Ok, stupid question here... I only have a vague understanding of "prediction error" in statistics. But since this figure seems to bounce all over the place in my test results, I'd like to know the best way to read this for the trading purposes.

I'm pretty sure this figure is only useful if the test data being analyzed has enough volume to be considered "statistically significant". I'm not sure exactly at what point that occurs. (Sidenote: if volume is not enough to create a useful "prediction error" -- then perhaps Zorro should not display it at all in those cases?)

So from a trader's perspective... what I'd like to know is... is there a way to read this figure and make some sort of judgments about a strategy based on this figure? For example we know from the manual that a tradeable strategy should have a Sharpe Ratio of > 1.

Btw, here is what the manual says about this stat:

You'll have to ask jcl how the prediction error is calculated in detail, but generally speaking it's a figure which shows you how likely your results can be reproduced in real trading.
For instance, if you have a strategy with a SR of 2, a UI of 3% and a prediction error of 200%, you might not get as good results in live trading as your results are likely to be influenced by a lucky simulation period.

Your prediction error should go down with the length of your test period and the number of trades.

In general I would aim for what the Z strategies have, that's < 30%. Z5 has a prediction error of 45%, though, but Z5 is kind of special since it does not rely on technical analysis.

blaub4r

The prediction error is an empirical value. All parameters that are known to have an effect on prediction accuracy, such as the number of trades or the profit distribution, are put together for calculating the error term. For instance, if the profit comes from only a few large wins, the strategy will have a relatively large error. Oversampling will normally reduce the error.