P&L Attribution · 6 min read
Two traders can report identical daily P&L for opposite reasons. What a broker net number hides, and why decomposing it by greek changes your decisions.
Your broker shows one number per position: up $412, down $187. It is accurate, it is legally meaningful, and it is nearly useless for improving as a trader. The number tells you what happened to your money. It does not tell you why, and in options, where every position is simultaneously a bet on direction, time, movement, and implied volatility, the why is the entire game.
Every options profit and loss decomposes into contributions from the greeks: Delta (the underlying moved), Gamma (the exposure curved as it moved), Theta (time passed), and Vega (implied volatility changed), plus a residual the simple model does not capture. The broker's number is the sum. The components are where the information lives.
Take a calendar spread showing +$412 after a week. Attribution might reveal: Theta +$310, Vega +$180, Delta −$55, Gamma −$23. That is a healthy trade: the position earned its money from time decay and a volatility uptick, exactly what a calendar is built to do, and paid a modest, expected toll to the underlying's wandering. Now the same +$412 with a different anatomy: Delta +$540, Theta +$120, Vega −$248. Same headline, rotten interior. This trade lost money on its actual thesis, since the volatility position bled, and was bailed out by an accidental directional gain it never planned to take. Run that second version twenty times and the direction luck washes out, leaving the Vega bleed as your true, negative expectancy.
Figure: The second +$412.00, drawn: Delta +$540.00 and Theta +$120.00 bailing out a −$248.00 Vega bleed. Same headline as the healthy calendar, opposite lesson.
This is the core argument for attribution: the headline number cannot distinguish skill from luck, and unexamined luck is how bad habits get reinforced. A green week teaches you to repeat whatever you did, whether or not it worked for the reason you think. Across a real book the effect compounds. A trader might discover that over three months, path effects, meaning Delta and Gamma combined, cost $2,100 while Theta and Vega together earned $1,850, with commissions and spread-crossing consuming another $600. The broker's summary of that quarter is simply −$850. The attribution says something far more actionable: the volatility engine works, the directional leakage and the friction are the problems, and the fixes are hedging discipline and fewer, larger trades, not a new strategy.
The broker's number tells you the score. Attribution tells you whether you are winning the way you intended to win.
There is a second layer of ambiguity inside the number itself: what is the position worth right now? The broker marks to market, typically at the midpoint of each leg's bid and ask. That is a fine convention in liquid markets during trading hours. But options quotes widen dramatically outside regular hours and in stressed moments, and the midpoint of a wide market is closer to fiction than fact. A four-leg position marked off overnight quotes can legitimately show +$738 by summing leg midpoints, +$216 by the account-level feed, and −$28 by a model repricing every leg off a consistent volatility surface: three defensible answers to "what is it worth," spread across $766.
None of the three is wrong; they answer different questions. The mark approximates liquidation value if markets were orderly. The model estimates fair value under consistent assumptions, which is the right basis for judging whether your greeks performed. The gap between them is itself a signal: a persistent difference between where the market marks your position and where a clean model prices it usually points at a stale quote, a leg trading below intrinsic value on a wide market, or genuine information in the order flow. Traders who track only the broker mark experience these gaps as unexplained overnight P&L swings, money that appears at the close and evaporates at the open, and mistake quote noise for market movement.
Beyond drivers and marks, a few bookkeeping conventions routinely distort what traders think their number means. Realized figures are often net of commissions, so subtracting commissions again, a common spreadsheet habit, double-counts them. Unrealized profit on open positions is not the same as banked profit: a week showing +$1,040 that is entirely open marks on wide quotes is a screenshot, not a result, and it can be gone by Tuesday. And expiration-day settlement can move money between "realized" and "unrealized" buckets in ways that make a single day look far better or worse than the trade actually was. When reviewing performance, always separate three quantities: cash actually banked on closed trades, model-basis value of what remains open, and cumulative friction. Any summary that blends them is answering a question you did not ask.
You do not need an institutional risk system to escape the single number. Start with a journal habit: for every closed trade, record what you expected to earn from (the thesis greek) and what attribution says you actually earned from. Platforms that decompose each position's profit and loss into Delta, Gamma, Theta, and Vega buckets automate this, but even a rough manual version, repricing the position at entry-day volatility and comparing, separates path from vol from time well enough to catch the big lies. Then review by bucket, not by trade: after twenty campaigns, the pattern across buckets is your actual track record, and it is usually a surprise.
The broker's number tells you the score. Attribution tells you whether you are winning the way you intended to win, and in a game where luck can impersonate skill for months at a stretch, that distinction is worth more than the number itself.
Educational content, not investment advice. Options involve substantial risk.