TRADING / EVIDENCE INTO PRACTICE
Trading Expectancy: Why Winning Often Can Still Lose Money
Calculate trading expectancy after costs with two worked trade logs. Understand win rate, average loss, break-even win rate and the limits of small samples.
General education, not individual investment advice. Capital is at risk; examples are hypothetical. Read the disclaimer.
Calculate the average net outcome per trade, then inspect the largest losses and the assumptions behind the sample. Win rate alone does not describe an edge.
In this guide
You finish ten trades with eight wins. Your friend finishes ten with four. Who had the better result? Without the size of the wins, losses and costs, the question has no answer. A trading journal should make those missing quantities visible.
Expectancy is the average outcome implied by a set of outcome probabilities and payoffs. In a trade log, we estimate it from the observed frequency and average size of wins and losses. That estimate describes the sample. Calling it “expected” does not make the next trade predictable.
Build the number from an actual trade log
For a simplified set containing only wins and losses, let p be the winning fraction, W the average gross win, L the positive magnitude of the average gross loss, and C the average cost per completed trade. Then estimated net expectancy is p × W − (1 − p) × L − C.
If the broker’s profit and loss figures already include the costs you are using, do not subtract those costs twice. If there are break-even trades, keep them in the denominator and calculate the mean directly from every net result. “Per trade” must also have a consistent meaning: a completed entry and exit is not the same unit as a single order fill.
Two traders, the same ten opportunities
Invented dollar examples: Trader A wins $10 on each of eight trades and loses $50 on each of two. Gross result: $80 − $100 = −$20. At $2 total cost per completed trade, the ten trades cost another $20. Net result is −$40, or −$4 per trade, despite an 80% win rate.
Trader B wins $40 on each of four trades and loses $15 on each of six. Gross result: $160 − $90 = $70. With the same $20 aggregate costs, net result is $50, or $5 per trade, despite a 40% win rate. Neither example represents a recommended strategy or observed performance.
The useful comparison is not “low win rates are good.” Trader B would also lose money if the large wins disappeared or costs rose enough. The lesson is to retain all three pieces: how often, how much, and at what cost.
Calculate the break-even win rate
Set the simplified expectancy formula to zero and rearrange it: p = (L + C) ÷ (W + L). With W = $40, L = $15 and C = $2, the result is 17 ÷ 55, or about 30.9%. This is a mathematical threshold under those fixed assumptions, not a forecast of how often a strategy will win.
If average costs increase to $6, the threshold becomes 21 ÷ 55, about 38.2%. The four-win example then has only $1 of estimated net expectancy per trade. A modest change in execution can consume most of what looked like an advantage.
FINRA flags the cumulative effect of trading costs, including when the cost per trade looks small. The exact commission illustration in its disclosure is not a current broker quote; use your own relevant fees and execution assumptions. [FINRA Rule 2270: Day-Trading Risk Disclosure Statement]
The average can hide the event that matters
Suppose 99 completed trades each make $2 and one loses $250, before costs. The total is $198 − $250 = −$52. A 99% winning record still loses. A summary that stops at trade 99 would tell a radically different story.
Keep the worst loss, the distribution of losses and any open positions beside the mean. If a strategy delays closing losers, a high percentage of completed winners can coexist with a large unrealised loss. This is one reason the CFTC warns against marketing based on near-perfect AI trading claims. [CFTC: AI Won’t Turn Trading Bots into Money Machines]
Do not turn ten observations into certainty
Our ten-trade examples teach arithmetic, not statistical confidence. Trades may cluster by market condition, several positions may share one underlying exposure, and changing rules midway produces a mixture of strategies. A larger spreadsheet does not automatically contain more independent evidence.
Write down how the sample was selected. Were all signals included? Did you choose this market because its results looked good? Did the rules exist before the period began? Review a later, separate period without repeatedly adjusting the strategy to repair its outcome. See the backtesting guide for the fuller research process.
For your next review, calculate the mean net result directly, then reconcile it with the win-rate formula. If they disagree, check costs, break-even trades and your definition of a trade. The first benefit of expectancy is often catching an accounting mistake, not discovering a winning system.
Sources and further reading
Sources checked on 6 October 2026. Links support the nearby factual claims; worked examples and checklists are our educational illustrations.
- FINRA Rule 2270: Day-Trading Risk Disclosure Statement
Trading costs, execution problems, leverage and loss risks. The articles do not claim this US member-firm rule applies to every reader.
- CFTC: AI Won’t Turn Trading Bots into Money Machines
Warnings about guaranteed AI returns and checking costs, providers and underlying risks.
Prepared with AI assistance for the Mika vs Guru publishing team. This is not a claim of professional accreditation or independent peer review. How we research, label examples and handle corrections.