Metrics ... gain per volume, cagr, gain per negative volume area, etc

[Matthew Simon Gream]

Hi,

Is anyone else working or experienced in advanced metrics? I have generated some analytics (that can analyse signals on this platform based on being able to export the historical trade data) that include metrics like:

gain per volume: normalising gains (e.g. what % gain per 0.01 lot ... including swap and commission), so as to normalise across signals (irrespective of currency units, account size, trade size, etc)
... and this metric in consideration of the total number or average number of trade hours for that signal (how long the trades are open) ... so to identify those that achieve the best gain per volume in the shortest amounts of time
cagr: on weekly and monthly basis ... again, also in consideration of the average number of trade hours / etc

The most interesting metrics I have developed is to look at all of those metrics (gains and returns) PER negative volume minimum and area, what this means is that I compute the amount of time that the trades have sat in negative equity (and the worst negative equity they reach) ... effectively the sum of the area under the curve, and the metrics against that, i.e. to select signals that give the best returns for the least amount of negative equity (e.g. risk) incurred. 

Also lots of known and standard metrics including max volumes traded, etc (e.g. aggressive averaging trades).

Also lots of cheating metrics: e.g. a signal that MQL5 says has been around for 60 weeks, but when you do the analysis, it only traded for 13 weeks of those: instant discard. 

These metrics were not easy to compute and required experimentation: I had to effectively use historical 1M data to replay historical trades and track second by second positions (and therefore, negative equity, etc). Basically a lot of nodejs running in docker containers in the cloud. It also required a lot of debugging and manual analysis and verification, and looking at error margins (e.g. when the trade history of the signal says a close price, but the 5M historical data at that point does not have the same close price in range).  

I've done this for all 30K active signals on the platform so I can develop insights and rankings. 

It seems to work. It's not perfect or silver bullet, nothing is, but it closes the gap further and improves "natural selection" to continue to evolve forward.

But I'm interested in people who know about or want to discuss advanced metrics like this, ones that that go beyond those computed by the MQL5 platform. Most people are just looking at standard metrics (which hide issues like negative equity loading, etc).

Would love to learn, discuss and share with like minded people.

Please connect or point me places.