Eight of your last nine trades lost half their risk. The ninth made twelve times its risk. Your average R-multiple now reads +0.9R, which looks like a real edge. The median of those same nine trades is -0.5R, which says you lose on a typical trade. Both numbers are correct. Only one of them is telling you the truth about your setup.
Your average R-multiple is the single number that says whether a setup makes money. If R-multiples are new to you, start with what an R-multiple is: every trade's result expressed as a multiple of the risk you took. The metric is fine. You read it too early, on too few trades, and let one outlier speak for the whole record. The fix is to read average R honestly: enough trades first, the median next to the mean, and a way to stop one freak trade from flattering your stats.
What your average R actually measures
Your average R-multiple is your expectancy: the profit, in units of risk, you make on a typical trade. Expectancy in R equals your win rate times your average winning R, minus your loss rate times your average losing R. Because the average loss is close to 1R by definition, the whole thing collapses to one risk-normalized number you can compare across any account size or position size.
The useful ranges are narrow. For most durable strategies, an expectancy of +0.2R to +0.5R per trade is a solid edge. Below about +0.2R, commissions and slippage eat most of it. Anything sustained above +1R is rare enough to be worth auditing before you believe it. For the full treatment of expectancy and trading edge, that companion post covers win rate and average R together. The focus here is narrower: whether the average R you are looking at is real yet.
Why one outlier can flip a small sample
On fewer than about 20 trades, a single outlier moves your average R more than your skill does. R-multiple distributions skew positive by design. A trend-following record is usually a pile of small losses near -1R to 0R and a handful of large winners of 10R or more, and those rare winners carry the entire result.
Run the numbers on the nine-trade example. Eight trades at -0.5R and one at +12R sum to +8R, an average of about +0.9R. Strip the outlier and the other eight average -0.5R. So the headline number describes a record that loses on eight trades out of nine. The single winner is real and you keep it. What you cannot do is let it set your expectation for the next trade, because you have no evidence the setup produces 12R winners at any reliable rate.
How many trades before your average R is real
Expectancy from 15 trades is close to meaningless, because one outlier can make a losing strategy look like a winner. A rough read needs at least 60 trades. Confidence needs 100 to 200 or more. Below about 50 trades, the luck of which outliers landed in the sample decides your average.
Van Tharp, who popularized R-multiples in the 1990s, built sample size straight into his System Quality Number: the square root of the trade count, times the mean R, divided by the spread of your R results. More trades raise the score. Tighter, more consistent results raise it. A big average built on a few wild trades does not. Sample size is part of the measurement.
Swingfolio flags any set under 20 closed trades as a low sample, so you see the warning before you trust a number that cannot yet hold weight.
Read the median next to the mean
The median R-multiple is the result of your middle trade, and outliers do not move it. Line your trades up from worst to best and the median sits in the centre. One 12R winner cannot drag it, because it only counts as one trade above the middle.
Put the two side by side and read the gap. A mean far above the median means your record leans on a few big winners, and your typical trade is weaker than the average suggests. A mean close to the median means the edge is broad: most trades pull the same way, not one outlier doing all the work. Swingfolio shows the median next to the mean, so you never read the average alone.
Clip the tails before you trust the mean
Winsorising replaces the most extreme results at each end with the next value in, so one freak trade cannot dominate the average. Clip the top 5% and the bottom 5%, and a lone +40R or -8R sits at the edge of the normal range instead of bending the whole mean around itself.
Swingfolio winsorises your average R once you have 20 trades or more, trimming 5% at each tail, and keeps the raw average visible too so nothing is hidden. Under 20 trades it leaves the average raw, because clipping a tiny sample throws away more signal than it cleans up. Winsorising keeps every trade in your record. It just stops your best and worst from being the only ones that move the average.
When you traded without a stop
A trade with no stop loss still gets an R in Swingfolio, computed from the deepest drawdown it suffered. That worst point becomes the implied risk, so a trade you opened without a plan still lands on the same scale as the rest, on a rougher but honest basis.
R goes wrong here too. A trade can lose more than 1R if your stop slips or you hold past your intended exit, and those tail losses beyond -1R do more long-term damage than any single winner repairs. Honoring the stop is what keeps R consistent. The number is only as honest as the discipline behind the exit.
How Swingfolio keeps your average R honest
Swingfolio computes R-multiple from each closed trade's entry, stop, and exit, with no manual maths and no spreadsheet. Position size cancels in the formula, so a 10,000-unit trade and a 200-unit trade land on the same scale. Trades opened without a stop fall back to their worst drawdown. Every R reading in the app, on the trade detail page, in the trades table, and across analytics, runs through one calculation, so the numbers never disagree with each other.
On top of the raw figure you get the three things that keep an average honest: a low-sample warning under 20 trades, the median beside the mean, and a winsorised average that clips the extreme 5% once the sample is large enough to trust. Tag the market regime on each trade as well and you can read average R by setup and condition, so you see which setups pay and when.
Frequently asked questions
How many trades do I need before my average R-multiple means anything? About 60 for a rough read and 100 to 200 for confidence. Under 50, one outlier can flip the number, so treat anything below that as a hint, not a verdict. Swingfolio flags samples under 20 as low.
Why is my median R-multiple lower than my mean? Because a few large winners are pulling the mean up while most of your trades sit lower. A mean well above the median means your record leans on outliers. A typical trade looks more like the median.
Can one big winner make my strategy look better than it is? Yes, on a small sample. One 12R trade in a set of nine can turn a losing record into a positive average. The winner is real, but it does not prove the setup repeats it. Wait for more trades and check the median.
What is a good average R-multiple? For most strategies, +0.2R to +0.5R per trade is a solid edge. Below +0.2R, fees and slippage erode it. Above +1R sustained is rare and worth checking for a few outliers inflating the figure.
Should I delete outlier trades from my stats? No. Deleting trades hides what happened. Use the median, which ignores extremes, and a winsorised mean, which trims them without removing the trade. Both keep the record intact.
General information only. Not financial advice.