Actually -- a change between 0.2 and 0.3 makes little difference -- just my opinion.  However, more importantly, the control uses a Windows NumericUpDown control.  So, now, if I want to increase the value from 0.2 to 0.3 I click once. But if Popov increases the resolution 10-fold then I have to click 10-times more -- and that would be a pain.  I'm much more likely to adjust the Sharpe ratio from 0.2 to 0.5 (3 clicks) than I am from 0.21 to 0.49 (28 clicks).  In general, I think simple is good and tend to resist added complexity without seeing evidence to support it, especially when you consider the current implementation of Sharpe ratio works well.

True -- I could simply enter a value.  But then, again, so could you -- and it wouldn't require Popov spending time making changes and testing it.  As you know, programming takes time -- and I'd prefer Popov spend as much time as possible implementing new and interesting features that benefit everyone, rather than customizing the software to suit one person.  The point wasn't about clicking -- the point was taking time away from him by requesting an inconsequential change to the UI that would affect everyone else simply to suit one person's preference.

Are you really suggesting because it doesn't support Sharpe ratios to 2 decimal places it's not a serious trading application?  Do you have any Live trading results or evidence that shows a marked difference between 0.14 and 0.18?  Seriously -- four one/hundredths makes a difference in a Live account?

You are right about one thing -- I am lazy, but I consider that a virtue and work very hard at it.

I don't think the behavior is ambivalent.  Back testing results will commonly report Sharpe ratios to two decimal places because those are computed values.

The UI control you are referring to under the Acceptance Criteria is called Minimum Sharpe ratio -- which implies strategies whose Sharpe ratio is below that minimum will fail the check.  In no way does that imply the resulting, computed value also has to be one decimal place.  The reason you can filter to two decimal places is because you are filtering computed values.

I suspect Popov has better things to do than play referee...

The MINIMUM value you set in the Acceptance Criteria is not computed -- you enter it, right?

The filter you mention is used to filter (i.e. sort) strategies in a Collection based on their statistics -- those statistics were computed by the back testing engine, right?

It's been awhile since I've had statistics, but there is a rule about how many significant digits are valid.  I mean, why just use 2?  Why not 5 or 17?  The number of significant digits implies accuracy.  For example, if you Google the time to drive from LA to NY it says 41 hours.  It doesn't say 41.372891 hours.  That's because if you were to make the drive 100 times it wouldn't come out 41.372891 hours each time, which is why the value is rounded to the nearest hour.  The number of significant digits that are statistically valid is related to how reproducible a measurement is.

When it comes to Sharpe ratio we all know very small changes in the data horizon can effect significant changes in the ratio -- hence, to think that using a MINIMUM of 0.18 versus 0.14 in the Acceptance Criteria will result in significantly better strategies is misleading.  It's just not that accurate, especially when you take into account how optimization works.  A difference of 0.5 to 0.2 -- sure, but not 4 one-hundredths.  I'm not making this up -- it's basic statistics.

If you don't believe me -- check out the Monte Carlo feature.  Do you see how minor perturbations can effect large changes in the equity curve?  The statistics of those equity curves can be quite different and their Sharpe ratios for sure differ by more than 4 one-hundredths.

Perhaps you misunderstand the purpose of Acceptance Criteria versus filtering -- or perhaps I do, I don't know.  Acceptance Criteria are used to determine which strategies make it into the collection in the first place.  Filtering is used to sort strategies that are members of the Collection (i.e. have already passed all Acceptance Criteria) and for whom their back testing statistics have been computed.  Acceptance Criteria and collection filtering are two different concepts.

Hi Mr. Popov,

the only problem I had with this was that you can enter fractional values like 0.14 (and they are important for shorter term strategies the way I generate them, even if you don´t think it is important at all, I guarantee you that it is for me and it indeed makes a huge difference here) on the Sharpe ratio filter in the "Collection" tab, but not within the "Settings -> Acceptance Criteria", where one can only enter values like 0.1 or 0.2. Is it really such a huge problem to allow fractional values? I mean your program calculates the Sharpe ratio to 2 decimal values, why not let users enter it like that then too? Especially since this is a code change that takes barely a few minutes as I know from my own Javascript coding.

@sleytus: sorry, can´t read you, you are on my ignore list since the previous post, no time to waste on my end here with your lengthy discussions about trivial topics :-) Thanks for your understanding.