Why a 35% win rate can still be net-positive

The first time most traders open the sandbox and see a bot with a 34% win rate, they close the tab. The brain reads “win rate” as “success rate” — and a 34% success rate at anything sounds awful.

But win rate, by itself, tells you almost nothing about whether a strategy is net-positive on history. The number that matters is expectancy — what a single trade is worth on average, when you include the size of the wins, the size of the losses, and how often each happens. This post is a short walkthrough of why a 35% win rate can sit comfortably in the net-positive column, and how to read a sandbox result without panicking at the first number.

This is educational only. Nothing here predicts the future or describes what any specific strategy will do on your account.

The one-line formula

Expectancy per trade is a weighted average:

expectancy = (win rate × average win) − (loss rate × average loss)

If the result is positive, the strategy is net-positive on the history it was measured against. If it is negative, it is net-negative on that history. That is the whole game. Win rate is one of four inputs.

The mistake most traders make is treating it like the only input.

A worked example — 35% wins, still ahead

Suppose a bot took 100 trades against a year of BTC/USDT history:

• 35 winners. Average win = +3.0R (3× the amount risked on each trade).
• 65 losers. Average loss = −1.0R (the stop-loss took 1R on each).

Run the formula:

expectancy = (0.35 × 3.0R) − (0.65 × 1.0R)
           = 1.05R − 0.65R
           = +0.40R per trade

Across 100 trades, the strategy banked +40R on history. Risking 1% of equity per trade, that compounds to a meaningful number — entirely because the winners are three times the size of the losers. The losses happen more often, but each one is small.

This shape — small frequent losses, occasional large wins — is what trend-following systems look like. The opposite shape — large frequent wins, occasional catastrophic losses — is what scalping and mean-reversion systems often look like. A 70% win rate strategy can be net-negative if the losers are 5× the size of the winners.

Why this is counter-intuitive

Win rate triggers an emotional reaction. The trader sees “65% losing trades” and feels every one of them — even though the math says they add up to a positive month. Three things make this harder to internalise than it should be:

• Recency bias. Five losers in a row feel like the strategy is broken, even when the math says the next win will more than cover them.
• Lottery framing. Humans intuitively want lots of small wins, not a few big ones. Most systematic strategies that worked on history pay the other way.
• Stop-losses look like failures. A stop-loss that takes 1R out of your account looks like a defeat. In a trend-following system, it is the cost of being in the market — the price of staying ready for the next big winner.

What to actually look at in a backtest result

When you read a sandbox result, walk the four numbers in this order instead of fixating on the first one:

1. Total return on the window. Did the strategy end ahead of buy-and-hold over the same period? This is the only top-level answer that matters.

2. Win rate. Not for celebration — for characterisation. A win rate below 40% almost always means a trend-following system. Above 60% almost always means a mean-reversion or scalping system. Different systems break in different conditions. Knowing which you’re looking at tells you what conditions to worry about.

3. Max drawdown. This is the test that matters more than win rate. A strategy that returns +40% on history but draws down 35% is the same strategy that 19 out of 20 traders abandon halfway through the drawdown. If you cannot psychologically hold a position through a 30% drawdown, the strategy’s expectancy doesn’t matter — you will exit before it pays.

4. Worst month. The month where everything went wrong. If the worst month was −18%, ask yourself honestly: would you have stayed subscribed to this bot through that month? If the answer is “no”, pick a different bot — or a smaller position size.

These four together give you a picture. Any one of them in isolation is a misleading number.

How to use this on Stralines

The sandbox lets you replay any of six bots against real Binance history. The result panel shows total return, win rate, max drawdown, worst month, and trade count — the same four numbers above — on whichever pair and lookback period you select. The Bollinger Breakout bot on BTC/USDT, last 12 months is a good worked example: sub-35% win rate, sub-10% max drawdown, net positive on the window.

That is a study of the past. It says nothing about the next 12 months, and Stralines does not predict that the same shape will recur. What the sandbox does give you is a chance to internalise the math before you fund an account — which, in our view, is the single biggest predictor of whether a trader sticks with a systematic approach long enough for the expectancy to actually show up.

A grown-up version of the same idea

If you have read this far, three further reads are worth your time — all from this blog, all sized to about 5 minutes each:

• Position sizing and risk caps — why the size of each trade is the lever that actually controls drawdown, not your strategy’s win rate.
• Manual account blow-up mistakes — what happens when traders override a systematic stop-loss because a loss “feels” worse than a number on a screen.
• Why the shift from manual to algo trading is essential in 2026 — the engineering case for moving execution into a system that doesn’t get tired, distracted, or emotional.

Closing — a quieter framing

A 35% win rate is not a bug. It is a description of a strategy’s shape. What matters is whether the wins are large enough, on average, to pay for the losses — and whether the drawdowns are small enough that you will actually stay in the seat long enough to find out.

Stralines is software for running this kind of strategy with the execution and risk-management discipline already encoded — so the trader is left to do the one job that matters most: keep their hands off the system through the drawdowns. The math will take care of the rest, if the math is honest in the first place.

Nothing here is investment advice, and the sandbox shows historical behaviour on past data only. Past performance does not predict future results. Capital is at risk.