K-ratio by Lars Kestner : which value desirable


Hi All,
after some years of testing with some Trading systems made by me, the best Value for K-RATIO i get is 0.12 with backtesting from 2000 to 2017 on 100 stocks.

Wich number are desiderable, wich number you get in your experience ?

regards luca


@lanas I don’t have a direct answer for you other than larger is better. This may have been discussed in the old forum.

But if you do some research into the K-Ratio you will be able to better compare systems and increase your knowledge about this useful measure of system performance.

Lars Kestner first introduced this measure in an article in TASC in ~ 1996. He wanted to know how smooth the equity curve of a system was, and here is where he claimed "I have developed a new method of evaluating performance that is more robust than current popular techniques. This method, which maintains the idea of measuring reward as compared with risk, utilizes more advanced statistical techniques to quantify performance. Rather than simply looking at returns independently, consistency of results through time will be the focal point of this new performance method. "

But the problem is the K-Ratio has had its formula changed twice since then. 2003 and 2013. I think AmiBroker is using the 2003 version (not the best IMHO, but then again I don’t look at K-Ratio often)

Zephyr Associates, Inc. put there spin on it in 2003

And Lars Kestner decided to “improve” it in 2013.

A summary of the three version found on-line,

Hope that is of some help to you.

Metrics - RAR%, RRRR, R-Sharpe

Hi Portfoliobuilder
thnks for answer me, it 's hepful.

We use in amibroker the oldest formula of K-ratio. Here you can see the line of my TS from 2000 to yesterday :

This line, for me is a good line …very consistency. But the score is “only” 0.12. If i start to backtest fromn 2003 the line is more more better, but the K-ratio valure is only 0.13. So, the question is… with the oldest formula, the objective is 1.0 ? or 5 ??? Seem impossible for me to arrive over 0.20…



Solved :

for the same result the different forumula gave very different results. In amibroker we use 2003 version formula, so i 0.12 are very good result.

k-ratio (1996) = 1.65

k-ratio (2003) = 0.0953

k-ratio (2013) = 1.51


Thanks very much for this material. The final (2013) version is the only one that IMHO could be useful PROVIDED THAT logarithms of values are used. That is because a linear regression loses contact with any non-linear trend, somewhere. With that in mind, using older K-ratio over any longish period is likely to give low values.