• 概要
  • レビュー
  • コメント

LTPD Statistic Calculator

LTPD Library


This Library Calcualtes LTPD.


Introduction


How to distinguish good signals or EAs which needs winning-ratio calculations?

I cannot distinguish it, I will not say an EA is good or not.

But I can say, I have mathematical and statistical methods to help developing or optimizing EAs.


What is Reliability ?


For Example, "A" EA won all of 5 trades, "B" EA won 9 trades from 10 trades.

If I only compare the winning ratio with LTPD (Lot Tolerance Percent Defective), results are below :

LTPD calculates statistical failure ratio.


FailsTestedConfidence(%)LTPD Results(%)
0 (A)59046.05
1 (B)109038.9


This Results says that "B" EA's result is more reliable than "A" EA when only compared with winning ratio.

I expected "A" is more good than "B".


Why "B" is more Stable than "A"

With Very Narrow Vision, if I see only simple results above, "B" is more stable than "A" because "B" is tested many times.

I don't have to compare tested times, because computers can do it.


Exact Prediction vs Prediction Calculations.


There is no Guranteed Statistical Methods in Mathematics.

FailsTestedConfidence(%)LTPD Results(%)
0 (A)599.999230.26
1 (B)1099.999142.37

I set very high confidence and recalculated LTPD. Because I want very very Exact Future Prediction.

I got meaningless results.


But the mathematics can predict the future within confidence. I recommend 95% (For Non Optimized Results) or 99% (For Optimized Results).

When you use 95% confidence, 5% risk still exists that the real-time's failure ratio is over the test's LTPD ratio.


I am assuming that the EA has no randomness situations on it. (Errors, Cheating etc).

Not Guaranteed, But can be predicted.


Using This Library


returnLTPD Function Definition : 

double returnLTPD(double fail,double try,double confidence)


Failure Ratio within Confidence = LTPD (%)

Success Ratio within Confidence = 100 - LTPD (%)


Examples :

#import "LTPD.ex4"

double returnLTPD(double fail,double try,double p);

#import

int OnInit()
  {

//Print Fail 0 Tested 5 Confidence 95% Failure Ratio
   Print(DoubleToString(returnLTPD(0,5000,95.0),2)+" %");

//Print Fail 1 Tested 50 Confidence 95% Success Ratio

   Print(DoubleToString(100-returnLTPD(1,50,95.0),2)+" %");

   return(INIT_SUCCEEDED);

  }


レビューなし