Algorithmic Trading in Crypto: What the Algorithms Actually Do

"Algorithmic trading" covers two different jobs that get constantly conflated. Alpha algorithms decide what to trade: they hunt for signals and generate positions. Execution algorithms decide how to trade: given a decision to buy or sell a certain size, they work the order to get the best possible price. Retail conversation is obsessed with the first; institutional results are disproportionately determined by the second. This guide covers both, weighted the way a fund should weight them.

Execution algorithms: the institutional workhorses

An execution algorithm takes a parent order (buy 500 ETH) and breaks it into child orders placed over time and across venues, minimizing the market impact and slippage that a single large order would suffer. The standard suite:

TWAP (time-weighted average price). Slices the order into equal pieces over a defined window, one piece per interval, indifferent to volume. Predictable, simple, effective for patient flows and thin markets. Its weakness is predictability: naive fixed-interval TWAPs can be detected and traded against, which is why production implementations randomize slice timing and size.

VWAP (volume-weighted average price). Distributes the order proportionally to the market's expected volume curve, trading more when the market naturally trades more. The benchmark of choice when the goal is "fill at the market's average price without standing out." Crypto's 24/7 sessions make volume curves less stereotyped than equities' U-shape, so good VWAP engines model per-venue, per-asset patterns rather than importing TradFi assumptions.

POV (percentage of volume). Tracks realized volume live and keeps participation at a fixed share, say 10% of whatever trades. Adaptive by construction: fast in busy markets, quiet in dead ones. The right default when urgency is moderate and volume is unpredictable.

Iceberg. Shows only a small visible tip of the order on the book, automatically replenishing as it fills. Hides intention, avoids signaling size, and pairs naturally with maker-side execution to earn spread instead of paying it.

Smart order routing (SOR). The cross-venue layer: a consolidated view of order books across exchanges, with logic that splits each child order to wherever the best prices and depth sit at that instant. In a market fragmented across hundreds of venues, SOR is not an optimization, it is table stakes for institutional size.

Selection logic in one line: urgency high and size small, just cross; urgency low, TWAP or maker-side Iceberg; benchmark-sensitive, VWAP; volume-uncertain, POV; size large in fragmented liquidity, SOR under all of the above.

Alpha algorithms: the strategy side

The signal-generating families in crypto are recognizable from other asset classes, with local flavor:

Arbitrage exploits price differences: across venues, between spot and derivatives (the basis and funding trades at the heart of delta-neutral strategies), and across correlated pairs (statistical arbitrage).

Trend and momentum systems ride persistence in price moves; mean reversion systems fade extremes. Crypto's volatility gives both families plenty to eat and plenty of ways to die, regime detection is the hard part.

Market making algorithms quote two-sided prices continuously and manage inventory, the strategy family we cover in our liquidity providers guide.

ML-driven approaches, including reinforcement learning agents that learn execution or quoting policies from market feedback, have moved from papers into production at sophisticated firms, mostly as components inside the families above rather than as standalone magic.

The honest note on alpha: signals decay, backtests flatter, and the marginal returns of most public strategies are competed away. Execution quality, by contrast, is a durable edge available to anyone willing to build or buy the infrastructure, which is exactly why this guide is weighted the way it is.

The infrastructure underneath

Every algorithm above assumes plumbing: normalized connectivity to each venue, real-time consolidated market data, order lifecycle management that survives exchange API failures, pre-trade risk checks, and post-trade TCA proving what the algorithm achieved against arrival price. That stack is the subject of our guide to HFT and execution software for crypto, and it is where build-versus-buy decisions actually get settled, because maintaining it across venues is a permanent engineering commitment.

And there is a layer after execution that algo discussions always skip: every fill has to land in the books. Positions, cost basis, realized and unrealized P&L, NAV, all of it must reconcile across every venue the algorithms touch, or the fund is flying a fast plane with no instruments.

That is the integration Renesis is built around: institutional execution (TWAP, VWAP, Iceberg, POV, smart order routing) wired directly into a portfolio management system with reconciled positions, P&L attribution, and NAV across CeFi and DeFi. The algorithms trade; the books stay true.