How spread capture works
You are paid the spread for standing in the market. You lose it back when the market knows something you don't. Everything else is bookkeeping.
What a spread is
Every market has two prices: the bid (what buyers will pay) and the ask (what sellers will take). The gap between them is the spread. A taker crosses the spread because they want to trade now; a maker earns it by being the one who was already there. If NVDA's true price is 100.00 and you quote 99.90 / 100.10, then every buyer pays you 10 cents over fair value and every seller gives you 10 cents under it. Buy at 99.90, sell at 100.10, and you made the full 20-cent spread while holding nothing overnight.
The maker's ledger
Per fill, a maker's expected profit is:
E[pnl per fill] = half_spread − adverse_selection − fees − hedging_cost- half_spread — your quote's distance from your true-price estimate. You collect it from every uninformed fill.
- adverse_selection — the amount the price moves against you, on average, immediately after a fill. Informed traders only hit you when you are wrong.
- fees — gas and venue fees per fill. On an L2 these are small but never zero; they set the minimum viable spread.
- hedging_cost — what it costs to keep inventory near zero.
The craft is entirely in keeping the first number bigger than the middle two. Quote too tight and adverse selection eats you. Quote too wide and no one fills you and you earn nothing. "Slightly wider than the true price" is the whole strategy stated in five words.
Why "slightly wider than true" and not "tighter than the pool"
The naive read is to undercut the stale pool. Wrong reference point. The pool's quote is not a price, it is a leftover. Your reference is your own true price estimate, built from the primary close, correlated symbols, and whatever signal is live. You quote around that, wide enough to cover your uncertainty. Sometimes that means you are inside the pool's effective price and you win the flow; sometimes it means you are wider than the pool because you know something the formula does not — and the pool, not you, gets run over.
The loop
- Estimate. Compute a fair value and a confidence for the pair.
- Quote. Post bid and ask at fair ± half-spread, sized to your inventory caps. Wider when confidence is low.
- Collect. Fills against your quotes book the half-spread.
- Reassess. Every fill is information. Two hits on the same side in a row means the market is telling you your fair value is off. Move it.
- Refresh or pull. Quotes carry a TTL. Re-post them only after re-running the estimate. On any signal you cannot price — halt, headline, gap in a correlated symbol — pull first and think second.
Why this is the boring, reliable one
Spread capture does not need a directional opinion, a catalyst, or a narrative. It needs discipline: thousands of small fills, each slightly in your favor, with the tail risk managed by pulling quotes fast. Over a session the PnL looks like a payroll, not a lottery ticket. The failure mode is equally unglamorous — one stale quote through one headline can refund a week of spreads, which is why the intelligence lives in knowing when not to quote.