We have started the Trading Strategy podcast. Listen to us on YouTube and Spotify.

What Is K-Ratio?

The K-Ratio measures the growth and consistency of a trading strategy’s equity curve. It compares the slope of a fitted trend with the standard error of that slope. Lars N. Kestner introduced it in 1996 as a complement to Sharpe. Unlike the ordinary Sharpe ratio, it depends on the sequence of returns: rearranging the same returns changes the cumulative wealth path and can change the K-Ratio.

Calculation and formula variants

For positive portfolio values V_t, construct the logarithm of cumulative wealth and fit an ordinary least-squares regression with an intercept against consecutive, equally spaced observation numbers t. Kestner’s 2013 version is:

\[ \begin{align}\begin{aligned}y_t = \log(V_t / V_0) = a + bt + \epsilon_t\\K_{2013} = \frac{\hat b}{SE_{\mathrm{OLS}}(\hat b)} \frac{\sqrt{m}}{n}\end{aligned}\end{align} \]

Here n is the number of period-end observations fitted, m is the expected number of observations per calendar year, and SE is the standard error of the regression slope, not the standard deviation of period returns or the residual standard deviation alone. For monthly data, m = 12; daily equity-market data often use m = 252, while daily cryptocurrency data covering every calendar day use m = 365.

The Zephyr K-Ratio uses the unadjusted slope-to-standard-error ratio, b / SE(b); earlier Kestner formulas also use different adjustments. Always name the version, sampling frequency, and evaluation window when reporting results. For example, a raw ratio of 20 over 252 daily observations gives a 2013 K-Ratio of approximately 1.26, not 20. A negative value indicates a downward fitted trend. A zero standard error makes the ratio undefined; a nearly straight curve can produce an extremely large value.

Comparison with other performance metrics

K-Ratio describes trend consistency, not every aspect of investment risk. The following measures answer different questions and can rank the same strategies differently:

  • Sharpe compares average periodic excess return over the risk-free rate with the standard deviation of those excess returns. K-Ratio instead compares the trend in cumulative log wealth with the standard error of its fitted slope; it does not subtract a risk-free return. With a constant per-period risk-free rate, rearranging the same periodic returns leaves the ordinary sample Sharpe unchanged, but can change K-Ratio. For example, grouping losing periods together rather than spreading them among gains changes the wealth path without changing the return sample’s mean or standard deviation. This order-independence does not apply to HAC Sharpe, which also uses serial dependence in returns.

  • Sortino relates return above a specified target to downside deviation: shortfalls below a minimum acceptable return contribute to its risk measure, while returns above that target do not. K-Ratio is not downside-only. Deviations both above and below the fitted log-wealth trend contribute to the slope’s standard error. A profitable strategy whose gains arrive in occasional large bursts can therefore have a lower K-Ratio than a more steadily growing strategy, without necessarily being less attractive on a downside-risk basis.

  • Compound Annual Growth Rate (CAGR) measures annualised compounded growth from starting to ending wealth, regardless of the intervening path. Strategies with identical CAGR over the same window can have different K-Ratios. Conversely, a nearly straight but slowly growing log-wealth curve can produce a high K-Ratio even when its absolute return is too small to meet the investor’s objectives.

  • Calmar ratio typically divides CAGR by the absolute maximum drawdown. Like K-Ratio, it is path-dependent, but its risk denominator focuses on the single worst peak-to-trough loss rather than deviations throughout the fitted curve. K-Ratio does not directly measure the worst drawdown or time needed to recover from it; a high value is not a guarantee that a strategy respects a drawdown limit.

  • Ulcer Index summarises the depth and persistence of percentage drawdowns across the evaluation window. The Martin ratio divides excess return by this drawdown measure. Both focus on being below previous highs, whereas K-Ratio evaluates departures from a fitted trend, including upside departures. A curve can keep reaching new highs yet have an uneven growth rate and a modest K-Ratio; a low Ulcer Index does not require a straight log-wealth trend.

Use K-Ratio to ask whether compounded growth has been consistent, Sharpe or Sortino to assess return relative to periodic variability, and drawdown measures to assess the experience of losing and recovering capital. Their numerical values are not interchangeable and have no common scoring scale.

Use in algorithmic trading

Use it as an additional diagnostic when benchmarking strategies with similar mandates over the same dates, observation frequency, and formula convention. It helps distinguish relatively steady growth from an uneven path whose profits arrive in isolated bursts. Evaluate it alongside Compound Annual Growth Rate (CAGR), Sharpe, maximum drawdown, and Ulcer Index: a smooth curve does not establish that the absolute return is worthwhile or that losses are tolerable.

Calculate it from periodic portfolio returns after fees, slippage, and financing costs, including the changing value of open positions and adjusting for external deposits and withdrawals. A curve based only on closed trades or stale valuations can hide risk. Assess it on out-of-sample testing periods and across market regimes; optimising only the K-Ratio can encourage overfitting to a particular return path.

There is no universal threshold for a good K-Ratio across versions and strategies. It does not subtract a risk-free rate, measure alpha, or quantify unobserved tail losses. Although the raw ratio resembles a regression t-statistic, cumulative wealth observations are dependent; an ordinary OLS calculation is not, by itself, a valid test of statistical significance or evidence of future profitability.

Research and literature

See also