search

owocki / pytrader

cryptocurrency trading robot
MIT License
1.93k stars 432 forks source link

`/admin/optimize` graphs are not helpful #38

Open owocki opened 8 years ago

owocki commented 8 years ago

These graphs:

were originally designed as a way to test whether trade.py parameters were correctly predicting the market, but they're just not all that useful. We need to come up with a better way to gauging how well our trade parameters are following the market.

NDuma commented 8 years ago

Finding the conditions where it was correct 11% As well as what was causing the other 89% is important.

Is there away to aggregate 'meta' conditions on these trades; or, find similarities in conditions and execution?

owocki commented 8 years ago

I'm not sure what you mean by 'conditions and executions'.

The meta information at the top of admin/*_charts, median, low, high, average, are the aggregated values for the group of selected tests.

If you scroll down admin/*_charts you will see the distribution of success for each tuning parameter .