What Is Risk-Multiple Efficiency?

How is RME calculated?

RME = net R kept ÷ gross R risked. Add up the R you actually staked across a window of trades, add up what came back, and divide. Worked through ten trades each risking 1R:

Trade	1	2	3	4	5	6	7	8	9	10
Result (R)	+2	−1	+1.5	−1	−1	+3	−1	+0.5	−1	+1

Net kept is +3R. Gross risked is 10R. RME is 3 / 10 = 30% — for every unit staked, three-tenths of it came back as profit. The win rate here is exactly half, which tells you almost nothing on its own; RME reads the size of the result against the size of the bet.

Why divide by risk instead of counting wins?

Because it catches the failure a win rate cannot see. Run the same ten trades at double the size — 2R staked each time — and suppose they produce the same +3R net. Gross risked is now 20R, so RME halves to 3 / 20 = 15%. Identical outcome, identical skill on display, half the score, because twice as much was put at risk to get there.

That is the whole point. Risk creeping upward is the single most common way an account gets into trouble, and it is invisible in PnL until the streak arrives. RME slides first: when the size balloons, the ratio moves before the equity curve does.

Why show a percentile instead of the ratio?

Because a raw ratio is not comparable across styles. A patient method taking few, large, low-probability trades produces a very different RME from a high-frequency scalping method, and neither number means anything next to the other. Ranked against your own history, both read the same when each is performing normally.

So kappi ranks your latest reading against your own last 20 and shows the percentile. Above 55% is strong form by your own standard; below 15% is your worst recent stretch. It is you, playing you — which also means a good RME is not a claim to be better than anybody else, and was never meant as one.

Two properties follow from ranking against yourself. It rewards patience: a sound low-win-rate, high-reward method stays green through the losing streaks it is supposed to have. And it catches drift early, because the percentile is sensitive to the denominator that PnL ignores.

When does it start reading?

At 11 closed trades, and not before. Below that there is genuinely not enough to read, so kappi says so instead of printing a confident number derived from one or two results — which is the same discipline every other figure on this site is held to. Between 11 and 20 readings it ranks against however many exist so far.

The comparison set is always your most recent 20 readings, so the percentile moves as you trade rather than describing a period you have left behind.

What happens to a trade with no defined risk?

It is not thrown out, which is a deliberate choice. Excluding unbounded-risk trades would let the riskiest positions escape the measurement entirely, and those are the ones the number exists to catch. Instead kappi assumes a normal 2% risk and clamps the loss at 100% of the position — an estimate standing in where a plan should have been.

When more than half of the last 11 were sized that way, kappi says so privately, at most once a week. Not a penalty: an estimate covering more than half your trades means the score is substantially kappi's guess rather than your stated risk, and you should know that about your own number.

What are the bands for?

RME reads green, yellow or red. Slipping into yellow raises a heads-up on the profile and by email unless you have opted out; red is red. The bands exist because a percentile is a thing you check and a colour is a thing you notice, and drift is only worth catching if it reaches you before the drawdown does.

Steady green over a long stretch is the reading worth having, and not because it is high. It means a trader held their sizing through the rough patches, which is the behaviour that survives contact with a bad month.

How does this connect to a track record?

The idea that the right measure is a ratio of what you keep to what you staked is old. Kelly's 1956 paper set out how much of a bankroll to commit given an edge, and the whole result turns on growth per unit risked rather than on winning frequency.^[1] RME reads the same axis backwards: not how much you should have staked, but how much of what you did stake came back.

What a kept-to-risked ratio catches, and a win rate does not, has been measured twice over. Odean showed investors realise gains far more readily than losses,^[2] which inflates a win rate and deflates the size of what is kept. And of Taiwanese day traders followed from 1992 to 2006, under 1% predictably earned positive abnormal returns net of fees^[3] — a population whose win rates were often fine.

RME is only meaningful over a complete set of trades. Computed across a selection, it measures the selection — and unlike a win rate, it can be raised simply by omitting the trades where the stake was largest. A metric is a summary of a record, so it inherits every weakness of that record. Computed from trades selected after the fact, it is a number about the selection. kappi commits each trade before it resolves and publishes it on a Merkle-anchored log, so PnL, RME, correlation to SPX, mean R:R and trade count over 30, 100 and 200-day windows are computed over everything that was committed, losses included. $15/month to keep a record; reading one is free.