What is the K-Ratio?

Guidance last reviewed 2026-08-18

The K-ratio answers a question a win rate cannot: did this account grow steadily, or did it get there in one lucky jump? Two traders can finish the year with identical profit and have completely different K-ratios.

It was introduced by Lars Kestner in 1996 and turns up in trading journals and analytics platforms — often behind a higher-priced tier — because it takes an equity curve and reduces its shape to a single number.

In one sentence

The K-ratio is the slope of your equity curve divided by how much the curve wanders away from that slope. Rising fast scores well. Rising smoothly scores well. Doing both scores best.

What it actually calculates

Three steps, and none of them need market data. Everything comes from your own closed trades.

1. Plot the cumulative equity curve — your running account value over time. The original uses a log scale, so a 10% gain counts the same whether it happens early or late.
2. Fit a straight line through it using linear regression.
3. Divide the slope of that line by the standard error of the slope, then apply a scaling factor for the number of observations.

The slope says you made money. The standard error says how tightly the actual curve hugged that line. Divide one by the other and you get return per unit of inconsistency.

Same profit, two very different scores: an account that gains a little most months fits its trend line closely and scores well. An account that sits flat for ten months and then jumps has a large standard error around the same slope, and scores poorly.

What the number actually means

There is no 1-means-lucky, 0-means-skilled scale. That is the honest answer, and it is the thing most explanations skip.

What the value does tell you:

Why no universal threshold exists. Kestner revised the formula more than once — the 1996 original, and later versions that scale differently for the number of observations and for the reporting frequency. The core, slope over standard error, is the same in all of them. The scaling is not. Two platforms can report two different K-ratios for the same trades and both be correct by their own definition.

That is why any published rule of thumb should be treated with suspicion unless it names the version it applies to.

The one comparison that is always valid

Your own K-ratio, over two periods, computed the same way. If it was 0.9 last quarter and 0.4 this quarter, something about the consistency of your results changed — and that is worth investigating regardless of which formula produced the numbers.

"Could it have been luck?" — the honest answer

Slope divided by standard error is a t-statistic, the same quantity statisticians use to ask whether a trend is distinguishable from noise. That is where the "it tells you whether your results were luck" description comes from.

It is a reasonable intuition and it is not literally a probability. The K-ratio does not output "there is a 12% chance this was luck." The scaling factors applied on top of the t-statistic mean it cannot be read straight off a statistical table.

Probability of chance, and what it actually means

The statistic people usually have in mind here is separate from the K-ratio, though it comes from the same t-statistic idea. It is often labeled probability of chance, statistical significance or a p-value, and it is calculated from your trades directly:

t = average trade result ÷ (standard deviation of trade results ÷ √number of trades)

Convert that t to a p-value and you get a number some platforms show as "probability of random chance."

⚠️ It does not mean what almost everyone thinks it means. A p-value of 5% does not say there is a 5% chance your edge is luck. It says:

If you genuinely had no edge at all, results at least this good would turn up 5% of the time anyway.

That is a statement about a hypothetical world with no edge — not about the probability that you have one. The difference sounds pedantic and is the single most common misreading in the field.

Three reasons to hold it loosely in trading specifically:

• Trade results are not independent. A losing streak affects the next decision. Classical tests assume each observation is unrelated to the last, and trading breaks that.
• The distribution is not normal. Trading returns have fat tails — rare large moves are far more common than the maths assumes, which makes results look more significant than they are.
• Selection bias swallows everything. If you tested forty strategies and kept the one that worked, the p-value on that survivor is close to meaningless. Something had to come first.

So use it as a rough sanity check on sample size, not as a verdict. It is genuinely useful for one thing: telling you when you simply do not have enough trades to say anything at all.

The better tool for the same question

Monte Carlo resampling takes your own trades, reshuffles their order thousands of times, and shows the range of outcomes the same set of results could plausibly have produced. That gives a distribution you can read straight off — this many of ten thousand reshuffles ended worse than break-even, and the worst drawdown seen was this deep — rather than a single index number to be misinterpreted.

It makes no assumption about the shape of the distribution, because it uses your actual distribution. For a trader, that is usually the more honest answer.

Use the K-ratio for consistency. Use resampling for luck.

What moves it

Note the third row: time you did not trade still counts. The regression runs against time, so a month away from the market widens the standard error even though nothing went wrong.

Where it misleads

Small samples. A few dozen trades will produce a K-ratio, but one more win or loss moves it substantially. Sample size belongs next to the number every time it is shown.

It ignores how the money was made. An account that climbed steadily by risking far too much on every trade can score well, right up until the trade that ends it. The K-ratio describes the shape of the curve, not the risk taken to draw it.

It is retrospective. It says nothing about the next trade, or the next hundred.

It rewards smoothness, which is not always a virtue. Some legitimate strategies are inherently lumpy — a small number of large wins is the entire design. Those will score poorly by construction, and that is a property of the metric rather than a fault in the strategy.

Definitions

Where it sits among the other measures

None of them replaces the others, and none of them replaces looking at the equity curve itself. The K-ratio is a summary of that picture, and summaries lose things.

Nothing on this website is trading or investment advice. Chart Ding is a market clock and alarm tool — it does not recommend trades, evaluate strategies, or take account of anyone's circumstances. Nothing here should be construed as a recommendation or relied on as the basis for a trading decision. Consult a licensed professional before placing real money at risk. Trading involves risk of loss.