Does a 70% win rate in gold trading mean you make money? First separate the number of trades from the amounts

Author: Thomas | Trading Knowledge | October 17, 2026

When looking at trading records, a 70% win rate easily creates a sense of reassurance. But win rate only answers "how many trades made money"; it does not answer "how much was made, how much was lost, and whether the records are complete." When evaluating a gold trading method, the win rate should be placed back into the context of amounts, costs, and sample background, rather than letting a single percentage replace all judgment.

Win rate only counts the number of trades, not the size of losses

Suppose there are ten closed trades, of which seven are profitable and three are losing, so the win rate is 70%. If each profitable trade makes 10 accounting units and each losing trade loses 30 units, the total is seven times ten minus three times thirty, resulting in a loss of 20 units.

This is a completely fictional teaching ledger; the units are not USD quotes and do not represent anyone's real trades. It shows that even when more trades are profitable, a few larger losses can still exceed many small profits. It cannot be inferred from this that a 70% win rate must lose money; the key lies in the amount per trade and the statistical basis.

Same number of trades, a different amount structure

Still ten trades: if four each make 30 units and six each lose 10 units, the total is 120 minus 60, for a profit of 60 units. This ledger has a win rate of only 40%, but under the assumed amount conditions the result is positive.

The two hypothetical sets are not intended to recommend a certain win rate, nor do they show a method for estimating real win rates. They simply separate "number of trades" from "amounts" so readers can see that the two dimensions cannot substitute for each other. A low win rate is not inherently an advantage, and a high win rate is not inherently a drawback.

Put costs into the same ledger

costs are temporarily excluded from the examples above. If each trade also incurs total costs of 2 units that are not yet included in the profit and loss, then ten trades subtract another 20 units, and the two ledgers become a loss of 40 and a profit of 40 respectively. This cost is purely hypothetical and is not the fee rate of any platform.

When checking real records, first confirm which items are already included in the reported profit and loss. Spreads already reflected in execution prices must not be deducted again; if commissions and holding costs are listed separately, they should be combined for reconciliation. Fields may differ across platforms, so seeing a "profit" label does not mean it is the final net result by default. Costs affect returns, but they are only one part of the evaluation framework.

Averages can describe the past, not guarantee the future

In a closed-trade sample with a consistent basis, you can multiply the proportion of profitable trades by the average profit, then subtract the proportion of losing trades multiplied by the absolute value of the average loss, to get the average result per trade; if costs are not included, subtract the average cost separately. When break-even trades exist, their count should also be retained and not arbitrarily removed to make the numbers look better.

This calculation is only a sample description. Future market conditions, execution conditions, and trading behavior may all change, so a historical average profit per trade does not mean the next trade will definitely make money. In particular, do not infer from a single positive number that a larger position size is more reasonable.

First ask whether the statistics are complete

Ten trades are usually not enough to represent how a method performs across various environments. When reviewing a win rate, you should at least know the start and end dates, the number of trades, whether failed orders are included, and whether only a certain smooth period is shown. This article does not give a fixed qualifying sample size applicable to all strategies.

It is also necessary to distinguish closed positions from positions still held. If profitable trades are settled promptly while losing trades are kept, looking only at the closed-trade win rate may miss floating losses. Partial closes split into multiple records may also change the definition of a trade count. First unify the rule for "what counts as one trade," then compare different months or different methods.

What a weekend review can look like

First export a complete record for a continuous period, retaining profitable, losing, and break-even trades; then list net profit and loss, average profit, average loss, and the largest single loss; finally check the risk exposure that has not yet ended separately. Do not temporarily change the start and end dates or hide unfavorable records in order to obtain a higher win rate.

If two records use different position sizes, the amount results also cannot directly prove which method is better. The scale of capital invested and risk conditions need to be recorded at the same time. A large loss after a series of small profits is worth checking to see whether loss control has changed, rather than merely reassuring yourself that "most of the time I was right."

When judging a trading method, the win rate can be retained, but it must be read together with the amount structure, costs, and completeness. All numbers in this article are teaching assumptions, not backtest or live results, and do not constitute investment advice or return promises.