Trade Distribution Analysis: Read Your Strategy Beyond Win Rate
Learn how to run a trade distribution analysis: build the histogram of your trade results, read skew and fat tails, and avoid sample-size and bias traps.
Trade distribution analysis means sorting every trade a strategy produced by its result and studying the shape of that spread: how many trades were small wins, how many were small losses, and how far out the biggest outliers sit. That shape tells you more about whether a strategy is tradable than win rate or total profit alone, because two strategies with identical totals can have completely different risk. This guide shows how to build the distribution, read it, and spot the traps that make a backtest look better than it is.
A note on terms: some pages use "trade distribution" to mean the spread of executed trade sizes in market data. Here we mean the distribution of your strategy's trade outcomes, which is the version that helps you decide whether to risk capital.
TL;DR
- A trade distribution is a histogram of per-trade results (in dollars, percent or R-multiples). Its shape shows what the average hides.
- Read it for skew, outliers and the size of the worst trades, not only for win rate and average profit.
- Real returns are often negatively skewed with fat tails, so a model that assumes a normal bell curve understates the worst case.
- Small samples, look-ahead bias, ignored costs and overlapping trades all distort the picture. Test for them before trusting any shape.
- The distribution describes what a strategy did. It does not forecast the next price move, but it does tell you what range of outcomes to prepare for.
Table of Contents
- What trade distribution analysis actually tells you
- How to build the distribution step by step
- Reading the shape: what each pattern means
- A worked example with made-up numbers
- Fat tails: why the bell curve understates your worst day
- Sample size, bias and costs: what distorts the picture
- Can the distribution predict the next trade?
- A checklist before you trust a distribution
- Run it on your own strategy
- FAQ
- Sources
What trade distribution analysis actually tells you
A backtest report usually leads with a few numbers: total return, win rate, profit factor, maximum drawdown. Each one compresses hundreds of trades into a single figure. The distribution is the uncompressed version.
The CFA Institute makes the same point in its discussion of performance measures: most metrics focus on the first two moments, mean and variance, and neglect higher-order moments such as skewness and kurtosis, which matter for judging tail risk. Skewness tells you whether the big outcomes lean toward gains or losses. Kurtosis tells you how heavy the tails are compared with a bell curve.
In practice, the distribution answers questions a summary table cannot:
- Does the profit come from many ordinary trades or from three lucky ones?
- How bad is the worst trade compared with the average trade?
- Are the losses clustered around one size (a stop doing its job) or scattered with a long tail?
- If you removed the best 5% of trades, would the strategy still make money?
How to build the distribution step by step
You need a complete list of trades from a backtest or a trading journal. Each row needs entry, exit, direction, size and fees. Then:
- Pick one common scale. Dollar profit lets large positions dominate, so convert each trade to a percent return or to R-multiples (profit divided by the amount risked on that trade). R-multiples make a $50 win on a small position and a $500 win on a large one comparable.
- Subtract costs. Use net results after commissions, spreads and a realistic slippage assumption. A distribution built from gross results is shifted to the right by an amount you will never keep.
- Choose bins. For 100 to 300 trades, 10 to 20 bins is a sensible start. Too few bins hide the tails; too many leave you staring at noise.
- Plot the histogram and mark the mean, the median and zero.
- Compute the companion statistics: average win, average loss, expectancy per trade, largest win, largest loss, skew and the share of total profit from the top 10% of trades.
- Slice it. Split by long and short, by setup, by instrument, by holding period and by calendar period. A healthy-looking aggregate can hide one slice that loses money every year.
If you want the full workflow from rules to report, our step-by-step guide to backtesting a trading strategy covers how to get a clean trade list in the first place.
Reading the shape: what each pattern means
Different shapes imply different strategies, and different ways to fail.
| Shape | What it looks like | What it usually means | What to check |
|---|---|---|---|
| Right-skewed | Many small losses, a few large wins | Typical of trend following and breakout styles | Is profit dependent on a handful of trades? Can you stay with the strategy through long losing runs? |
| Left-skewed | Many small wins, rare large losses | Typical of mean reversion, premium selling and no-stop systems | How large is the worst loss relative to months of gains? What caps it? |
| Roughly symmetric | Wins and losses of similar size | Fixed target and stop with roughly equal distance | Win rate must carry the edge, so costs matter a lot |
| Bimodal | Two clusters of results | Two different behaviors mixed together, such as stop hits and target hits | Split by exit type and setup |
| Heavy single tail | One or two extreme points | An outlier driving the whole result | Re-run without it, and ask whether it can repeat |
The important habit is to compare the median trade with the mean trade. When the mean sits far above the median, a few big winners are carrying the strategy. When it sits far below, a few big losers are eating the small wins.
A worked example with made-up numbers
The figures below are hypothetical and exist only to show the mechanics.
Strategy A takes 100 trades. 45 win an average of $120 (total $5,400) and 55 lose an average of $60 (total $3,300). Net profit is $2,100, expectancy is $21 per trade, and the win rate is 45%. Looks mediocre by win rate.
Strategy B takes 100 trades. 90 win $20 each (total $1,800). Nine lose $25 each (total $225). One loses $1,500. Net profit is $75, expectancy is $0.75 per trade, and the win rate is 90%. Looks excellent by win rate.
On a win rate leaderboard, B wins. On a histogram, B is a pile of small wins with one point far out on the left, and a single repeat of that outlier erases months of gains. Strategy A has a modest, bounded left side. If you are going to trade real money, the distribution tells you to prefer A.
This is also why a stop-management rule can change the whole shape. Moving a stop to breakeven turns some would-be losers into scratches and cuts off the left tail, at the cost of clipping some winners. If you want to see how that logic is coded and tested, this walkthrough of a breakeven stop strategy is a good companion to a drawdown-based reading of your results.
Fat tails: why the bell curve understates your worst day
The phrase "fat tails" describes distributions where extreme outcomes happen more often than a normal curve predicts. Trading results are a common example.
The evidence is consistent across asset classes. Research on the momentum factor found that these strategies have historically generated high positive returns but suffer infrequent yet severe losses. A study of currency speculation found payoffs that are leptokurtic with fat tails, which mean and variance alone do not capture. The CFA Institute's 2026 reading on backtesting adds that standard backtests often assume a multivariate normal distribution, which ignores the negative skewness and fat tails seen in real asset returns.
What this means for you:
- Do not size positions off the average loss. Size them off the worst plausible loss, including gaps through your stop.
- Do not trust standard deviation alone. A strategy can show calm volatility for years and still contain a rare, severe trade.
- Look at the extremes directly. Report the worst 1% and 5% of trades, the longest losing streak and the maximum drawdown.
Maximum drawdown is the natural partner of the distribution. The CFA Institute describes it as the accumulated loss from a peak to a trough and notes that, unlike variance, it is path dependent and captures an expected worst case. Our guide to maximum drawdown for traders explains why a 50% loss needs a 100% gain to recover.
A single historical path is only one draw from the possible outcomes. To see how much worse the same trades could have gone in a different order, resample them. We cover that in Monte Carlo backtesting that exposes tail risk, and the same idea applied to survival is in the piece on risk of ruin.
Sample size, bias and costs: what distorts the picture
A distribution is only as honest as the trades behind it. Four problems come up again and again.
Too few trades
With 30 trades, one outlier can move the mean, the skew and the histogram shape visibly. Tail statistics are the least stable of all, because the tail is made of the rarest events. Treat a shape from a small sample as a hypothesis, not a result. Statistical work on non-normal data notes that nonnormality can affect Type I and Type II errors, and that bootstrap methods and robust estimators are particularly effective for handling nonnormality and outliers. In plain terms: resample your trades and see how much your key numbers move.
Look-ahead and survivorship bias
The CFA Institute's backtesting reading tells analysts to watch for survivorship bias and look-ahead bias. Either one fattens the right side of your histogram with trades that could never have been taken. If the left tail looks suspiciously thin, check for these first. We list four fast checks in exposing look-ahead bias in quant backtests.
Costs and slippage
Costs shift the whole distribution left, and they hurt high-frequency, small-edge strategies most because the edge per trade is already tiny. An NBER study of pairs trading found that bid-ask bounce and transaction costs reduced returns, although they stayed positive and significant in that sample. The lesson is not that costs always kill an edge, but that you must model them before you read the shape.
Overlapping and correlated trades
Ten trades opened during the same market event are not ten independent decisions. If your list contains clusters, the effective sample size is smaller than the row count, and a single event can produce several of your worst results at once. Group trades by day or event when you check how many truly independent observations you have.
Can the distribution predict the next trade?
It can tell you what range of outcomes to expect if the strategy keeps behaving as it did. It cannot tell you whether the next trade will win.
The distribution is a description of past results under past conditions. Three limits matter:
- Regimes change. A strategy that thrived in a trending year can produce a different shape in a choppy one. Slice the distribution by period and market condition and see whether the shape is stable.
- Costs and behavior change. Spreads widen in stress, and so do your own mistakes.
- Tails are under-sampled. The worst outcome in your data is not the worst outcome possible.
What the distribution does support is planning: how much to risk per trade, how deep a drawdown to expect, and how long a losing streak to budget for. Probabilistic forecasts built from historical analogs can add a view of what has followed similar setups, but they come with confidence ranges for a reason, and you should treat them as odds, not promises. The sizing side is covered in our risk-reward and 2% sizing backtest.
A checklist before you trust a distribution
- At least 100 trades, ideally many more, across more than one market regime.
- Net of fees and a realistic slippage assumption.
- No look-ahead or survivorship bias.
- Median and mean compared, and results rechecked without the top 5% of trades.
- Worst trade and worst drawdown sized against your account, not against the average.
- Results split by side, setup and period, with no slice carrying all the profit.
- Trades resampled to show the range of outcomes, not just the one that happened.
If you test strategies you saw online, apply the same checklist. A creator showing a handful of winning charts is showing you a handful of points, not a distribution. Our approach to this is laid out in backtesting a trading strategy like a regulator.
Run it on your own strategy
Backtestify lets you describe a strategy in plain English, run it on real historical market data and see every trade plus the performance metrics that matter: win rate, profit factor and maximum drawdown. You can compare an original strategy with an improved version on recent data, and export to TradingView Pine Script. We also publish results for winning and losing strategies, so you can see what a full report looks like in the published strategy library and in our methodology.
You can start on the Free plan. Pro, which adds unlimited backtesting, improvement and forecasting, costs $29 per month or $190 per year, and you can see the Pro plan here.
FAQ
What is trade distribution analysis?
It is the study of how your individual trade results are spread out, usually as a histogram of profits and losses per trade. It shows skew, outliers and the size of your worst trades, which win rate and average profit hide.
What do fat tails mean in trading?
Fat tails mean extreme gains or losses occur more often than a normal bell curve predicts. Research on momentum and currency strategies has documented this, so risk measures built on a normal distribution can understate your worst case.
How many trades do I need for a reliable distribution?
There is no magic number, but 30 trades is too few to trust the shape, and tail statistics need far more. Aim for at least a few hundred trades across different market conditions, and resample them to see how much your numbers move.
Does a high win rate mean a good distribution?
No. A strategy can win 90% of the time and still lose money if its rare losses are large. Check average win against average loss, expectancy and the worst single trade alongside the win rate.
Sources
- Why do we need another Greek letter in portfolio management? An introduction to Omega (CFA Institute)
- Sculpting Investment Portfolios: Maximum Drawdown and Optimal Portfolio Strategy (CFA Institute)
- Tail Risk in Momentum Strategy Returns (NBER)
- The Returns to Currency Speculation (NBER)
- Backtesting and Simulation (CFA Institute, 2026)
- Pairs trading study on bid-ask bounce and transaction costs (NBER)
- Hypothesis Testing in Business Administration (Oxford Academic)