MARKET MAKING DEVELOPMENT
PROPELOO engineers market making systems — quote engine, inventory management, spread calculation, skew logic, hedging execution, and the risk controls that prevent a market maker from becoming an unintentional directional holder. Market making is a defined-edge strategy. The infrastructure that executes it must keep the edge intact under real market conditions.
Market making earns the bid-ask spread by continuously quoting on both sides of the market. The risk is inventory accumulation: if the market moves consistently in one direction, the market maker fills more on one side and builds up an inventory position. An unmanaged inventory position in a trending market turns spread income into directional losses. The market making system must continuously monitor inventory, skew quotes to attract flow on the over-held side, widen spreads when inventory risk is high, and hedge residual inventory via external venues when it exceeds tolerance. PROPELOO engineers the full market making stack — quote generation, inventory tracking, skew model, hedging execution, and the risk limits that halt quoting when market conditions become unfavourable.
Quote generation is one component. Inventory management is where the edge is protected.
Continuous bid/ask quote generation at configurable refresh rates (1ms–1s). Spread set as fixed, volatility-adjusted, or order-book-spread-referenced. Quote size by inventory level — reduce size as inventory risk grows.
Real-time inventory tracking: long/short position per symbol, average entry price, unrealised P&L, inventory age. Automatic skew: quotes shift toward the side that reduces inventory. Widening spreads as inventory approaches limits.
Residual inventory hedged on a reference venue (spot exchange, futures) when it exceeds tolerance. Hedge order execution optimised for cost — limit orders with timeout fallback to market. Hedge P&L tracked separately from spread P&L.
Spread components: base spread (target profit) + volatility premium + inventory skew. Inventory skew shifts mid-price toward the side that reduces position. Volatility premium calculated from realised or implied volatility.
Maximum inventory per symbol, maximum total net delta, maximum daily loss. Quoting halt on: price feed loss, excessive volatility (market emergency), inventory limit breach, kill switch activation.
Spread P&L vs inventory P&L attribution. Fill rate by side. Adverse selection detection (toxic flow identification). Quote-to-fill ratio. Competitor spread tracking.
Market making infrastructure is an adversarial system. Informed traders will fill your quotes when doing so is profitable for them — and costly for you. The system must detect and respond to adverse selection.
The most important control in a market making system is not the spread width — it is inventory skew. When inventory is long, mid-price shifts down (making the bid less attractive, the offer more attractive). This passively attracts selling flow without requiring active hedging. The skew model determines how aggressively this rebalancing happens.
Axiom: SKEW BEFORE HEDGE
Informed traders take from market makers systematically. Signals: consistent fills on one side, fills correlated with subsequent price movement, fill rate spikes. When adverse selection is detected, widen spreads or pause quoting until market conditions normalise. Taking losses from a toxic flow situation for longer than necessary destroys the edge.
Axiom: IDENTIFY TOXIC FLOW EARLY
Hedging on an external venue has costs: spread paid, fees, market impact. The decision to hedge residual inventory must weigh these costs against the risk of holding the inventory unhedged. For liquid assets with tight external spreads, aggressive hedging is cheap. For illiquid assets, hedging cost may exceed holding risk for small positions.
Axiom: HEDGE COST MUST JUSTIFY ITSELF
These choices define quote quality, inventory risk tolerance and hedging efficiency.
Impact: Volatility-adjusted spread with order-book floor. Fixed spread is only appropriate for testing or very simple implementations.
Impact: Continuous streaming for exchange-based market making. RFQ for OTC/institutional market making where large block execution is the primary use case.
Impact: Futures on the reference exchange for crypto market making — lowest latency and best liquidity. Model the funding rate cost if using perpetuals as the hedge instrument.
Impact: Gradual response is preferable — it allows the system to continue operating while naturally reducing inventory. Hard stop as a backstop for extreme situations.
Market making system for a crypto exchange — continuous quoting, inventory management, on-exchange hedging, spread analytics.
OTC quote engine for institutional FX or crypto — RFQ handling, two-way pricing, inventory management, hedge execution.
LP infrastructure for DEX liquidity provision — Uniswap v3 concentrated liquidity management, rebalancing strategy, fee harvest, IL tracking.
Multi-venue market making with centralised inventory management, cross-exchange position netting, best-venue hedging.
Market making for stablecoin pairs or FX with tight mean-reverting spread models, high fill rate optimisation.
Low latency where it matters; correctness everywhere.
Stack: Go / Rust (latency critical), In-memory spread model, Lock-free quote updates, Event-driven architecture, WebSocket quote push
Stack: In-memory position tracking, Real-time skew calculation, Volatility estimator (EWMA), Redis (state persistence), Alert on limit approach
Stack: Exchange REST + WebSocket, Limit-with-timeout pattern, Hedge P&L tracking, Slippage measurement, Position netting
Stack: Spread P&L dashboard, Adverse selection metrics, Fill rate by side, Inventory age heatmap, Competitor spread monitor
The most dangerous time for a market maker is a fast market — exactly when the kill switch must respond fastest.
The kill switch that cancels all quotes must execute in under 100ms, even under maximum system load. It must work when the quote engine process is unresponsive — implemented as a separate watchdog with direct exchange API access.
Trading keys have trade-only permissions, no withdrawal. Per-symbol or per-venue keys limit the blast radius of a compromise. Key rotation procedure that does not interrupt trading operations.
If the price feed fails (stale data, disconnection), the system must immediately cancel all quotes — not continue quoting on stale prices. Staleness detection with automatic quote cancellation is a required safety mechanism.
Daily and per-session loss limits that halt all quoting when breached. These must be independent of the strategy logic — enforced by a separate risk process that cannot be overridden by the quote engine.
Spread model, inventory limits, skew function, hedge trigger parameters, risk limits.
Bid/ask generation, spread calculation, quote refresh, inventory-adjusted sizing.
Real-time tracking, skew calculation, limit enforcement, alert system.
Hedge order logic, venue connectivity, slippage management, hedge P&L.
Kill switch, loss limits, price feed staleness detection, adverse selection detection.
Simulated market making with historical data, inventory simulation, P&L attribution.
Gradual volume ramp-up, monitoring dashboards, risk desk access, incident response.
Market making is a specific strategy: quoting on both sides of the market to earn the spread, with inventory management as the primary risk control. Algo trading is a broader category — any strategy executed algorithmically. Market making infrastructure has specific requirements (continuous quoting, inventory tracking, hedging) that a generic algo trading framework typically does not address well.
Yes. DEX market making (Uniswap v3, Curve) requires different mechanics: concentrated liquidity position management, rebalancing when price moves outside the range, impermanent loss tracking, fee harvest and gas cost optimisation. We build the management layer that monitors positions and executes rebalancing transactions.
Adverse selection detection monitors fill patterns: consistent fills on one side, fills correlated with subsequent price movement, fill rate spikes during low-volatility periods. When detected, the system widens spreads, reduces quote size, or pauses quoting temporarily. Long-term adverse selection from a specific venue can be addressed by restricting order flow from that venue.
Target spread depends on the asset, competition and exchange fees. We build the spread model to your parameters and instrument it to measure actual realised spread vs target. For crypto majors with tight competition, maker rebates often subsidise the strategy — the net economics are rebate minus adverse selection losses.
We implement stochastic inventory control models including Avellaneda-Stoikov and Ho-Stoll frameworks. The algorithm continuously shifts the reservation price and skews bid-ask spreads based on current inventory deviation from target, incentivizing opposing trades to mean-revert inventory automatically.
When spot inventory fills exceed risk thresholds, our hedging sub-module immediately fires opposing market or aggressive limit orders on perpetual futures venues (Binance, Bybit, Deribit). This delta-neutral hedging locks in the bid-ask spread while eliminating asset directional volatility exposure.
We engineer hardware and software kill-switches: WebSocket heartbeat disconnect detectors, max drawdown circuit breakers, exchange balance reconcilers, and anomalous fill rate throttles. If any safety boundary is breached, the bot instantly cancels all outstanding quotes within 5 milliseconds.
The real-time management dashboard tracks total trading volume, P&L attribution (spread capture vs inventory revaluation vs maker rebates), Sharpe ratio, fill ratio, inventory skew timelines, and latency profiling across all active exchanges and currency pairs.