State of Crypto Market Making

Main Paper

WPRC SG [35] Crypto Market Making: A look, back into the future  Ervin (Herring Global) + Rongxin (WPRC) + Thanks Koopman     24 Feb 2026

Crypto market makes token markets tradable by continuously posting buy and sell quotes (~350 firms). BUT why do 52% of the community not trust them and post-Jan 2026 liquidity has collapsed (survey lo.tech), forcing industry toward transparency, enforceable contracts & new assets. Note:There’s too much to talk about - will discuss venues like HyperLiquid in the future!

Crypto markets evolved from thin, retail-driven chaos — BTC swings of +7,600% or -85% in a year were normal, spreads ran 50–200 bps, and a single exchange collapse (Mt. Gox) almost killed the market. Now: a professionally market-made system where BTC spreads now sit at 1–5 bps, volatility has halved, and $86T in annual volume flows through algorithmic quoting infrastructure run by a handful of firms like Jump, Cumberland, and Wintermute. But we introduced a new and arguably more dangerous fragility: markets are now structurally dependent on those same firms. If one rumored exit (Jump, August 2024) can wipe $314B in a single day, the risk didn't disappear — it just centralized. Layer on top $2.57B in suspected DEX wash trading, $123M Jump settlements, and academic estimates that wash trading inflates 50%+ of CEX volume. We now have a market that looks deep/efficient on the surface, but are reliant on a few powerful actors who provides this depth.

The real edge is moving from "liquidity promised" to "liquidity observable + contractible": dashboards + enforceable KPIs + shorter exit paths beat brand-name desks and reduce the option-convexity extraction vector baked into Option+Loan deals [1]. Key shifts define transition: (i) retainer+KPI contracts with real-time dashboards that make MM performance enforceable, (ii) on-chain CLOBs and vault-based strategies that make MM activity publicly auditable, (iii) expansion into structurally different asset classes (tokenized treasuries, binary prediction markets) where competitive moat is compliance infrastructure and probability modeling rather than raw capital/speed Note:We find in these markets, speed is king.

(i) App Developers (Drift, Phoenix, etc.): Want to build competitive, specialized exchanges (e.g., for retail or institutions) but need control over order flow sequencing (e.g., cancel priority, speedbumps) to manage liquidity, which is currently impeded by the shared, immutable leader schedule. (ii) Validators & Node Operators: Want to maximize revenue from block production and MEV while minimizing operational complexity. incentives align with adopting performance upgrades (e.g., BAM, DoubleZero) but could conflict if custom app rules reduce their extractable value or require significant new infrastructure. (iii) Market Makers: Want to provide liquidity without being picked off by toxic arbitrage; they benefit from conditional liquidity and faster inclusion to reduce "gap risk," but their need for predictability may conflict with takers' desire for low-latency access. (iv) Takers & Traders: Want lowest latency to capitalize on new information and best execution prices; their incentive to "pick off" stale quotes conflicts with MMs, creating a classic adverse selection problem that different market structures attempt to solve.

(i) Wintermute · 110 employees · London HQ, Singapore, NYC · $15B+ daily volume, $126B on-chain gold volume in 2025, $1.17M realized profit in a single week (Dec 2025, Nansen). Shifted to a 4% deal approval rate; lowest controversy among top-tier desks. Founder: Evgeny Gaevoy left Optiver/Wall Street in 2017, built a mining operation with personal savings, survived the 2018 bear market without external capital. Known for being publicly combative on manipulation allegations and maintaining explicit public distance from DWF Labs. (ii) GSR Markets · 275 employees · London HQ, Singapore, NYC, Dubai · Est. net revenue $150M–$200M (2025), driven by institutional OTC flow and treasury management for 300+ portfolio companies. Pivoted to a "crypto investment bank" model via GSR One; first crypto liquidity provider with both FCA and MAS licenses; acquired a U.S. FINRA broker-dealer (Equilibrium Capital). Partner: Josh Riezman — Deutsche Bank, Société Générale, Circle background; mined ETH on personal servers for initial capital. At Consensus 2025: "The integration of DeFi and CeFi is the future." (iii) Flow Traders (Euronext: FLOW) · 635 FTEs · Amsterdam HQ, London, NYC, Singapore · FY2025: €485.8M NTI, €133.6M net profit (down 16% YoY on tech spend), 41% EBITDA margin. Primary liquidity engine for the AllUnity MiCAR-compliant euro stablecoin; dominant in global ETF/ETP ecosystem. The only publicly traded pure-play MM — fully audited, lowest opacity in the industry. CEO: Thomas Spitz (from Jul 2025) — 20+ years at Crédit Agricole CIB; appointment signals pivot toward regulated tokenized asset infrastructure. (iv) DWF Labs · 90 employees · Singapore HQ, Dubai, Hong Kong · Est. trading profits >$100M (2025); deployed $250M+ into liquid funds, $25M into World Liberty Financial. Operates as a liquidity provider / VC hybrid simultaneously — invested in 400+ projects ($200M+) between 2023–2025. WSJ wash-trading allegations ($300M on Binance) remain unrefuted; no regulatory action taken. Managing Partner: Andrei Grachev — entered logistics at 18, traded traditional markets from 2014, made early ETH profits ($7 → $350), previously headed Huobi Russia. Quote: "As long as we operate within legal boundaries, if a method proves effective, we will adopt it without worrying about what others say." (v) Cumberland (DRW) · 100+ crypto / 2,000+ total · Chicago HQ, London, Singapore · Est. net income >$150M (2025); SEC unregistered dealer case dismissed with prejudice (Mar 2025); primary institutional bridge for USDT-to-CEX inflows; Goldman, Nomura, Circle partnerships. Regarded as the most institutionally orthodox U.S. desk, backed by DRW's balance sheet for high-volume arb and bank-sourced OTC. (vi) Keyrock · 170 employees · Brussels HQ, London, NYC, Geneva · Est. revenue >$25M (2025, up significantly YoY); pivoted to on-chain yield products and prediction markets via acquisition of fija Finance; dominant liquidity provider on Polymarket and Kalshi. MiCA filing in Liechtenstein; NYC entity established Mar 2025. Note: Market Makers is on a spectrum, also many are secretive - hard to know players. Also platform and ecosystem focused. Smaller desks in SG and other areas producing extreme returns.

Chooses a point on the triangle of (a) cost, (b) latency, and (c) control/transparency; mis-choosing shows up as slippage, blowups at launch, or governance/legal fallout. (i) Latency vs. transparency: CEX colocation offers speed but opacity is default; on-chain venues add block-time/MEV constraints but reduce "black box" behavior via auditability. Slower quote updates increase stale-order adverse selection; MMs respond by widening s or thinning D, degrading user experience [1]. (ii) Cost vs. control: Option+Loan looks "cash-cheap" but can be economically expensive via option value + launch sell pressure. Retainer is explicit cost but buys exit leverage + clearer accountability. Dashboards/telemetry add operational overhead but are the only scalable enforcement mechanism [1]. (iii) Centralization vs. fragmentation: Top 8 CEXs control 91.7% of global depth; Binance alone ~42% spot share [2]. Exchange listing dynamics can force projects into specific MM relationships and token allocations with limited visibility. MM firms bundling advisory/investment/MM services concentrate conflicts—any firm simultaneously acting as MM, investor, and advisor has irreconcilable conflicts [1]. (iv) Vault-based models (Hyperliquid): Protocols retain control while community LPs earn yield based on trading performance. Flips the traditional relationship but shifts risk to community LPs who may not price adverse selection correctly. Note: here are both market markers and vaults. Today most liquidity comes from participants in the books not vaults. (v) Standardized legal frameworks (CMIC standards, MiCA): Reduce negotiation friction and level the playing field for smaller projects, but may not capture crypto-specific dynamics like token loan mechanics.

1. The Lo:Tech survey:  60% of crypto community understands what MMs do, and order book liquidity as core function, but deeply distrusts how they do it. (i) 89% believe MMs affect token prices—technically true via spread management, but 37% sees coordinated price manipulation. (ii) 69% claimed to understand MM revenue, but <48% identified legitimate methods; top "answers" included insider access (31%), wash trading (38%), and backroom exchange deals (45%). (iii) 34% of respondents who held tokens in projects using MMs reported losing money; 48% experienced pump-and-dump outcomes. (iv) 70% support criminal prosecution of MMs.

Key events: (i) The FTX/Alameda collapse (Nov 2022): Alameda was running >20% of global volume on $10B in misappropriated client funds. When it failed, global market depth halved overnight. U.S. altcoin liquidity has still not recovered, sitting ~38% below pre-FTX levels in early 2026.

Feedback: Alameda vs. Wintermute: Wintermute was aggressively eating Alameda's maker flow with taker orders, contributing to Alameda's collapse. Then Wintermute got hacked, Jump pulled out of Solana entirely (also ate the $350M Wormhole hack). Nearly wiped out the major MM players simultaneously, causing an on-chain liquidity crisis. Only survivor was Aphelion Trading (Space Monkey). (ii)The FBI's Operation Token Mirrors (2024–25) confirmed what people already knew: wash trading was a standard commercial service, not a fringe practice — Gotbit's $23M forfeiture and CLS Global's $853K in penalties. (iii) MOVE (Apr 2025) was the option+loan model at its worst — a dual-role entity (Rentech/Web3Port) sitting on both sides of the contract loaned 66M tokens (5% of supply), dumped $38M day-one post-listing, and left the token down 98% from ATH - Binance banned the desk. (iv) MANTRA/OM (Apr 2025) exposed how thin liquidity can hide beneath a large reported market cap: $6B+ was wiped in a single hour via a collateral-borrow-inflate-liquidate loop that OKX called premeditated, a 90%+ drawdown that erased the illusion of depth entirely.  (v) 10/10 crash (2025) A tariff announcement triggered a market-structure failure more than a simple price crash. Whale short positioning began at 14:27 UTC, the macro shock hit at 14:57, and by 20:50 a mechanical liquidation cascade ignited. In the 40 minutes that followed, BTC perp spreads blew out from 0.20 bps to a peak of 26.43 bps, visible order book liquidity collapsed from $103.64M to $0.17M (a 98%+ wipeout), and liquidations accelerated from $0.12B/hour to $10.39B/hour — culminating in $3.21B liquidated in a single minute at 21:15. Open interest dropped 27.5% ($60B) in roughly 25 minutes. The cascade was amplified by three compounding failures: (1) market makers' synchronized volatility circuit breakers pulled quotes simultaneously, turning displayed liquidity into a mirage; (2) Binance's transfer subsystem degraded between 21:18–21:51, preventing users from re-margining across Spot/Earn/Futures as the cascade unfolded; and (3) unified/cross-asset margin mechanics meant venue-local collateral mispricings — most notably USDe printing at ~$0.65 on one venue while trading near $1 elsewhere — fed directly into oracle engines and triggered cascading liquidations of unrelated positions. The event permanently impaired post-event liquidity, with BTC 1% market depth remaining lower afterward, confirming that 10/10 wasn't just a volatility spike — it was structural proof that crypto liquidity is deeply conditional, and the very infrastructure meant to stabilize markets became the primary transmission mechanism for the crash.

1. A Market Maker’ Perspective: The Liquidity Regime Shift (Jan–Feb 2026): (i) Majors liquidity down ~40% comparing first two months of 2026 vs. last two months of 2025. (ii) altcoins materially worse; retail absent; desks cutting risk. (iii) Directionally consistent with Kaiko data showing U.S. altcoin depth already 38% below pre-FTX peaks. Mechanism: (i) Balance sheets tighten → fewer desks warehouse inventory through shocks → thinner books. (ii) Retail absence increases flow toxicity → higher adverse selection → makers quote wider. (iii) Selective liquidity → majors remain tradable, alts get punished harder on spread/depth. (iv) Capital reallocates to arenas with structural edge: RWAs (compliance moats) and prediction markets (microstructure + probability modeling on transparent rails) [3]. Feedback:  Dominant problem in Solana DeFi is toxic taker flow — same dynamic as HFT in TradFi. Currently a speed race (infra advantage), but Solana is building core protocol-level infrastructure to prioritize maker orders, which would shift competition toward intelligence/algorithms rather than raw speed.

2. Next Frontier: Prediction Markets and RWAs: 

Prediction Markets: Polymarket drove >$3B volume on the 2024 U.S. election alone [2]. A temporal latency arbitrage existed between spot confirmation on CEXs and Polymarket oracle settlement—near risk-free during the window; Polymarket patched oracle speed mid-2025; opportunity largely closed. Post-patch equilibrium shifts from "free latency arb" to true market making: inventory management, adverse selection, and event-time jump risk. Ongoing strategies: (i) continuous quoting on both YES/NO sides of binary markets earning spread from uninformed flow, (ii) Bayesian probability modeling as core edge vs. retail, (iii) cross-market hedging using CEX perps/options, (iv) specialization in crypto-native markets (governance votes, upgrade outcomes) which are underserved and mispriced. Regulatory note: prediction market MM is not permitted in Singapore under MAS's Digital Payment Token framework [3]. Expect continuous hardening—faster repricing loops, friction design to reduce pickoff risk, improved surveillance. Note: Thank you Herring Global for the perspective and Notes

RWAs: BlackRock's BUIDL ($500M+), Franklin Templeton's BENJI, Ondo Finance, and dozens of tokenized treasury products created a new asset class . RWA MM is fundamentally different from native tokens: (i) inventory risk bounded by redemption value (unlike native tokens that can go to zero), (ii) spreads reflect credit risk + redemption queue delays + smart contract risk, (iii) compliance is layered across jurisdiction of underlying asset, tokenization platform, and trading venue, (iv) oracle quality (reliable NAV feeds) is existential. RWAs inherit constraints crypto-native assets don't: whitelists/transfer restrictions, settlement mismatch, issuance/redemption windows, basis vs. NAV. "Good RWA MM" likely converges toward ETF/FX-like patterns: redemption rails, inventory financing, compliance workflows, distribution. The convergence play: firms bridging institutional compliance (SEC/MAS/FCA licensing, AML/KYC, segregated custody) with on-chain execution at DeFi speed. Fewer than 10 firms globally positioned today. Meaningful secondary market liquidity for non-stablecoin RWAs is realistically 2027–2028.

(i) Traditional token MM competes on capital and speed. (ii) Prediction market MM competes on probability modeling and oracle-latency awareness, access to custom data. (iii) RWA MM competes on compliance infrastructure and redemption-rail integration. The common thread: from "how much capital can you deploy" to "what structural advantages do you have that others cannot easily replicate."

Feedback: Three tiers: TradFi (ETFs, futures — legally required to have MMs), CeFi (CEXs), and DeFi (AMMs + order book DEXs). TradFi now includes crypto ETFs in US, Europe, Singapore, and HK. (i) AMMs still dominant, but private/prop AMMs (e.g. SoFi) now carry the lion's share of liquidity with no UI. (ii) Order book DEXs only succeed if they do their own market making — Hyperliquid proved this (founders came from a MM background). (iii) Innovative order types emerging, e.g. Manifest Trade's multi-pair shared liquidity orders. (iv) Auction-based systems (e.g. Drift's JIT, Jupiter's auction with Wintermute) are increasingly popular — MMs prefer them because they eliminate toxic flow risk entirely.

Feedback: On-chain = transparent = harder to be fraudulent (vs. CeFi where exchanges trade their own books, wash trade, etc.). Future is on-chain, intelligence-driven, and fairer — closer to the economist ideal of efficient capital markets.Pure CLOB market making doesn’t fit prediction markets well, because prices don’t move continuously they jump around news and oracle updates, and informed flow is extremely toxic in short-horizon events.

That makes resting quotes stale instantly, so market makers either get picked off or pull liquidity together, creating “liquidity mirages.”

A prediction native MM therefore needs a stateful microstructure on top of CLOB. they switching between continuous trading and short batch modes or rfq, with conditional quoting and auditable so liquidity is observable, contractible, and resilient at event time.Yes, Order placement speed isn't king in prediction market. King is getting information faster and updating probabilities more quickly.there are both market markers and vaults. Today most liquidity comes from participants in the books not vaults.Good pointAlso -added a note at the top about future discussion on HL. lol