Trading Win Rate: Why Winning 70% Can Still Lose Money

Define a trade and a win consistently

A basic win rate divides the number of winning completed trades by the total number included in the sample and multiplies by 100. Seven out of ten is 70%. That calculation still needs definitions: what constitutes a completed trade, whether a win is net of fees and how break-even or partial exits are treated.

One purchase order may generate many fills. Counting each purchase fill as a separate success while treating an entire losing exit as one trade creates an inconsistent statistic. Choose a review unit and apply it across every row. If the definition changes, recalculate earlier comparisons instead of displaying percentages that answer different questions.

Follow a 70% record to a negative total

Assume ten hypothetical closed trades. Seven gain ten reference-currency units each, producing total gains of 70. Three lose thirty each, producing losses of ninety. Net aggregate before any excluded costs is 70 − 90 = −20. Every loss offsets three gains of the stated size.

The 70% win calculation is correct. The mistake would be treating that count as proof of profit. A balance responds to the amount of each outcome, whereas the simple frequency measure gives every trade one vote regardless of size.

The CME mathematical-expectation lesson discusses the relationship between frequency and outcome size. Our arithmetic describes a constructed sample, not a guaranteed distribution for a future trading method.

Distinguish a sample average from a prediction

Using the observed frequencies, the average result is 0.70 × 10 − 0.30 × 30 = −2 per trade. Multiplying by ten reproduces the −20 total. Here average loss is expressed as a positive magnitude that the formula subtracts. If results are already signed, simply sum them and divide by the count rather than reversing loss signs twice.

Substituting assumed probabilities for observed frequencies creates a model expectation. That is different from describing the previous ten results. The mathematical form can look identical while the inference changes. An observed average does not establish that the future distribution remains unchanged or that a particular number will occur on the next trade.

Calculate a simplified break-even frequency

With fixed gains of ten, fixed loss magnitudes of thirty, no break-even trades and no additional costs, the break-even win rate is 30 ÷ (10 + 30) = 75%. In twenty trades, fifteen gains of ten total 150, while five losses of thirty total 150. The constructed sample then breaks even.

With equal net gain and loss magnitudes, the same simplified threshold is 50%. Changing either magnitude changes the threshold. There is no universally good win rate independent of those amounts and cost definitions. Even matching the theoretical frequency does not ensure that a later finite sample realizes it.

Include costs without deducting them twice

Consider another hypothetical sample: six gains of four and four losses of five, all before commissions. Gross result is 24 − 20 = +4. A cost of one per completed round trip totals ten, turning the result into −6. The initial 60% gross win rate did not reveal that reversal.

Fees can also turn individual small gross winners into net losers, changing the win rate itself. Label whether classification happens before or after costs. A gross frequency placed beside a net total can be valid only if the differing definitions are clear.

If averages already include commissions, do not subtract the same fees again in the expectation formula. A transfer belongs in the review only when included in the task’s scope, and should be documented separately. The transfer fee calculator can isolate that component without changing what each trading result means.

Position size changes the financial weight

Seven small wins and three large losses can arise from different capital allocations even when percentage price moves look similar. Use actual amounts and a stated reference currency. A simple average of trade percentages is not automatically the account’s return, especially when sizes vary.

For example, two hypothetical 2% gains on allocations of 100 produce four overall. A later 1% loss on an allocation of 1,000 loses ten. The total is −6 despite two of three trades winning. The larger position has greater financial weight than its one-third share of the trade count.

Keep timing and external cash flows visible too. If position size changes with account equity, compounding differs from adding percentages. Use the actual cash-flow and valuation record when calculating account performance rather than treating every percentage as though it applied to the same starting amount.

Open losses and split exits can distort the picture

Closing small winners while leaving a large losing position open can produce a strong closed-trade win rate alongside a weak account valuation. Do not label the open loss realized before an exit, but show its separate valuation effect so it does not disappear from the explanation.

Likewise, ten partial exits from one winning position should not silently become ten independent victories while a complete losing position is counted once. Either group related parts under a stated rule or present a fill-based analysis with the correct label. Consistency matters more than choosing whichever counting method yields a higher percentage.

Inspect sample selection, money totals and sequence

Seven wins out of ten and seventy out of one hundred have the same percentage but contain different amounts of information. More observations do not alone establish future representativeness: results can come from one asset, market condition or simplified simulator. Report period, number of trades and whether results are actual, simulated or backtested.

The Investor.gov performance-claims bulletin discusses presentation limits, including selected periods and hypothetical performance. Preserve errors and losses within the stated scope, and distinguish rules developed on a sample from independently evaluated results.

A separate amount-based ratio divides total gains by total loss magnitudes. Our first sample gives 70 ÷ 90, approximately 0.7778. That is not a win rate. If losses are zero, its denominator is zero and the statistic cannot be read as unlimited proven strength. Retain the original amounts and chronological sequence. The trade-journal guide helps keep those definitions and records together.

Continue reading