3b. The Families of Trading Strategies

Before diving deeper into market making, it helps to see the whole map — where market making sits among the other ways people try to make money in markets. Almost every strategy ever traded belongs to one of a handful of families, and each family is really a single sentence: a bet about what price will do next. Knowing the families tells you three things at a glance — what a strategy assumes, what market it thrives in, and (crucially, after Chapter 2b) what data it needs.

1. Trend-following / momentum

The bet: "a move in motion tends to continue." Buy what's going up, sell what's going down, and ride it until it turns.

This is the oldest and most intuitive family. The classic implementation is a moving-average crossover (Chapter 2): when a fast EMA rises above a slow one, go long; when it falls below, go short. Breakout systems are momentum too — buy when price clears a prior high. Trend-following is convex: many small losing whipsaws paid for by a few enormous winners when a real trend arrives. It is the strategy behind the managed-futures industry (CTAs).

• Thrives in: strong, persistent trends — a bull run, a currency devaluation, a commodity shock.
• Dies in: choppy, range-bound markets — it buys every false breakout and sells every false breakdown, bleeding on costs.
• Data it needs: klines. Direction is all it reads. (This is the family the EMACrossBracketAlgo in the companion backtester belongs to.)

2. Mean-reversion

The bet: "a move stretched too far tends to snap back." Sell what's risen too fast, buy what's fallen too far, and profit from the return to normal.

The mirror image of momentum. Implementations fade extremes: buy when price pierces the lower Bollinger band, sell at the upper; short a stock that's spiked on no news. The bet is that prices oscillate around a fair value and that deviations are temporary. Statistically it's a bet on negative autocorrelation of returns.

• Thrives in: calm, range-bound, liquid markets that oscillate around a stable level.
• Dies in: trends and regime breaks — "the market can stay irrational longer than you can stay solvent." Fading a real trend is how mean-reverters blow up.
• Data it needs: klines suffice; tick data sharpens entries.

3. Arbitrage & relative value

The bet: "two prices that should be equal, aren't — yet." Buy the cheap one, sell the rich one, and collect the difference when they converge.

The purest family — in its ideal form, riskless. Variants span a spectrum of how "sure" the convergence is:

• Pure arbitrage — the same asset at two prices (BTC cheaper on one exchange than another). Nearly riskless, fiercely competed, won on latency.
• Triangular arbitrage — a loop of currency pairs whose rates don't multiply to 1.
• Statistical arbitrage / pairs trading — two correlated assets (e.g. Coke vs Pepsi) whose spread has widened; bet it reconverges. Not riskless — the relationship can break.
• Basis / funding trades — spot vs future, or perp funding on Hyperliquid/Binance (Chapters 11–12).

• Thrives in: fragmented markets and temporary dislocations. Its edge is speed and infrastructure, not prediction.
• Dies in: a slow stack (someone faster takes the trade first) or when a "should converge" relationship structurally breaks.
• Data it needs: often quotes / the book on multiple venues at once — you're comparing live prices, not history.

4. Market making — the flow bet

The bet: "I don't know where price is going, and I don't need to." Quote a bid and an ask, earn the spread as both fill, and manage the inventory that piles up when they don't.

This is the family this entire book is about, and it is categorically different from the three above. The others are directional or convergence bets — they profit from price moving (or two prices converging). Market making profits from providing liquidity: it's paid the spread as a fee for immediacy, regardless of direction, and its enemy isn't being wrong about direction but adverse selection (getting filled by someone who knows more) and inventory risk (the price running away while you're stuck holding). Those two forces are the whole of Part II.

• Thrives in: liquid, mean-reverting, two-sided flow — many small uninformed traders paying the spread in both directions.
• Dies in: trending or toxic markets — one side fills relentlessly, inventory piles up against the move, and informed flow picks off stale quotes. (The failure mode of naive quoting and grid bots, Chapter 15.)
• Data it needs: quotes and the order book — it lives inside the spread and cannot see it on klines. This is exactly why the trailing-stop backtest failed on kline data in Chapter 2b.

The honorable mentions

• Scalping — very-short-horizon momentum/mean-reversion on tiny moves; blurs into market making at the fast end.
• Carry — holding a position that pays you to wait (funding, dividends, roll yield). You're paid for bearing a risk, not for predicting.
• Event / news trading — betting on the reaction to a scheduled event (earnings, a Fed decision, a sports outcome — Chapters 8–9). For a market maker, events are when you widen or pull quotes, not when you predict.
• Grid trading — a lattice of buy/sell orders around a level; mechanically a market-maker-lite, and it shares the maker's fatal weakness to trends (Chapter 15).

How they fit together