Options win rate vs. expectancy: why 90% can still lose

Start with one rule for what counts as a trade. A two-leg spread opened and closed as one position is one observation, not a winning short leg plus an unrelated losing long leg.

Use the same rule for partial exits and rolls. Keep the cash flows of a roll connected to its original position, even when the journal also records the closing trade and new opening separately.

Otherwise, closing small winners while leaving large losers open can produce an impressive realized win rate without describing the account's overall result.

Distinguish a historical win rate from a platform's estimated probability of profit. The former describes a sample; the latter depends on a model and a specified future outcome.

The separate probability of profit guide explains the model question. Here, the task is to reconcile a trading record with the dollars actually gained or lost.

For a simplified model with wins and losses only, define the inputs before calculating. Use the same currency and position-size convention throughout.

Average net outcome = p × W − (1 − p) × L − C.

Using historical averages gives the sample's average result. Using assumed probabilities and payoffs gives a scenario's expected value. Neither calculation proves those inputs will persist.

For records containing flat trades or irregular outcomes, sum every completed trade's net profit or loss and divide by the total trade count. This avoids quietly dropping inconvenient observations.

If W and L already include all costs, do not subtract C again. Classify a win consistently: a small gain before fees may be a loss after fees.

• p: fraction of trades that win, written as a decimal.
• W: average winning trade before the separately recorded costs.
• L: average losing trade's absolute dollar loss before those costs.
• C: average cost per completed trade, charged once in the calculation.

Consider a hypothetical 100/95 bull put spread with the same expiration, a 0.50 net credit, and a 100-share multiplier. The two strike prices are $5 apart.

Its full expiration profit is $50. Its full expiration loss is ($5 − $0.50) × 100 = $450, before costs, assuming the spread's two legs are handled together as intended.

This follows the payoff structure in the [OIC bull put spread guide](https://www.optionseducation.org/strategies/all-strategies/bull-put-spread-credit-put-spread).

To isolate the arithmetic, assume 90 trades make the full $50 and 10 lose the full $450. Actual spreads can also finish between those outcomes or close early; this is not a backtest.

The average net outcome is −$2 per trade. In this example, one full $450 loss consumes nine full $50 gains before any costs are paid.

A received option credit is not the same as a completed profit. The liability to close or settle the spread must be included in the record.

• Winning trades: 90 × $50 = $4,500.
• Losing trades: 10 × $450 = $4,500.
• Result before costs: $0 despite winning nine trades out of ten.
• Hypothetical costs: 100 × $2 = $200; net result: −$200.

Set the expectancy equation to zero and solve for p. With positive W and L, the required win rate is (L + C) ÷ (W + L).

For the example, ($450 + $2) ÷ ($50 + $450) = 0.904, or 90.4%. The assumed 90% is below that threshold.

That is a win-rate threshold, not the spread's stock-price break-even. The latter is $100 − $0.50 = $99.50 at expiration before costs.

Changing strikes, taking profits earlier, or adding a stop changes the outcome distribution. Recalculate both average gains and average losses rather than keeping whichever old input looks attractive.

Hold the example's $50 win, $450 loss, and $2 cost fixed. Changing only p gives three different expected outcomes.

These are sensitivity cases, not forecasts. They show how much the result depends on a probability that may not be known accurately.

Also test larger execution costs or a lower average gain. When the modeled advantage disappears after a small assumption change, record that fragility instead of reporting only the best case.

• At 85% wins: 0.85 × $50 − 0.15 × $450 − $2 = −$27 per trade.
• At 90% wins: 0.90 × $50 − 0.10 × $450 − $2 = −$2 per trade.
• At 95% wins: 0.95 × $50 − 0.05 × $450 − $2 = $23 per trade.

Profit factor divides total gains on winning trades by the absolute total losses on losing trades. State whether those trade results are before or after costs.

Before costs, the example's factor is $4,500 ÷ $4,500 = 1.00. Allocating $2 to every trade gives net winners of $4,320 and net losses of $4,520, a factor of about 0.956.

The same record therefore has a 90% win rate, a negative average net outcome, and a cost-adjusted profit factor below one. The metrics agree once their definitions match.

With no losing trades in the sample, the denominator is zero. An undefined or infinite-looking factor does not establish that the strategy has no downside.

Eighteen wins in 20 trades and 900 wins in 1,000 trades both equal 90%. They contain different amounts of evidence, and neither ratio establishes the next trade's probability.

Overlapping trades on the same underlying may share one market shock. Counting each as an independent success can exaggerate how much independent evidence the record contains.

Position size matters too. If losing positions are larger than winning positions, a per-contract average can conceal the actual account loss. Review both standardized trade outcomes and the cash ledger.

A positive estimated average is not a drawdown limit. Several losses can arrive before any gains, and a leveraged account may be unable to sustain that sequence.

See [OIC's leverage discussion](https://www.optionseducation.org/optionsoverview/leverage-risk) for why exposure and downside must be considered separately from the frequency of wins.

Choose the review period before selecting trades. Reconcile the complete sample with broker records, and display remaining open-position exposure alongside realized performance.

Do not charge a modeled bid-ask penalty again when actual execution prices already include its effect. Backtests using midpoint quotes need a separate, explicit execution assumption. [!TRYMARK] TryMark expectancy checkpoint At the next journal review, compare average net P/L with zero. Recheck the result using the actual closing fills, all position costs, and a worse loss assumption before interpreting a high win rate. [!WARNING] A payoff cap is not an account-management promise Early assignment or mismatched leg handling can create stock exposure and funding needs. The simplified spread example is not a guarantee that every account outcome stays inside its expiration payoff.

• Record the full position, quantities, multiplier, and all entry and exit cash flows.
• Link partial exits, rolls, exercise, assignment, and resulting stock positions.
• Separate commissions and fees from slippage already reflected in actual fill prices.
• Report net dollars, trade count, average win, average loss, and largest loss together.
• Keep the sampling dates, exit rules, and changes in position size visible.

A low option delta is a price sensitivity, not evidence that your trading process wins at the complementary percentage. The delta explanation covers that distinction.

The calculations here are original hypothetical arithmetic, not live quotes, personal trading results, or a tested strategy. They evaluate a record or assumption set; they do not recommend buying or selling an option.

For the execution side, continue with options trading costs and the trade-record checklist.