v0.54 Strategy Fails

[BradT7]

No final results after 800 Epochs and RBB Strategy Fails? Never had an error like this prior to v0.54 update.

    ======  EPOCH 799/800  ======
Locale9
first unevaluated: 4
2 individues removed due to equality
[2]
EPOCH 262   &2
Maximum profit 30.441   Average profit 28.829   
Minimum profit -13.754  Profit variation 7.827  
Population size 71  Max population size 67  
Avg trade number 27 Avg sharpe ratio 0.320  

Elder dies 0

Locale10
first unevaluated: 5
4 individues removed due to equality
[1]
EPOCH 237   &1
Maximum profit 15.938   Average profit 14.693   
Minimum profit -13.636  Profit variation 5.648  
Population size 86  Max population size 86  
Avg trade number 183    Avg sharpe ratio -0.438 

Elder dies 0

Epoch runs in 9.26 seconds;
Backend power [4.443584442138672]

4 selected;
1 selected;
Selecting 1+4 individues, random test;
Annual profit -23.524%
Testing Strategy of Locale9 @ EPOCH 263:

Strategy Fails.
4 selected;
1 selected;
Selecting 1+4 individues, random test;
Annual profit 13.115%
Testing Strategy of Locale10 @ EPOCH 238:

Strategy Fails.
[<datasetOperations.CandlestickDataset object at 0x7fbb38863cc0>, None]
4 selected;
1 selected;
Selecting 1+4 individues, random test;
Annual profit -33.267%
Testing Strategy of Locale9 @ EPOCH 263:

Strategy Fails.
4 selected;
1 selected;
Selecting 1+4 individues, random test;
Annual profit -6.190%
Testing Strategy of Locale10 @ EPOCH 238:

Strategy Fails.

        .RUN ENDS.

[BradT7]

Issue localized to one separate gekko/japonicus folder, error did not happen with a second gekko/japonicus folder I was backtesting on. I've deleted that folder and started fresh again, hopefully that will resolve.

[Gab0]

Hey, strat failed because of score below zero. japonicus will protect u against dangerous parameters unless you change genconf.showIndividualEvaluationInfo ;)

[BradT7]

What's odd is it was the same parameters as a successful run in another gekko/japonicus folder, just different dataset with different date ranges, it was actually the bull date range that this failure happened too, and the bear date range completed normally. I'll test again tonight with the fresh install and fresh bull dataset. :)

[BradT7]

Hey Gab, I've had negative results still spit out successful results in a manual gekko backtest. Just take a look at some past logs I've sent.
I did just get this error again with the bear dataset using 3 and 2 min candles, but 5 and 10 displayed results, I still would have been interested how negative the 3 and 2 would have been, or manually tweak the results. Perhaps it could still display the final result, even if it's negative. I just hate to see it do 800 epochs and I'm left with nothing. The problem with enabling that setting you mentioned, is I imagine I would need to look through 800 results to try and find the most favourable result, even if negative.
Thanks for all your work!