P&L Attribution For A Market Maker

A market maker closes the day up $4,000. That single number is nearly useless on its own — it could be $20,000 of clean spread capture wiped down to $4,000 by a bad afternoon of adverse selection on one name, or it could be $4,500 of spread and rebates minus $500 of inventory drift, a completely different, and much healthier, business. P&L attribution splits the one number into the pieces that actually drove it, because the fix for "too much adverse selection" (quote wider, or stop making markets in that name) is nothing like the fix for "too much inventory drift" (tighten position limits or hedge faster).

The four buckets

$\text{Total P\&L} = \text{Spread P\&L} + \text{Rebate P\&L} + \text{Inventory P\&L} + \text{Adverse Selection P\&L}$

• Spread P&L: the difference between where you bought and where you sold, on matched round trips, holding the mid fixed. This is the "textbook" market-making profit.
• Rebate P&L: exchange payments for posting liquidity (or fees for taking it) — see Maker-Taker Fees. Often a meaningful fraction of total P&L on thin-spread, high-turnover names.
• Inventory P&L: mark-to-market gains or losses on whatever position you're carrying, purely from the price moving while you hold it — has nothing to do with how good your quotes were.
• Adverse selection P&L: the markout-based cost (see Markouts: Measuring Post-Trade Drift) of counterparties who were, on average, right about where the price was headed.

Worked example: one symbol, one day

A maker in a mid-cap stock does 8,000 round trips at an average captured spread of $0.015 and posts 16,000 shares' worth of maker volume earning a $0.0015/share rebate. Markouts at 10 seconds average −$0.008 per fill across 16,000 individual fills. Inventory ends the day 3,000 shares long, and the stock closed $0.20 higher than the maker's average holding price on the residual.

• Spread P&L: 8,000 × $0.015 = +$120
• Rebate P&L: 16,000 × $0.0015 = +$24
• Adverse selection P&L: 16,000 × (−$0.008) = −$128
• Inventory P&L: 3,000 × $0.20 = +$600

Total: 120 + 24 − 128 + 600 = +$616.

Look what happened: spread and rebates together made $144, adverse selection took back $128 of it, leaving barely $16 of genuine market-making edge — and the entire day's profit actually came from inventory, i.e., getting lucky on a directional position the desk happened to be carrying. Reported as a single $616 number, the day looks great. Attributed, it says the core market-making engine is nearly break-even and the desk got bailed out by an accidental long. Those call for very different Monday-morning conversations.

Where this feeds back

• Model tuning. A persistently negative adverse selection bucket in one symbol is the trigger to widen quotes, skew more aggressively (see Skewing Quotes To Manage Inventory), or drop the symbol.
• Risk limits. A P&L dominated by the inventory bucket is a sign position limits are too loose relative to what the strategy is actually designed to do — see Inventory Management for Market Makers.
• Compensation and evaluation. Desks and individual traders are usually judged on the spread + rebate − adverse selection components, not raw P&L, precisely to avoid rewarding lucky directional bets.

In interviews

If asked to explain a market maker's P&L, resist naming just "spread." List all four buckets, and be ready with the punchline: total P&L can look great while the actual market-making edge is flat or negative, and attribution is the only way to tell the difference.