18.7M Americans have ever traded options (FINRA Foundation 2024: 21% of ~89M investors — the ~89M investor base is not sourced unverified); est. 5–10M active (Robinhood alone implies 3.7–7M est = 13–25% of its 28.4M Q2-26 funded customers, where 13% is Robinhood's 2021 options-adoption share and the 25% is unsourced) . Pro: est. 10–40k people whose job is options trading (the top-5 option market makers employ ~13–14k people in total (whole-firm headcounts; the five largest listed in §3 sum to ≈ 12.5–13k); 17 proprietary option MMs; 14 specialist firms). ≈ 125 : 1 to 1,000 : 1 on 5–10M active (1,900 : 1 using the 18.7M ever-traded pool); central ≈ 400 : 1 — the average professional trades ~400× the contracts of the average retail participant.
Volume High · headcount Med
India equity F&O (the largest retail options market)
Individuals = 30% of F&O notional turnover; proprietary traders and FPIs took ₹33k cr + ₹28k cr of gross profit in FY24, 96–97% of it via algorithms
9.6M individual F&O traders (FY24) → 9.8M (FY25) → 7.9M (FY26); 93% (FY22–24), 91% (FY25) and 87.7% (FY26) lost money; aggregate losses ₹1.8 lakh crore over FY22–24 and ₹91,685 cr in FY26; 43% under 30
High
Crypto options (CEX)
Deribit (49% of volume, 75–80% of crypto-native CEX OI, ≈ 41–43% of all BTC options OI): 80–85% institutional / ~15% retail (CEO, Jan-25 and Oct-25); block trades 27.5% of Deribit volume. Bybit/Binance/OKX (49% of H1-26 volume) disclose no mix. Whole-market retail ≈ 15–35% (central ~25%) → retail : pro ≈ 1 : 3 by volume est
No venue publishes user counts. Proxies: Paradigm network = 1,000+ institutional counterparties, 50+ market makers → ~5–15k professionals. Retail active options traders unknown; Derive's ~700/week (median ≈ 725) is the only primary on-chain count . Order of magnitude ≈ 10 : 1 to 50 : 1 — an order of magnitude more "professional" than TradFi by headcount, 3× by volume est
Volume Med · headcount Low
On-chain options
Intermediated by a handful of professional market makers; MM share of volume not disclosed
Why retail participation in crypto options is low (sourced)
The volatility dimension. Deribit's CEO: options "are a bit more complex than anything else… especially because of volatility" — a trader can be right on direction and still lose; retail reaches for perps for leverage instead.
Data and education gap. Bybit (Aug-26): implied vol, OI distribution and gamma exposure data "has historically been a key barrier to entry for retail traders". Derive (Mar-26): gamma risk is "the biggest challenge for new crypto options traders"; "quality educational content… remains scarce".
Access. Deribit is ex-US; US retail routes to IBIT options. Its CEO frames the Coinbase deal as reaching "previously unreachable user groups, such as the retail market".
Product conventions. Deribit's linear USDC options were $10.9B of $1,354B in 2025 — the coin-settled inverse contract, crypto-denominated margin and lot conventions are unfamiliar to retail.
Outcomes invite friction. Retail options traders lose on average (US academic estimate $2.1B over 2019–21; India 88–93% of individuals lose (87.7% in FY26) ), which is what drives regulator curbs like SEBI's.
Why DeFi options have struggled (sourced)
Fake notional, thin premium. DefiLlama switched its default to premium in Dec-2024 after Aevo's notional was gamed; premium is ~2.9% of notional across DeFi today.
Exploits and shutdowns. Moby (Jan-25), Typus (Oct-25), Thetanuts (2026, twice); Opyn abandoned options for perps; Premia, Stryke, Ithaca, IVX and Moby show zero current volume.
LP economics. Hegic's 30-day fees are negative; Aevo's token is −99%; category TVL ex-Derive is ≈ $103M (≈ $85M excluding Paradex's shared perps TVL).
Market-maker dependence. The one scaled venue (Derive) is a CLOB reliant on professional MMs, 79% BTC, with record days driven by single $130M structures — and it is ≈ 0.8% of BTC options OI (BTC-only basis at its Mar-26 peak; ≈ 1% all-asset) while being ~89% of DeFi .
Regulated competition. IBIT/CME/Bullish absorbed the institutional hedging flow on-chain venues hoped to win (IBIT options OI $27–33B vs all DeFi options OI < $1B).
$85.7T (+47%); ≈ $235B/day (85.7T/365 — Coinglass's own "$264.5B/day" implies a ~324-day window); peak day 10 Oct ≈ $748B
H1: $35.1T (−16% YoY); Q1 $18.8T, Q2 $16.3T; derivatives ≈ 76% of all CEX activity (≈ 3.2× spot on that basis; the "9.6× spot" figure is on a different, unreconciled basis unverified)
Venue (CEX)
Aug-26 volume
Share (15-venue)
H1-26 volume (Coinglass)
H1-26 share
Open interest (9 Sep)
Users
Binance
$1,594B
50.9%
$9.34T
26.6%
$32.6B
323M registered (Jul-26); no derivatives-user disclosure
OKX
$639B
20.4%
$4.19T
11.9%
$8.9B
"120M" (Jun-26) unverified (same figure as Bitget's row; OKX's own 2024–25 disclosure was 60M+); 2.5M active on a 60M base in 2024 (≈ 4%)
DEX share of crypto perpetuals volume, monthly, Jan 2025 – Aug 2026 (The Block basis: 17 DEX protocols vs 15 CEXs). 2024 average was 3.4%.
Perp DEX volume
2024
2025
2026 YTD
Note
The Block (17 protocols)
≈ $1.8T
$6.91T
$3.72T (Jan–Aug)
Record month Oct-25 $1.19T; DEX/CEX ratio avg 10.3% in 2025, 12.7% in 2026 YTD
CoinGecko (top-10/12)
$1.50T
$6.38–6.7T (+346%)
Jan–Apr avg $612B/month
DEX share of perp OI 13.5% (Apr-26)
DefiLlama (50+ protocols)
≈ $2.4T
$7.9T (H1 $2.1T, H2 $5.7T)
Aug-26 $649B; trailing 30d $593B
Wider scope incl. HIP-3 and incentive-heavy venues; Q3-25 alone ($1.8T on The Block's 17-protocol basis) roughly equalled The Block's ≈ $1.8T for all of 2024
Venue (DEX)
Aug-26 (The Block)
30d volume / OI (DefiLlama, 9 Sep)
2025 / 2026 YTD
Users & caveats
Hyperliquid
$208B core + $86B HIP-3 = $242B
$217B / $14.7B
$2.93T / $1.66T (+$0.55T HIP-3)
1.2M users (Apr-26 feed) / 1.73M trading addresses (Mar-26); DAU ~50k (peak 99k, Oct-25); fees $67M in Aug-26; top 100 addresses = 81.3% of volume
Aster (BNB)
$50.5B
$62.3B / $2.6B
$791B / $579B
Self-reports "45.9M users" unverified (figure not traceable to Aster; its published self-reports are 2M, 5M+ and 7.9M) — implausible vs 256k token holders; Coinglass flagged low liquidation-to-volume as incentive/points signal; The Block revised its Sep-25 Aster print from $420B to $18B unverified (no such revision found: The Block still shows ≈ $420B for Sep-25 and $18B was a 24-hour milestone; the delisting was DefiLlama's, 5 Oct 2025)
Lighter (zk L2)
$39.2B
$48.1B / $1.5B
$1.31T / $458B
188k accounts, 50k+ DAU (Sep-25); points-farming volume pre-TGE ($232B in the 30 days before its Dec-25 token)
The 2023–24 leaders have shrunk to <2% of the sector; dYdX lifetime $1.59T; Drift exploited for $285–295M (Apr-26), now $0
Hyperliquid in numbers (the on-chain benchmark)
Cumulative perp volume $5.37T (The Block, single-sided) vs $8.4T on Hyperliquid's own feed (double-sided) — every third-party share in this brief uses the single-sided basis.
Record month $400B core perps (Aug-25; third-party single-sided basis — the own stats feed shows $840B two-sided); record day $29.1B (10 Oct 2025; third-party figure unverified — the own feed shows $65.8B two-sided); OI $14.7B on 9 Sep 2026 (DefiLlama; The Block reports $14.3B for 8 Sep; series max $15.9B, Aug-25 unverified — no OI series is saved in the bundle); HIP-3 builder markets now 26% of OI and $86–115B/month, 98% of it trade.xyz's TradFi perps.
Share of global perps: 7.5% record (Jun-26); 9.3% of global perp OI (Jul-26); ≈ 15% of Binance's volume (Aug-26: $242B vs Binance $1,594B per the §5.1 tables derived).
Fees: $972M in 2025 (revenue $818M); $542M Jan–Aug 2026; gross revenue fell from $457M (Q3-25) to $202M (Q2-26); ~97% of fees go to HYPE buybacks.
HLP vault: $188M TVL (down from a $604M peak in Sep-25); lifetime PnL $138M — of which ≈ 44% was earned in two liquidation events (10 Oct 2025 and 31 Jan 2026).
Users: 291k (end-24) → 917k (end-25) → 1.2M (Apr-26); daily uniques averaged 33k in 2025, ~45k over 1 Jan–3 Apr 2026 (≈ 51k in March; feed ends 3 Apr 2026).
5.2 Behavior and outcomes — holding time, liquidation, PnL, funding
No exchange publishes holding-period, liquidation-frequency or PnL distributions. To get them we sampled Hyperliquid directly: 420 accounts drawn at random from the leaderboard (stratified by lifetime volume, all active in the last 30 days), their last ≤2,000 fills each — 686,363 fills over a median 138-day window — reconstructed into 80,501 completed position episodes (position leaves zero → returns to zero or flips). Figures from this are labelled own sample; they are exact for Hyperliquid and a proxy for CEX retail, who by Binance's own disclosure use more leverage.
How long positions stay open
Holding time of closed perp positions, Hyperliquid, own sample (n = 80,501 episodes, Mar–Sep 2026). Share of positions by time from open to flat.
Cohort (lifetime volume)
Episodes
Median hold
p25 / p75 / p90
Closed < 1 h
Closed < 24 h
Open > 7 d
Ends in liquidation
Positive PnL
All sampled
80,501
57 min
8 min / 6.9 h / 26.7 h
50.7%
88.9%
1.8%
3.7%
47.2%
Active retail ($10M–$100M)
48,996
67 min
10 min / 7.9 h / 29.7 h
48.4%
87.9%
2.1%
4.3%
48.0%
Mid ($100M–$1B)
25,252
36 min
5 min / 4.7 h / 21.8 h
56.6%
90.8%
1.5%
2.9%
47.4%
Whale / market maker (>$1B)
6,253
88 min
12 min / 8.3 h / 27.7 h
44.8%
88.3%
1.2%
2.2%
39.2%
Accounts with < $10k equity
46,949
42 min
—
55%
93%
—
4.4%
—
Accounts with $100k–$1M equity
12,624
153 min
—
37%
81%
—
3.4%
—
Notional-weighted median hold is 2.5 h — bigger positions live a little longer, but 86–93% of users in every cohort have a median hold under one day.
Caveats: excludes dormant/wiped accounts and accounts under ~$10M lifetime volume; positions opened before the fill window or still open are dropped, so multi-week holds are under-represented; the whale tier mixes market makers with directional whales.
Corroboration: a Nov-2024 study of 10,000 random Hyperliquid wallets found 40.6% stopped trading within the month; NYU Stern's simulation of 125× BTC perps has a median time-to-liquidation of 46 seconds (98.3% liquidated); BitMEX's own trollbox study describes attention "focused on price action from the last hour or so".
Liquidations — how much, how often, who
Period
Market-wide (Coinglass)
Hyperliquid on-chain (own calc from its stats feed)
Note
2024
not published
$15.4B (avg $42M/day; 1.55% of volume)
—
2025
≈ $150B ($400–500M/day)
$92.6B (avg $254M/day; 1.53% of volume)
Hyperliquid alone is 62% of the reported market-wide total, which is impossible — Coinglass notes Binance reports only one liquidation order per second, so the true market total is materially higher
H1 2026
$73.4B (longs 62%, shorts 38%; ≈ $405M/day)
$17.0B (1 Jan–3 Apr; feed stale after)
Jan $15.0B, Feb $12.7B, Mar $9.8B, Apr $9.3B, May $10.4B, Jun $16.1B
~$203M ETH/USDT (Hyperliquid, per Coinglass; not cross-checked against the bundle's per-user liquidation feed)
100%-tariff post; $3.21B liquidated in one 60-second window; Hyperliquid $10.3B (Coinglass) / $15.1B (its own feed); Bybit $4.65B; Binance $2.41B
21 Nov 2025
$1.91B
93%
391,164
$36.8M BTC (Hyperliquid)
BTC breaks $82k
30–31 Jan 2026
$1.68B (Coinglass H1: $2.59B on 31 Jan)
93%
267,370
$80.6M BTC (HTX)
BTC $84k → $76k
5 Feb 2026
$2.14B unverified (no source found for this print; the Feb-2026 coverage cited in §9 reports ≈ $1.8B on 6 Feb, other coverage $1B+ to $2.6B)
—
—
—
BTC low $62k
20 Aug 2026
$3.02B — largest short-liquidation day since 2021
8% (shorts $2.77B)
171,045
$48.8M BTC (Hyperliquid)
Short squeeze into the current rally
Who gets liquidated (own sample)
53.6% of the 420 active accounts were the liquidated party at least once inside their window (median 138 days): 64% of active retail, 54% of mid-tier, 25% of whales/MMs.
3.7% of all position episodes end in liquidation (4.4% for accounts under $10k equity, 1.1% above $1M).
0.93% of all fills are liquidation fills; liquidations ≈ 1.5% of Hyperliquid volume (Kraken 1.9–3.0%, BitMEX 1.3% on the long side per academic studies not cited here — unverified; the published BitMEX figure (arXiv 2102.04591) is 3.51% long / 1.89% short of daily positions, not a share of volume ).
Catastrophic outcomes
27.2% of 10,000 random Hyperliquid wallets lost more than 85% of their capital within 30 days (Nov-24).
10 Oct 2025 on Hyperliquid: 1,010 traders lost >$100k, 206 lost >$1M; 358 accounts wiped (CCN) vs "6,300 wallets eliminated" (Blockscope) — conflicting counts.
NYU Stern simulation: 125× → 98.3% liquidated, 75× → 97.3%, 25× → 93.0% before a take-profit is hit.
Leverage actually used
Binance Futures (2019): "over 80% of traders… trade at a leverage of 20x or higher", 20% at 100×+; institutions (81% of volume) ≤ 20×. Binance capped new users at 20× in Jul-21. BitMEX 2018–19 weighted average 20–35× (unsourced — unverified).
Hyperliquid today (own sample, 1,265 open positions): median leverage setting 7×; 48% ≥ 10×, 15% ≥ 20×, 4% at the 40× cap; effective account leverage median 6.4×; 13% of positions within 10% of their liquidation price.
ESMA capped retail crypto CFDs at 2:1 in 2018 after finding 74–89% of retail accounts lose money.
The vault on the other side (HLP, own calc from API)
Lifetime PnL $138M: ≈ $49.7M in 2024, $67.5M in 2025, $19.7M 2026 YTD.
+$41.4M in the 1–15 Oct 2025 window and +$18.8M around the 31 Jan 2026 ETH whale liquidation — two events ≈ 44% of lifetime profit. HLP's best days are its traders' worst.
Losses: −$4.0M (Mar-25 ETH whale), −$4.7M (29 Oct–12 Nov 2025 bucket of the 14-day PnL feed, which spans the POPCAT bad debt reported as $4.9M elsewhere in this brief) ; TVL fell 69% from its Sep-25 peak.
PnL distribution — how many perp traders make money
Study
Sample / period
Profitable
Detail
ENVY Protocol
10,000 random Hyperliquid wallets, 30 days, Nov-24
16.5%
73.8% losing, 9.7% flat; median wallet −$67; profitability rises with size: $0–100k equity ≈ 15%, $1M–10M 42%, $10M–100M 86%
Hyperdash / BeInCrypto
1,000 Hyperliquid traders, Jun-25
13.5%
86.5% losing (135 of 1,000, Hyperdash); 170 traders > $10M profit and 1,589 > $1M are analyst DeFi Mochi's leaderboard counts, not the Hyperdash sample, on 499k users at the time
TechFlow / HTX
~43,000 Hyperliquid addresses, May-26
≈ 25%
Winners are either algorithms (261k trades, 64.8% win rate) or high-conviction accounts (50 trades, 28% win rate, $4.5M)
Own calc — Hyperliquid leaderboard
35,529 accounts with ≥ $10M lifetime volume, 9 Sep 2026
40.7% (upper bound)
58.8% losing; median −$7,510 / −24% ROI; top 1% of accounts = 57% of all gains, top 10% = 93%; 2.4% are down > $1M; survivorship bias (wiped accounts drop off) and HYPE appreciation inflate this
Own sample — position level
80,501 episodes
47.2% of positions
Mean net PnL per episode +$109; whales −$2,105 per episode (they lose per trade and win on size/fees elsewhere)
Reference: day-trading studies
Brazil equity futures 2013–15; Taiwan 1992–2006; ESMA CFDs 2018
1–3%
97% of Brazilian day traders persisting > 300 days lost money; < 1% of Taiwanese day traders predictably profitable net of fees (~3% with any positive net return) ; 74–89% of retail CFD accounts lose
Binance's only public statement on outcomes is qualitative ("majority of users… do not use high leverage levels"); no CEX publishes a profitability disclosure. The wallet-sample studies (14–25% profitable) are the right retail number ; the leaderboard's 41% is a ceiling.
Mechanics. Hyperliquid's interest component alone is 0.01%/8h ≈ 11.6% APR paid to shorts; funding settles hourly and is capped at 4%/hour (35,040% APR). Rule of thumb: +0.05%/8h ≈ 55%/yr, +0.10%/8h ≈ 109%/yr. OKX's BTC-USDT swap averaged 4.4% annualized for Jun–Sep 2026 — consistent with Hyperliquid.
What it costs a correct trade. A BTC long held through 2024 paid 24.2% of notional — 121% of its initial margin at 5× — while BTC roughly doubled. BTC 15 Feb–15 Apr 2024: 8.3% of notional in two months (42% of 5× margin). DOGE 15 Feb–1 Apr 2024: 15% (75% of 5× margin). HYPE 2025: 22.1%. A 30-day trend trade at 5× during a 100%-APR regime gives up ~41% of its margin to funding. Monthly regimes above 100% APR: DOGE Mar-24 (127%), HYPE Dec-24 (119%), TRUMP Jan-25 (107%); BTC ran 67% in Mar-24 and 33% in Nov-24.
Extremes cut both ways. On 10 Oct 2025 FARTCOIN printed −4,832% APR (annualized) and TRUMP −2,978% APR (annualized) (shorts paying) ; TRUMP averaged −60% APR through Mar–Apr 2026.
Total transfer. No market-wide figure is published. Own calc for Hyperliquid (daily OI × summed hourly rates, 8 coins ≈ 81% of OI): $177M in 2024, $768M in 2025 (BTC $282M, ETH $192M, HYPE $161M) → ≈ $0.95B for all coins, ≈ 13% of average OI. Extrapolating 8–13% to market-wide OI ($124–236B in 2025) gives an order-of-magnitude $10–31B net paid long→short in 2025 (8–13% of $124–236B; the upper bound uses the 7 Oct record OI) estLow.
No survey systematically ranks perp-trader pain points (the only behavioral data found: after the 22 Sep 2025 wipeout, "liquidation checks rose 32%" and "funding-rate re-checks climbed 35%" across 106k trade setups on Leverage.Trading — a vendor press release on The Block's press-release channel reporting usage of its own leverage calculator; figures self-reported unverified). The ranking below is by how much money each complaint has demonstrably cost, with the sourced incident behind it.
#
Complaint
What actually happened (sourced)
Voices
1
Auto-deleveraging — "punished for being right"
10–11 Oct 2025: Hyperliquid ran ~35,000 ADL executions across ~19,300 wallets and 162 tickers in about ten minutes, closing $2.1B of positions in 12 minutes and, per Gauntlet's Tarun Chitra, ≈ $654M of winning positions; his paper puts the excess haircut on those traders at $45–52M versus an optimal policy. He also finds Binance "overutilized ADL far more than Hyperliquid" (Binance account data is not public). Hyperliquid's rebuttal: ADL "does not transfer P&L to HLP", is symmetrical, and closed shorts at favorable prices. OI on Hyperliquid halved afterwards unverified (no OI series is saved in the bundle).
Wintermute's Evgeny Gaevoy (on Binance, not Hyperliquid): institutions that were short BTC and long alts as a hedge were "slapped in the face" by ADL — "the market price was $1, while our short position was forcibly liquidated by the system at $5". Anonymous trader: "It felt like robbery. My short was closed at the worst possible time — right before the market tanked further."
2
Oracle / index failures and venue-only depegs
Same day on Binance: USDe printed $0.62–0.66, wBETH and BNSOL 5–7% off underlying, ATOM to ~$0.01 — the depeg happened "only on Binance", collateral was marked down and liquidations cascaded. Binance blamed old limit orders, paid $283M in compensation and launched a $400M "Together" fund. Hyperliquid JELLY (26 Mar 2025): a trader pumped thin spot to squeeze a short into HLP (−$13.5M unrealized); validators delisted and force-settled at $0.0095 vs a $0.50 oracle; $293M left HLP in two weeks.
Bitget CEO Gracy Chen: Hyperliquid "is moving towards the model of the collapsed FTX". Arthur Hayes: "Let's stop pretending Hyperliquid is decentralized."
3
Outages and stop-losses that don't fire
10 Oct 2025: Binance's interface froze and API connections lagged; dYdX was offline 8 hours, Lighter 4.5 hours; Coinbase and Robinhood halted briefly; "stop-loss orders… failed to execute or filled far below the stop price". Binance's insurance fund absorbed ~$188M (≈ 8% of its reported $2.4B liquidations — a denominator that Coinglass and row 6 below call understated; on Coinglass's $30–40B scaling the share is ≈ 4–5%) .
"$19B liquidated and somehow nobody at Binance is responsible lol." Binance co-CEO Richard Teng: "The US equity market plunged $1.5 trillion… The crypto market is much smaller. It was about $19 billion." Hyperliquid's Jeff Yan on his own systems that night: "I'm just looking at it and praying that it's good."
4
Liquidation hunting, scam wicks, whale games
12 Mar 2025: a whale opened a ~$306M ETH long at 50×, withdrew collateral and let himself be liquidated onto HLP (+$1.86M for him, −$4M for HLP) — Hyperliquid cut max leverage to 40×/25×. 26–27 Aug 2025: four coordinated addresses ripped XPL pre-market from ~$0.60 to $1.80 in two minutes, booking >$46M while OI collapsed from $153M to $22M. 13 Nov 2025: a $20M POPCAT buy wall was pulled and $20–30M of longs liquidated in seconds (HLP −$4.9M). 25 Nov 2025: a whale spent $14.5M in 15 minutes to push HYPE funding to −800% and liquidate $43M of shorts. Cross-venue dislocations reached 10% on 10 Oct.
James Wynn alleged market makers "deliberately target his liquidation prices"; a liquidated XPL hedger: "1x hedge, account destroyed, and lost half of my XPL allocation."
5
Funding bleed
Quantified in 5.2: a BTC long paid 24% of notional in 2024; HYPE longs 22% in 2025; ZEC shorts paid $536k in five to six weeks (11 Oct–19 Nov 2025) on a thesis that then played out (story B1). f(x) Protocol markets explicitly against it: "hold on to your leverage positions longer without fear of rising funding rates because we don't charge any funding fees".
Presto Research on Wynn's trade: Hyperliquid BTC funding ran 160%+ annualized, 35.6 pp above the CEX average — every long paid it.
6
Opaque liquidation engines and under-reporting
Binance reports one liquidation order per second to data providers, so public totals (Coinglass $19.1B) understate the day — Coinglass itself says $30–40B. Hyperliquid's fully transparent 2025 total ($92.6B) is 62% of the "market-wide" $150B.
—
7
Counterparty risk
FTX (Nov-22): an "$8 billion hole", >1M users, recoveries only through bankruptcy — FTX is the commonly cited motive for self-custody perps, but the brief's own series shows the on-chain share stayed under 6% until mid-2025 before reaching ~11–13%.
5.4 Trader stories — right on the thesis, wrong on the mechanics
Every story below is sourced to an X post, an on-chain tracker (Lookonchain/Arkham/Hyperdash) or a news write-up, and the "thesis outcome" line checks where the price actually went afterwards against Yahoo Finance / CoinGecko daily data. Exchange wicks are often deeper than these index prints. Anything not confirmed is tagged unverified. Grouped: A right direction, liquidated before the move · B right direction, bled by funding · C stop hunts, scam wicks, oracle incidents · D ADL victims · E who collected.
A1 · James Wynn · Hyperliquid · May–Jun 2025
The $1.25B BTC long, liquidated 4.8% below entry — BTC was 14–20% higher within months
Long BTC at 40×: 7,764 BTC ($830M) at $105,033 avg on 21 May, scaled to 11,588 BTC (~$1.25–1.27B), avg entry $106,144–108,243; liquidation ~4.8% below.
30 May: liquidated at $104,950 / $104,150 / $104,620 (1,044 BTC, ~$110M) as BTC printed $103,686. ~$37M realized loss on the final BTC long (the cited Presto and The Block pieces headline $85M and >$100M — the peak-equity drawdown and the principal plus peak paper profit given back); account $90.3M → ~$5M in 8 days; Lookonchain's recap: $21.8M principal + $87M peak profit, all gone, account ended at $23.
Funding: Hyperliquid BTC funding ran 160%+ APR during the position, 35.6 pp above the Binance/Bybit average — he paid to hold the right view.
4 Jun: re-longed ~$100M at 40× from $106,066 (liq $103,930); ~240 BTC liquidated the same day; withdrew his last $700k on 6 Jun. Returned Mar-2026 with $3,911 and a 40× short; liquidated six times in two weeks, $900 left.
Thesis outcome: BTC $110,380 on 10 Jun (five days after his last wipe), $123,092 on 14 Jul, $126,198 on 6 Oct 2025 — 14–20% above his entry. His own words: "I took a large and calculated bet at making billions."
BTC printed $102,822 on 13 Jun (Israel–Iran strikes) and $98,286 on 22 Jun → $12.5M loss; re-longed from ~$106k, was up ~$10M at ~$108.8k, closed at −$2.5M on the pullback. Nov-25 recap: $37.6M principal, $41.7M peak profit, account wiped.
Thesis outcome: BTC $123,092 on 14 Jul and $124,457 on 14 Aug — 13–15% above entry.
A4 · Andrew Tate · Hyperliquid · Jun 2025 (and 108 liquidations since)
25× ETH long at $2,516 liquidated within hours; ETH doubled in eleven weeks
10 Jun 2025: posted a 25× ETH long at ~$2,515.90; liquidated the same day (post deleted). Lookonchain: 76 trades, 27 wins (35.5%), −$583k by 12 Jun; $727k deposited, $0 withdrawn; account hit zero on 18 Nov 2025; 108th liquidation on 17 Jun 2026 (57 BTC at 40×, liq $65,216).
Thesis outcome (Jun-25 ETH long): ETH wicked to $2,444 on 13 Jun, then $3,080 on 14 Jul and $4,954 on 24 Aug — +97% from his entry. (His Nov-25 BTC longs near $90k were simply wrong: BTC fell to $62k by Feb-26.)
Short BTC at $117.8k, liquidated at $122.8k by a $123.1k wick — BTC was below his entry within a day and 5% lower in three weeks
3–9 Jul: eight forced closures ($12.5M); "liquidated 8 times in 5 hours" — 1,177 BTC ($128M) and 34,466 ETH ($87M); weekend of 5–6 Jul five more; account $16.3M → $610k.
12 Jul: redeemed 10M USDC from Maker to keep shorting — 1,273 BTC 40× short at $117,828 (liq $122,789) + 27,700 ETH 25× short at $2,966 (liq $3,196). Liquidated within ~3 hours on 13–14 Jul as BTC printed its $123,092 all-time high; ~$334M notional closed (the liquidated book had by then grown to 1,743 BTC + 33,743 ETH + 15M FARTCOIN per BeInCrypto/CCN; the two positions listed above were ≈ $232M at entry); cumulative −$25.8M (CCN; CoinDesk had already reported $37M lost in the week to 7 Jul, so the cited sources conflict) unverified.
Thesis outcome (BTC leg): the ATH was $300 above his liquidation; BTC $115,766 the next day, $114,760 on 25 Jul, $112,006 on 2 Aug. (The ETH leg was wrong — ETH went to $4,954.)
Machi Big Brother, "Cyantarb" and four $13–19M wallets: longs closed at the wick, ETH +24% within 72 hours
Trump's 100%-tariff post at 14:57 UTC; BTC $122,510 → $104,582, ETH $4,396 → $3,460 (index; deeper on venues); $19.13B liquidated in 24h (Coinglass; a later ~$19.3B print also circulated), 87% longs, ~1.6M accounts; $3.21B in a single 60-second window; Hyperliquid ~$9.3–10.3B of it (third-party estimates; Coinglass $10.3B, $15.1B on Hyperliquid's own feed). USDe printed $0.65 on Binance only.
Machi Big Brother (Huang Licheng): ~$79M ETH long at 15× liquidated; account swung from +$44.5M cumulative to about −$10M; >$53M lost in the month; 145 liquidations after 11 Oct, 335 lifetime; $30k left by Mar-26 with $75.2M total losses.
"Cyantarb" (8th-largest account by volume): ~$18–19M lost, ~$100M notional wiped. Four wallets in one Lookonchain post: −$18.7M, −$16.4M, −$15.7M, −$13.7M — two emptied to $0, the others left with $140 and $104.
Aggregate on Hyperliquid: >6,300 wallets lost $1.23B+; 205 lost >$1M; 1,070 lost >$100k.
Thesis outcome: BTC $116,020 on 13 Oct (+11% from the low), ETH $4,293 (+24% from the index low, more vs exchange wicks) — the liquidated longs were right for three days. Honest caveat: both then rolled over (BTC $80,660 on 21 Nov; ETH $2,627), so the vindication here is the wick recovery, not the quarter.
Shorted BTC with $7M, lost ~$6.2–6.9M unverified on a bounce — BTC then fell 22%
Deposited $7M on 6 Nov; short peaked at 2,222 BTC (~$226M); liquidated repeatedly as BTC bounced from $98,990 to $106,565; $560k left by 10 Nov (loss figures conflict: $7M → $560k implies $6.44M, Lookonchain's 10 Nov post says $6.19M lost, headline $6.9M). Rebuilt to $3.1M and on 17 Nov held Hyperliquid's largest BTC short: $123M at 40×, entry $95,270, liquidation $96,100 — 0.5% away, with that day's high at $95,928.
Thesis outcome: BTC $80,660 on 21 Nov — 22% below where his first short was liquidated. Whether the 17 Nov position survived the $95,928 print is not documented unverified.
A8 · Wick events with verified recoveries · 2024–2026
Event slides where the price came back within days
5 Aug 2024: BTC $58,269 → $49,121 intraday, $62,674 three days later; ETH $2,696 → $2,123, $2,722 three days later.
3 Feb 2025: ETH $2,919 → $2,159 intraday and closed at $2,885 — a 26% round trip inside one session; BTC $102,514 → $91,243, closed $101,405; $2.23B liquidated, 729k accounts.
21 Nov 2025: five Hyperliquid liquidations above $10M inside one minute (largest $36.8M) as BTC broke $82k; BTC $91,898 six days later (+14%).
5–6 Feb 2026: BTC $62,354 low then $71,681 high within 24 hours (+15%); ETH $1,819 → $2,090. Named victim: Garrett Jin's ~$472M ETH long liquidated below $2,500 on 31 Jan, ~$250M loss unverified — and here the thesis was not vindicated; ETH kept falling.
The ZEC max-short: $536k of funding in five to six weeks (11 Oct–19 Nov), $22M drawdown , $6M liquidated — then ZEC fell 70%, exactly the thesis
Short ZEC from 11 Oct at $305.91 avg (later ~$360), 10×. ZEC ran $229 → $713 (7 Nov) → $657 (21 Nov).
7 Nov: −$21.5M unrealized on a $37.5M position; 16–17 Nov: −$22.0M, $5M margin added, liquidation $1,112; 19 Nov: −$13.8M unrealized "of which $536K funding fee loss alone"; stop-loss closed $8M for −$1.78M realized; 12–15 Dec: ~$6.1M more liquidated; 16 Dec: the remainder finally in profit.
Thesis outcome: ZEC $429 on 1 Dec → $340 on 31 Jan → $220 on 1 Mar 2026 — 70% off the peak and below his entry. Right idea; funding and margin calls took most of it. (ZEC then ripped to $1,265 in Sep-2026 — see B4.)
Paid 254–800% APR to be right by 85–95% — if they survived the wick
After the 27 Aug squeeze (C3), XPL traded 22–36% above Binance pre-market with Hyperliquid funding at 242–254% annualized (28–29 Aug), spiking to ~800% after the mark-price change — shorts paying longs. WLFI pre-market funding hit 100–105% APR the same week.
Presto's forensic on the Wynn trade: Hyperliquid BTC perp funding reached 160%+ annualized, 35.6 pp above the CEX average (22–24 May) — every long paid it, and BTC went on to $126k.
B4 · Wallet 0x92ea (attributed to a Jin-linked entity) · Hyperliquid · Sep 2026 — ongoing
$2.05M of funding paid on a BTC long that is up $4.4M; a ZEC short 26 wins deep and one 10% rally from zero
EmberCN (7 Sep 2026): ~$107M BTC long with $4.42M unrealized after $2.05M paid in funding, alongside a 32,760–39,760 ZEC short from ~$444–576 now −$21.9–25.7M (ZEC $1,228; liquidation $2,540). Attribution unverified. A second whale (0x362a, 26-win streak) is short 15,785 ZEC from $866.90 with liquidation $1,317 vs mark $1,203 — "a 10% rally erases the $3.76M account."
B5 · The other side of the funding ledger · Jun 2025
0xa31 collected $2.275M of funding in three days shorting 16 alts
$57.5M of shorts across ETH, PEPE, DOGE, XRP, HYPE and 11 others; 14 of 16 profitable; $2.275M funding received in 72 hours — paid by longs unverified (on $57.5M this implies ≈ 480% APR across the basket, far above the 2025 averages in the funding table; the tracker total may be cumulative for the account rather than three days).
Funding was trivial; the liquidation line was decisive — a $46M realized loss two months before the thesis paid
0x8def: short 1.8M HYPE (~$103M) from $44.96 at 5× cross, liquidation ~$69; −$22.2M unrealized on 21 May 2026 with only $204k funding collected.
loracle.hl: short 1.8M HYPE, liquidation $69.90; dumped 616,675 HYPE ($36.8M) on 22 May to add margin; closed 3 Jun for a $46.46M realized loss (funding paid ~$54k) with HYPE at ~$69.7 — then flipped long.
Thesis outcome: HYPE peaked $73.56 on 17 Jun then fell to $64.97 (1 Jul) and $52.52 (1 Aug) — 25% below where the short was capitulated. (HYPE then rallied to $88 in Sep-2026.)
C1–C2 · JELLY and the ETH self-liquidation · Hyperliquid · Mar 2025
Two ways to make the vault pay
JELLY (26 Mar): a trader paired a Hyperliquid JELLY short with spot buying in thin DEX liquidity, pumped spot, and let the short self-liquidate into HLP; HLP's unrealized loss peaked ~$13.5M; validators delisted and force-settled at $0.0095 vs a ~$0.50 oracle; HYPE −20%.
ETH whale 0xf3F (12 Mar): 50× ETH long >$300M, liquidation 1.6% from entry; withdrew collateral instead of closing, got liquidated at ~$1,915, booked +$1.86M while HLP absorbed −$4M; >$130M left HLP; max leverage cut to 40× BTC / 25× ETH. The same wallet later closed a $520M 40× BTC short for +$9.4M (ZachXBT: "a cybercriminal gambling with stolen funds").
"1x hedge, account destroyed" — shorts right by 85–95% (if they survived the wick), wiped by a 200% two-minute wick
XPL hyperp ripped ~$0.60 → $1.80 in ~2 minutes (Binance pre-market ~$0.55); OI collapsed $153M → $22.4M; user losses est. ~$60M; four coordinated addresses booked >$46M, one of them $16M in a minute.
Named victims: KOL @Cbb0fe lost ~$2.5M hedging 10% of his airdrop allocation; 0x64a4 −$2M; one address ~−$7M; two shorts deposited $44M + $29M of emergency margin to push liquidations to $4.03 / $10.73. Hyperliquid: "no technical issues"; added a 10× EMA mark-price cap and an external pre-market price input.
Thesis outcome: shorts at $0.60 were right by 85–95% (XPL $0.09 by 2026) — if they survived the wick.
HYPE (25 Nov): a whale deposited 32M USDC, TWAP-bought $14.5M HYPE in 15 minutes targeting shorts within ~5% of liquidation, pushed funding to −800% annualized, then market-sold 100k HYPE; short OI fell from 2.1M HYPE ($67M) to 727k ($24M). HYPE then fell from $33.75 to $21.09 by 21 Jan 2026 — the squeezed shorts were right by 37% within eight weeks.
ENA (23 Oct): −10% in two minutes on Hyperliquid; a Kang-linked address had $47.4M of ENA liquidated plus ~$8.5M of top-ups; ENA closed higher the next day, then collapsed to $0.25 in November (bounce vindicated, thesis not).
POPCAT (13 Nov): 19 wallets built ~$30M of longs behind a ~$20M buy wall at $0.21, pulled the wall, and $20–30M was liquidated within seconds; HLP lost $4.9M.
D1–D2 · ADL victims · Hyperliquid and Binance · 10–11 Oct 2025
Hedged correctly, closed anyway
Hyperliquid: ~35,000 ADL events across ~19,300 wallets and 162 tickers ; Chitra: "ADL was used repeatedly to close $2 billion of positions in 12 minutes… over 40 times in a 10 minute period"; ≈ $654M of winning positions haircut, $45–52M more than an optimal policy would have. @seanlippel noted ATOM/APT shorts ADL'd right before rebounds. Hyperliquid's Jeff: ADL "does not transfer P&L to HLP" and is "completely symmetrical".
Binance: Wintermute's Gaevoy — hedged institutions "slapped in the face" by ADL, short closed "at $5" when the market was "$1"; "Although we incurred some losses on ADL, we also made a lot due to high volatility." Forbes reported he was weighing legal action; a $450M BTC-USDT long closed, a Chinese hedge fund −$180M, and "Wintermute over $300M" unverified — the last conflicts with Gaevoy's own statement. Arthur Cheong (DeFiance) solicited victims for a class action.
The widely shared "$5M spot long + $5M perp short, ADL closes the short, naked long liquidated" case (Odaily) is illustrative, not a documented account — don't attribute it to a person.
0xb317 / Garrett Jin, 10 Oct: ~30 minutes before the tariff post, ~$450M BTC + ~$350M ETH shorts on Hyperliquid funded with $80M USDC deposited that day; profit $160–200M (CoinDesk: $192M); then a $340M 10× short from $116,009 and a $227M short closed +$6.4M. Jin: "this isn't insider trading… clients' fund". His later ETH long was liquidated on 31 Jan 2026.
Abraxas Capital: 10× BTC/ETH/SOL shorts hedging spot with >$112M unrealized (Jun-25); $760M notional across 22 shorts, account $165M → $256M in a week (Nov-25).
Squeezers and collectors: the XPL cartel +$46M in minutes; the HYPE hunter; 0xa31's $2.3M of funding in three days; "The White Whale" +$30M in a week longing ETH/SOL; HLP took the JELLY, ETH-whale and Jin liquidations onto its book and 44% of its lifetime profit from two crash days.
The "six whales" recap (EmberCN/Lookonchain, Nov-25): a "rolling long whale" $125k → $43M (344×) → liquidated; an "insider whale" $3M → $26M → zero; a "100%-win-rate whale" $30M principal, $26.6M peak, liquidated after sizing up — average loss >$40M per whale.
James Wynn (A1) — "Long $1.25B of BTC at $105–108k, liquidated at $104k, paid 160% APR funding for the privilege. BTC: $126k four months later."
Qwatio (A5) — "Short BTC at $117.8k, liquidated at $122.8k by a $123.1k wick. Twenty-four hours later BTC was below his entry; three weeks later 5% below."
XPL pre-market (C3 + B2) — "Shorts were right by 85–95% — if they survived the wick. They were still wiped by a 200% two-minute wick, then charged 254–800% APR to stay in. '1x hedge, account destroyed.'"
The ZEC max-short (B1) — "$536k of funding in five to six weeks (11 Oct–19 Nov), $22M drawdown, $6M liquidated. ZEC then fell 70% — exactly the thesis."
10–11 October 2025 (A6 + D1 + D2) — "$19B liquidated, 1.6M accounts, 35,000 ADL events; a $79M ETH long closed at the wick, ETH +24% in 72 hours; Wintermute (ADL'd on Binance): shorts ADL'd 'at $5 when the market was $1'. "
HYPE shorts (C4 + B6) — "A whale spent $14.5M to push funding to −800% and liquidate $43M of shorts; HYPE fell 37% in eight weeks. In 2026 loracle.hl ate a $46M loss at $70; HYPE was $52 two months later."
Pair any of these with E (the $192M tariff short, Abraxas' $760M book) for the zero-sum slide: the victims supplied the winners' liquidity. Not verified for this brief and best left off slides : TRUMP and ASTER funding extremes, Bitget VOXEL and MEXC wick cases, the exact Aug-2024 and Feb-2025 liquidation totals, the "$203M single ETH liquidation" on 10 Oct, and Garrett Jin's Jan-2026 figures.
5.5 Why perps win, and what liquidation-free products have achieved
Liquidity and simplicity. Perps show spreads 49–83% tighter and 5–8× more daily trades than dated futures; on stress days perp spreads widen 34% vs 161–249% for futures (Gornall, Rinaldi & Xiao). One instrument, one number to watch, no expiry, no Greeks.
Scale gap. On-chain, options DEX volume is ≈ $2.5B/30d against perp DEX volume of $370–600B/30d — options are well under 1% of on-chain derivatives flow.
Exchange incentive. NYU Stern estimates ~$296k of fees per 10,000 trades at 125× "regardless of trader outcomes"; liquidations are a revenue line (HLP's two best fortnights ≈ 44% of its lifetime profit).
Liquidation-free / defined-risk product
Claim
TVL now (9 Sep 2026)
Peak TVL
f(x) Protocol (xPOSITION, ≤ 7×)
No liquidation ("Liquidation Brake" rebalancing), no funding
$122.1M
$271.3M (Aug-25)
Derive V2 (on-chain options)
Defined risk
$158–159M
$164M (Aug-26)
Index Coop leverage tokens
No liquidation
$14.4M
$550M (Nov-21)
Toros leveraged tokens
No liquidation
$7.7M
$114M (Nov-24)
Opyn Squeeth (power perp)
No liquidation for longs
$0.6M
$27M (Apr-22)
TLX Finance
Leveraged tokens
≈ $0
$13M (Jul-24)
Read: real but small. f(x) is the only liquidation-free leverage product above $100M and its $122M TVL (deposited collateral, including the fxUSD/stability-pool side — not open notional, so not like-for-like) is under 2% of Hyperliquid's average 2025 OI ($7.4B unverified — no OI series is saved in the bundle). The demand signal sits in trader behavior (half of positions closed within an hour; a quarter of wallets losing >85% in a month; a majority of active accounts liquidated within months) more than in adoption of the existing alternatives — which either rebalance (f(x)'s black-swan mode), decay (leveraged tokens) or require options literacy (Derive).
Prediction markets — size, traders, and how much of it is real
The measurement trap
Three "volumes" circulate and differ 2–4×: notional (contracts × $1 face — Kalshi, The Block, most Dune dashboards), dollar (contracts × price — DefiLlama "volume", Polymarket event pages), and naive on-chain aggregation that double-counts CTF mint/burn legs (one study: $958M "naive" vs $391M exchange-equivalent for the Oct-2024 Trump market). Kalshi's own $23.8B for 2025 matches the notional series exactly, so "Kalshi volume" in the press is notional. Both bases are shown below.
6.1 Size and venues
Monthly notional volume, 2026 — Kalshi vs Polymarket (international + US), DefiLlama notional basis, $B. The Block's monthly sums run slightly lower (Jul $50.6B as first reported, ≈ $53.0B on The Block's revised Kalshi July of $40.1B; Aug $45.3B). Monthly series transcribed from the DefiLlama API on 9 Sep 2026; pull not saved.
KalshiPolymarket (intl + US)
2026 monthly, $B notional
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Kalshi
9.45
10.32
13.21
14.65
17.85
32.09
41.24
39.85
Polymarket international
7.66
7.94
10.57
9.01
7.08
10.70
7.89
4.59
Polymarket US
0.28
0.26
0.70
1.26
1.77
4.43
4.79
3.82
Kalshi, dollar basis
—
—
—
—
—
9.4
12.4
11.4
Venue
Volume
Open interest
Users
Valuation / economics
Status
Polymarket (on-chain, Polygon; plus Polymarket US, a CFTC DCM/DCO)
2024: $9.0B notional ($4.4B dollar); 2025: ≈ $21.5B notional ($10.5B dollar; The Block's $44B combined implies ≈ $20B); 2026 YTD to 8 Sep: intl $66.5B + US $18.5B = $85B notional ($37B dollar). Records: Mar-26 month $10.5B intl / $11.2B intl + US (same data at two scopes; DeFi Rate's $12.2B is a separate series) — but on the DefiLlama series June 2026 is higher at $10.70B intl / $15.13B intl + US, $425M day (28 Feb 2026). Aug-26 −35% MoM.
intl-only ATH ~$510M (Nov-24); $457M (Mar-26); combined ~$550M (start Jul-26) → $422M (end Jul-26); now intl $354M + US $128M = $481M
Monthly actives 314k (Dec-24) → 478k (Oct-25) → 733k (Mar-26) → 643k (Apr-26); 1.7–2.5M lifetime addresses; ~30% of offshore volume estimated to come from US traders
ICE up to $2B (Oct-25, ~$9B post); $15B (Apr-26); $1B at $21B led by 1789 Capital (Sep-26). Revenue $0 in 2025 (fee-free) → annualized >$1B (Jun-26) after 2026 taker fees. POLY token confirmed "in principle", not launched.
US: QCEX bought for $112M (Jul-25), CFTC amended designation 25 Nov 2025, politics added Apr-26, waitlist lifted May-26, "Combos" parlays >20% of US volume. Polymarket Perps launched 3 Sep 2026 (20×, hourly funding, US/Canada blocked).
Kalshi (off-chain, CFTC DCM + DCO)
2024 ≈ $2.0B; 2025: $23.8B notional (+1,108%; $10.4B dollar); 2026 YTD: $148B to 8 Aug, $173B to ~28 Aug (company), $191.9B notional / ≈ $75B dollar unverified to 8 Sep (DefiLlama; the dollar figure is not reproducible from the monthly dollar figures) . Records: $41.2B month (Jul-26), $10.17B week (w/e 30 Aug 2026), >$1B days in June (87.9% sports + parlays). Kalshi platform volume during the tournament window ≈ $27B (World Cup contracts $12–14B per CNBC/DeFi Rate); parlays ~50% of volume vs 2% before.
≈ $1.0–1.07B now; ~$1.4B (1 Jul) → $788M (31 Jul); Crypto Briefing: ~$2B World Cup peak
Active traders 240k (Dec-24) → 1.2M (Dec-25); +3M new users in the World Cup; 1,000+ institutional clients; 18–21-year-olds = 3.1% of volume
$2B (Jun-25) → $5B (Oct-25) → $11B (Dec-25, Paradigm) → $22B (May-26, $1B Series F, Coatue) → talks at $40B (Sequoia/Wellington, Aug-26). Fee revenue 2025 $263.5M (≈ 1.1% of notional; 89% sports); annualized $4B by Jul-26.
Sports ≈ 88% of Aug-26 volume; 15-minute crypto $4.8B in July (12%) unverified (KalshiData via BitRss; no primary source); "Timeless" crypto perps $43.3B cumulative unverified (Kalshi's own homepage figure, 9 Sep 2026) since launch on 3 Jun 2026 (announced 27 Apr; CFTC approved BTCPERP 29 May; CME sued the CFTC over that approval). 9th Circuit ruled 3–0 against Kalshi vs Nevada (28 Aug 2026) → circuit split, Supreme Court expected 2027; NY AG suit seeks >$36B.
Robinhood (broker, routes to Kalshi/ForecastEx/Rothera/OG.com)
2025: 12B+ contracts; Q1-26 8.8B; Q2-26 13.6B, $156M revenue (more than equities $129M or crypto $100M); 30B Jan–Aug 2026
—
>1M customers have traded event contracts
~1.15¢ revenue per contract; equity stake in OG.com (Sep-26)
Susquehanna is its Rothera partner; >50% of Kalshi volume once came via Robinhood
Opinion: $10.1B notional in 2025 (points-driven), 30% of all notional in the first week of 2026, then −90% to $0.75B in Aug-26; predict.fun $6.2B notional YTD (peak $647M/month); Limitless $0.86B YTD; HIP-4 (May-26) claims its first BTC 15-min market did ~3× the equivalent Polymarket+Kalshi volume
Opinion backed by YZi Labs; predict.fun by YZi Labs and Susquehanna Crypto
Volume tracks incentive programs, not users
Betfair Exchange (legacy off-chain reference)
£84B matched in 2025 (+~10%) unverified (sole source: a Covers trade-press article on the Betfair Predicts beta; no Flutter/Betfair statement cited)
—
—
Flutter
"Betfair Predicts" beta Apr-26
2025 total
$44B
Kalshi + Polymarket notional (The Block); ~$63B including Opinion and the long tail (DefiLlama); ≈ $26.7B dollar basis.
2026 YTD (1 Jan – 8 Sep)
$277B
Big three notional (Kalshi 191.9 + PM intl 66.5 + PM US 18.5); ~$301B all tracked; ≈ $109B dollar basis unverified (not reproducible from the monthly dollar figures; components sum to $112B). Record month Jul-26 $50.6B as first reported by The Block (≈ $53.0B after The Block revised Kalshi's July to $40.1B; the −14.5% August drop is against that revised base).
On-chain share, Aug 2026
12% / 20%
Notional / dollar basis. Off-chain (Kalshi + Polymarket US) = $43.7B notional vs ~$6.1B on-chain. A year earlier the split was roughly 40/60 on-chain — the flip is entirely Kalshi's sports growth.
Open interest now
≈ $1.56B
Kalshi $1.07B + Polymarket intl $354M + US $128M (DefiLlama, 9 Sep); DeFi Rate $1.5B. Combined ATH ≈ $2B at the World Cup.
400k+ wallets; $20B on-chain World Cup volume before and during the tournament; ~$5.7B wagered during the five-week tournament itself, when World Cup markets were ~63% of prediction-market activity
Chainalysis
Estimate — real people
~6–7M people have traded a real-money prediction contract at least once by mid-2026 (Kalshi ≥ 4M incl. the World Cup cohort; Robinhood >1M; DraftKings 0.6M; Polymarket ~1–1.5M humans after removing bot/sybil wallets; components sum to ≈ 7.3M, and Robinhood's users largely trade via Kalshi so they overlap rather than add). Monthly actives plausibly 2–4M across venues estLow
Derived
How long positions stay open
Capital duration (Polymarket, Apr-26; 1,843 markets, $348M):79% of capital held under 30 days, over 40% under 5 days; 21% "patient capital" ($73.6M) held ≥ 30 days, $19M > 90 days — geopolitics, the 2028 presidential cycle, season winners. $82.75M sits in markets paying Polymarket's 4% holding yield.
Trader tempo (Pew, 11,989 wallets, May–Jun 2026): median 46 trades in six weeks, average trade $6.50, median 10 active days of 42; sports specialists 69 trades, crypto 59, politics 13.
Market durations: crypto 5-minute / 15-minute up-or-down (Polymarket 5-min BTC since 12 Feb 2026; Kalshi 15-min), daily and weekly price ladders; sports hours–days (≈ 88% of Kalshi volume); politics and geopolitics months–years ("Democratic Nominee 2028" $1.27B and "Republican Nominee 2028" $701M of volume, still open).
Short-dated share: Kalshi — sports 75–88% + 15-minute crypto ~12% → roughly 85–95% of volume resolves within days. Polymarket — cumulative since Jul-24: sports 39%, crypto 20%, politics 32%; among crypto events ending in Aug-26, 5-minute BTC up/down windows were ≥ 54% of volume (1,822 windows, median $79k each; top-2,100-event Gamma API sample of $288M, dollar basis — smaller windows are excluded by the API cap, so the true share is higher) , daily multi-strike ladders 19.5%, weekly/monthly hit-price ladders 17.5%.
Liquidation — none, and the three closest analogs
Fully collateralized. Polymarket mints YES+NO pairs against $1 of collateral (Gnosis conditional-token split/merge — the same tranche split/merge arithmetic as StrikeSimple's primitive 2); Kalshi contracts are fully collateralized at its clearinghouse; the CFTC's Aug-2026 roadmap explicitly addresses "fully collateralized event contracts". There are no margin calls on binary event contracts at either venue.
Analog 1 — expiring worthless. Every binary sends one side to zero; no venue publishes the share of contracts that do. Proxies: 84.1% of Polymarket addresses are net-negative; Kalshi retail takers have lost $583.5M cumulatively (Roosevelt Institute; disputed by Kalshi); ~70% of closed Polymarket markets had under $10k of volume (CNBC) — thin markets where positions ride to zero.
Analog 2 — forced or contested resolution. Polymarket disputes go to UMA's optimistic oracle; $972M of volume has traded in disputed markets; nine anonymous UMA whales dominate dispute votes; the $80M "Strategy sells BTC by 31 May" market resolved No after two rounds (607 voters, 98.6%), two whale wallets earning $299k and $370k from the vote. Kalshi refunded the $54M Khamenei contract at last price under a "death carveout" and is being sued over it unverified (sole cited source is Forbes' Kalshi company-profile page).
Analog 3 — leveraged prediction products. Both incumbents added leverage on prices, not on event shares: Polymarket Perps (Sep-26, 20×, hourly funding, liquidations) and Kalshi "Timeless" crypto perps (live 3 Jun 2026 after CFTC approval on 29 May, $43.3B cumulative per Kalshi's own homepage unverified, margin via an FCM — its hidden margin-API demo explicitly excludes event contracts and parlays); Hyperliquid HIP-4 binaries live on a margin DEX. No protocol offering margin or looping on Polymarket shares was found. The reason is structural: any leverage L > 1 on a 0/1 claim creates a guaranteed adverse-outcome shortfall; stress tests on 13,298 Polymarket markets (the analysis sample passing adequacy gates, from 61,087 ingested; 21–27 Apr 2026) found dynamic margin only cut drawdowns 5.1% and staged halts made terminal shortfalls worse (Nechepurenko, arXiv 2605.10400). A bounded scalar payoff is a far more natural substrate for leverage than a binary.
6.3 Bots, wash trading and arbitrage vs real users — and who makes money
Wash-trading study
Scope
Finding
Columbia Business School (Sirolly, Ma, Kanoria, Sethi; SSRN 5714122, Nov-25)
> 2 years of on-chain data through Oct-25
≈ 25% of all historical volume; ≈ 60% of weekly volume in Dec-24; > 90% in some sports/election weeks; one cluster of 43,000 coordinated wallets; wash traders earned nothing → motive is airdrop/ranking farming
Chaos Labs / Inca Digital (Fortune, Oct-24)
2024 presidential market
≈ 1/3 of volume and ≈ 1/3 of users; part of the gap ($2.7B reported vs $1.75B on-chain) is a share-vs-dollar counting artefact
Solidus Labs (Dec-25 – Feb-26)
Politics markets
≈ 15% of volume in some markets consistent with self-trading / economically neutral positions
Dubach (arXiv 2604.24366, Apr–May 2026)
600-market panel, 52 days of tick data
Self-counterparty wash share median 1%, 22% upper tail — far lower after 2026 taker fees
Opinion (BNB), DefiLlama
2025–26
Volume fell ~90% after its points program ended (Jan $3.28B → Aug $0.37B); DefiLlama excludes wash-flagged wallets
Kalshi
—
No credible wash-trading study; quadratic taker fees make it uneconomic. The WSJ "fake trades" story (Jun-26) concerned Polymarket influencer marketing, not order flow
Bots, market makers, arbitrage
Concentration (Polymarket fills, 25–28 Apr 2026 — a 2.3-day, legacy-CTF-exchange-only block range; negative-risk markets absent; 13.4M fills, 77,204 addresses with ≥ 5 records):12.6% of addresses control 81.4% of notional; 2,952 high-frequency operator addresses (3.8%); 68 addresses (0.09%) = 28% of notional; 82% of addresses are episodic, low-notional (Nechepurenko).
TRM Labs (Jan–Mar 2026): high-frequency market makers (> 10,000 fills) = 35.2% of trades; mid-frequency 44.7%; first-time bettors < 0.2%; median bet $30 (first-timers) vs $12 (HFT).
Kalshi (Roosevelt Institute; 400M+ trades, $32B matched-trade dollar basis — not notional, Jul-21 – May-26): retail takers lost $583.5M to professional makers (Kalshi disputes the study, saying it counts institutional market-maker flow as "ordinary users" and casual app trades as "professional"); 63% of retail revenue came from 6% of matched orders (≥ $200); median loss rate 8% of amount wagered (vs ~5% at sportsbooks). Kalshi runs an in-house market maker, Susquehanna is Robinhood's Rothera partner, maker rebates up to 1%; institutional volume +800% in six months; DraftKings estimates 80–90% of prediction volume in sportsbook states is professional. Kalshi's public trade data went behind a $40k Dune paywall on 15 May 2026.
Documented arbitrage: $40M extracted via within-market rebalancing and combinatorial arb (Saguillo et al.); $1.12M via NegRisk basket arb; NBA in-game mispricings persist a median 3.6 seconds; arb half-lives fell from hours (early 2024) to under a minute (Oct–Nov 2024); Polymarket quotes lag Binance by a median 347 ms yet a 43-feature model found no tradable edge in 15-minute BTC markets; Polymarket vs Binance-implied probabilities differ by 5.6–6.3 pp (11 pp vs Deribit) on a single Sep-2023 BTC contract (214 hourly observations; Portnaya), most at low probabilities and long maturities.
Settlement manipulation: 821 manipulator-profile traders (1 in 300 of 243,000 participants) extracted $8.2M from 5-minute BTC contracts, 93% from retail — absent in 15-minute contracts (Dai, Jia & Yu). Moving a prominent market enough to get it cited in the news costs ~$0.7–1.0M.
Bots / market makers / professionals ≈ 65–80% of notional; wash 1–25% depending on period
Kalshi (off-chain)
Overwhelmingly retail accounts (millions)
A professional counterparty on one side of most trades; DraftKings' 80–90% "pro" estimate for regulated sports flow
PnL distribution
Study
Sample
Result
Sergeenkov / The Defiant (Apr-24 → Apr-26)
2.5M Polymarket addresses
84.1% in the red; 16% profitable; only 2% ever made > $1k; 840 addresses (0.033%) > $100k; 0.015% earn ≥ $5k/month four months running; of 6,600 wallets averaging $5k+/month, 2.6% stayed active over a year
DeFi Oasis (Dec-25)
1.7M addresses, 124M trades
~70% lost; < 0.04% of addresses captured > 70% of $3.7B realized profit; 668 addresses above $1M = 71% of gains; 149 addresses lost > $1M
Pew (May–Jun 2026)
11,989 active wallets
Typical user ≈ break-even: 56% lost (9% lost > $1k), 44% gained (7% > $1k); among 1,000+-trade users 33% lost > $1k, 27% gained > $1k
Solidus (politics, Dec-25 – Feb-26)
—
Top 0.55% of maker wallets and top 0.26% of taker wallets each ≈ 50% of gains; ~3% of traders drive most price discovery (attributed to an LBS/Yale study, not the Solidus/CoinDesk piece citation missing)
Kalshi (Roosevelt)
400M+ trades
~3× more losers than winners; median loss 8% of stake; no official disclosure
Who profits, in practice: "Magamyman" $553k on Khamenei's death ahead of the 28 Feb 2026 strikes; 38 fresh accounts with ~100% hit rates netted > $2M on the same strikes; a serial Iran-bets trader ~$1M; Master Sgt. Gannon Van Dyke indicted over the Jan-2026 Venezuela/Maduro-operation market; two people charged over Israel–Iran bets on classified information; Polymarket referred "dozens" of military-related cases to DOJ (Aug-26). Academic screens (Gomez-Cram et al.; Mitts & Ofir screen 210,000+ wallet-market pairs) classify 3.1% of accounts as "skilled winners" and flag ~1,950 "insiders" citation missing. Théo ("Fredi9999") made $85M on the 2024 election.
6.4 Structural notes — fees, ladders, and the scalar gap
Fee models converged on p(1−p). Polymarket international was fee-free through 2025 (2025 revenue $0); its 2026 taker fee is C × rate × p(1−p) with crypto 7%, sports 5%, most categories 4–5%, geopolitics 0%; makers never pay and receive 15–25% of taker fees (max taker fee at p = 0.5 ≈ 1.75¢/share on crypto). Polymarket US uses Θ = 0.06 × C × p(1−p). Kalshi: ⌈0.07 × C × P(1−P)⌉, max 1.75¢/contract, effective take ≈ 1.1–1.2% of notional ($263.5M on $23.8B), plus ~4% interest on idle cash. Robinhood earns ~1.15¢ per contract. Every incumbent charges most at 50/50 and least at the tails — a continuously priced range product does not map onto this curve and could undercut on tail-strike trades.
Range views are already expressed — as strips of binaries. Ladder-style events (price-hit ladders, "above X", counts, temperatures) were 295 of Polymarket's top 2,100 closed events and $4.0B of $55.5B (7.2%) of event volume. Largest: "What price will Bitcoin hit in 2025?" $188.8M across 30 strikes; monthly BTC ladders $107.7M (Jan-26), $121.3M (Feb), $99.6M (Mar), fading to $25–35M by Jul–Aug; ETH 2025 $75.9M; WTI crude ladders $60.6M (Apr-26); "How many Fed rate cuts" $31–52M per year ($49.4M 2024, $31.4M 2025, $51.7M 2026); Elon tweet-count ladders $30–47M per week at peak. Open: "What price will BTC hit in 2026?" $64.3M. Users express a range view through 20–50 binaries per event that must then be arbitraged into consistency (the $40M "combinatorial arbitrage", the NegRisk adapter, the 5.6–6.3 pp option-implied wedge measured on a single Sep-2023 BTC contract) — a native scalar tranche is what those ladders approximate, without the YES+NO ≠ $1 leakage.
Crypto-price share is rising, and the shortest horizon is the manipulable one. Polymarket ~20% crypto cumulatively; daily crypto records $176M (Polymarket) and $108M (Kalshi) on 2 Jun 2026 (Artemis; Kalshi's 14 Jun day at 10.4% crypto of $1.24B ≈ $129M already exceeded it); Kalshi 15-minute crypto $4.8B in July unverified (KalshiData via BitRss); Kalshi crypto perps $43.3B in about three months (Kalshi's own homepage figure); Polymarket 5-minute BTC windows ≥ half of its crypto event volume in Aug-26 (top-2,100-event Gamma sample, dollar basis). The 5-minute market lost $8.2M to settlement manipulation; the 15-minute did not — a design lesson on settlement windows for any price-referencing product.
Demand signals for bounded range payoffs: (1) BTC/ETH/WTI ladders sustain $25–190M per event; (2) both incumbents launched leveraged perps rather than leverage on binaries, because binaries cannot be safely margined — leaving capital-efficient, bounded price-range exposure unserved; (3) ICE distributes Polymarket data and prediction-market ETFs (Roundhill, GraniteShares, Bitwise) are queued at the SEC; (4) the CFTC roadmap carves out "fully collateralized event contracts" for lighter reporting — a fully collateralized scalar tranche fits that lane; (5) Polymarket's own pUSD collateral and CTF split/merge architecture make USD-collateral tranche minting the accepted on-chain standard.
What the evidence supports, what it doesn't, and which numbers to keep off the slides. Each point states the evidence first and the caveat last.
1. The perp pain point is real and now quantified — but most perp flow is intraday
Evidence: 53.6% of active Hyperliquid accounts liquidated within ~4.5 months (64% of the retail tier); 14–25% of wallets profitable; a BTC long paid 24% of notional in funding in 2024; $19B liquidated in one day with the price back within 72 hours.
Caveat: the median position lives 57 minutes and 89% close within a day. A settle-at-maturity claim does not serve that flow. The addressable slice is the ~11% of positions (larger by notional — weighted median 2.5 h, p90 27 h) held overnight or longer, and the holders who were right and got wicked. Size the wedge on that slice, not on total perp turnover.
2. "No funding, no liquidation" has a dollar value — state it as a range
Evidence: BTC longs paid 10.6–24.2% of notional per year (2025/2024); HYPE 22%; on Hyperliquid alone ≈ $0.95B of funding changed hands in 2025; market-wide $10–31B est . The ZEC, XPL and HYPE stories show funding and liquidation lines, not the thesis, deciding outcomes.
Caveat: a prefunded claim replaces funding with upfront time value and caps upside per the tranche design — the honest comparison is total cost of holding a correct view for N days, side by side, not "free vs paid".
3. Crypto options headroom is real; DeFi options is not where it is being filled
Evidence: options are 2–3% of crypto derivatives volume vs. options dominating futures in US equities; BTC options OI already exceeds futures OI. But the growth went to IBIT ($27–33B OI), CME ($9B) and Bullish ($6B), while all of DeFi options is $2.5B/30d, ~89% of it Derive with 24,604 lifetime traders and ~235 daily actives.
Caveat: the on-chain retail options user base is tens of thousands, not millions. The wedge is a simpler payoff product for perp-native users, not a better options exchange — the memo's "primitive plus app" framing matches the evidence better than an options-venue framing.
4. Prediction markets validate the fully collateralized split/merge model at scale
Evidence: $277B notional YTD across the big three with zero margin calls; the CFTC's Aug-2026 roadmap explicitly carves out "fully collateralized event contracts"; users already express range views as $25–190M ladder strips per event; both incumbents chose price-perps over leverage on binaries because binaries cannot be safely margined — a bounded scalar can.
Caveat: 65–80% of on-chain notional is bots and professionals; 84% of Polymarket addresses lose; the 2026 growth is Kalshi sports, off-chain. The validated demand is for crypto-price ladders specifically — cite those volumes, not total prediction-market volume.
5. Settlement design is a first-order risk, not a detail
Evidence: 5-minute BTC markets lost $8.2M to settlement manipulation while 15-minute ones did not; JELLY, XPL and the Binance-only USDe print show single-venue marks can be moved; $972M of Polymarket volume sits in disputed markets; Oct-10 cross-venue dislocations reached 10%.
Caveat for the execution-based settlement idea: an averaged achieved-price settlement inherits the liquidity of the venues it swaps across at the settlement block — the Oct-10 data says that liquidity is thinnest exactly when settlement matters most.
6. On-chain liquidity concentrates in one venue per category
Evidence: Hyperliquid = 62% of DEX perp OI and 81% of its volume from 100 addresses; Derive = 85–90% of DeFi options; Polymarket ≈ 75% of on-chain prediction dollar volume. New options liquidity (Rysk $52M total TVL, ≈ $45M of it on Hyperliquid; Derive ≈ $52M, its Hyperliquid-collateral slice of $158M TVL; Hypersurface) is forming on Hyperliquid L1/HyperEVM, and HIP-3 builder markets are already 26% of Hyperliquid OI.
Caveat: distribution, not product, decided every one of these categories. Venue choice for launch is a go-to-market decision with evidence behind it.
7. TAM framing that survives diligence
Bottom-up, with sources: ~10–30M active crypto derivatives traders monthly (est.); ~0.3–1M monthly on-chain perp traders; 2–4M monthly prediction-market users, of whom the crypto-price segment is ~20% of Polymarket volume; on-chain options users in the low tens of thousands. Present SAM as on-chain perp traders + crypto-price prediction traders, with the ranges and their confidence levels shown.
Caveat: no exchange publishes derivatives-user counts — say so on the slide rather than let a diligence analyst find it.
8. Numbers to keep off the slides
Aster's "45.9M users" (untraceable to Aster; its own figures are 2M–7.9M) and any incentive-era DEX volume (Aster, Lighter, ApeX) as organic.
Coinglass liquidation totals as "losses" — realized losses were only a fraction of the $19B headline (5–15% unverified: ratio not derived in §5; the one venue data point, Hyperliquid's $1.23B lost on $10.3–15.1B, gives 8–12%).
DeFi options notional (use premium); Kalshi's "$3.3B day" (marketing claim, not carried in §6); the "26% DEX share" figure and CME's "$207B OI" (that is FX) — neither figure appears in this brief (body DEX-share peak is 21.6%).
Anything tagged unverified in §5.4.
§8
Conflicts & confidence
Topic
Conflicting figures
Resolution
Conf.
CEX perps volume basis
The Block 15 venues: 2025 $64.9T · Coinglass/CoinGecko all venues: $85.7T / $86.2T
Scope (MEXC, BingX, WhiteBIT, CME, Coinbase etc.); several small venues show implausible volume/OI ratios. Use The Block/CCData for "credible", Coinglass for "reported".
High on ratios
Hyperliquid volume
Own feed $840B (Aug-25), $8.4T cumulative · third parties $400B, $5.4T
Double- vs single-sided counting; all shares here use single-sided.
High
Perp DEX totals 2025
$6.38T (CoinGecko) · $6.91T (The Block) · $7.9T (DefiLlama)
Protocol scope; DEX/CEX ratio (3.4% → 10–13% → 11–13%) is the robust metric.
Med
Oct 10–11 2025 liquidations
$19.13B / 1.62M accounts (Coinglass; own caveat $30–40B) · Hyperliquid $10.3B (Coinglass) vs $15.1B (its feed) · wallets wiped 358 (CCN) vs 6,300 (Blockscope)
Direction certain, exact figures medium; cite Coinglass with its caveat.
Med
Perp trader profitability
14–25% (wallet-sample studies — Hyperdash's 1,000 is a leaderboard sample, ENVY is a 30-day window) · 40.7% (leaderboard)
Leaderboard is survivorship-biased; use 14–25% for retail.
High
Holding periods, leverage, funding totals
Own Hyperliquid sample only
High for Hyperliquid, medium as a market proxy (CEX retail used more leverage per Binance 2019). Market-wide funding ($10–31B) is an extrapolation.
Headcount ratio (10–50 : 1) is an order-of-magnitude estimate.
Low
Prediction-market volume basis
Notional vs dollar differ 2–4× (Kalshi Aug-26 $39.9B vs $11.4B)
Methodological; both shown.
High
Polymarket 2025 and Mar-26 record
2025: ~$21.5B / ~$20B / $28.9B (DefiLlama notional) / $10.5B dollar · Mar-26: $10.5B (intl only) / $11.2B (intl + US — same data, wider scope, not a conflict) / $12.2B (DeFi Rate)
Use ~$21B notional for 2025; DeFi Rate's monthly series treated as unreliable; June 2026 ($10.70B intl / $15.13B intl + US) exceeds March on the DefiLlama series.
Not obtained for this brief: OKX/Binance/Bybit/Bitget options volumes and user mixes; CME options-only ADV/OI for 2026; BIS FX/equity options notional; a 2025 market-wide liquidation total for 2024; The Block monthly series behind JS charts (only current windows verified — the DefiLlama monthly series in §6.1 was transcribed from the API on 9 Sep 2026, but the pull was not saved, so it is not reproducible from the bundle); Kalshi trade-level data (paywalled since 15 May 2026); Coalition Greenwich front-office headcounts; any exchange's derivatives-user count.
§9
Sources
Grouped by section. Where a figure is our own computation, the raw inputs are the Hyperliquid API/stats feeds and DefiLlama/CoinGecko APIs listed under "Perpetuals — behavior"; scripts and raw pulls are saved with this brief.