Here are examples of how I've tried some ideas using genetic algorithms and genetic programming.
https://www.elitetrader.com/et/thre...your-edge-for-2019.329802/page-8#post-4809209
https://www.elitetrader.com/et/thre...ine-learning-for-astronomical-profits.334373/
https://www.elitetrader.com/et/threads/oscillators.337471/page-23#post-4960785
https://www.elitetrader.com/et/threads/machine-learning-for-price-wave-analysis.339646/#post-4997882
https://www.elitetrader.com/et/thre...-prediction-models.357511/page-3#post-5368574
https://www.elitetrader.com/et/thre...ession-model-experiments.357998/#post-5373435
I currently use genetic optimization of floating point values for hand-coded rules like
Code:
$R0 = $R1 = 0;
if ( $cyFitProp_lo >= 0.22716 ) { if ( $cyFitProp_lo <= 0.732629 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $cyFitProp_hi >= 0.29458 ) { if ( $cyFitProp_hi <= 0.904422 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $cphase0_lo >= 2.8355 ) { if ( $cphase0_lo <= 3.95609 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $cphase0_hi >= 2.80795 ) { if ( $cphase0_hi <= 3.76181 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $rlh_per0 >= 0.316866 ) { if ( $rlh_per0 <= 1.02584 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $rlh_phase0 >= 0.000702903 ) { if ( $rlh_phase0 <= 6.28273 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $detrPropNextBar_lo >= 2.92023e-09 ) { if ( $detrPropNextBar_lo <= 0.32465 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $detrPropNextBar_hi >= 8.10645e-06 ) { if ( $detrPropNextBar_hi <= 0.635021 ) { $R1 = 1 + $R1 ; } }
$R0 = 1 + $R0 ;
if ( $R1 >= $R0 ) { $return = 1 ; }
When the optimized rules perform well on in- and out-of-sample data, I consider combining them with other (base) rules to try to improve the performance of using the base rules alone. The final combination would be forward tested and might be used for actual trades.
This is in response to something
@stochastix posted but deleted (it's always possible I imagined that too
).
That code listing described the genetic optimization
after other software created functions that model price movement like:
Code:
y_hi = 110.319160461426 - 0.0421270467340946 * x
+ 7.39092588424683 * skewed_cos(twopi / 88.1613007177421, 0.646005928516388, -0.747746825218201, x, 100) ;
y_lo = 107.114105224609 - 0.042427085340023 * x
+ 7.60985136032104 * skewed_cos(twopi / 87.9424281352233, 0.619945704936981, -0.802092313766479, x, 100) ;
where y_hi and y_lo are predicted high and low prices of a particular asset
x is the number of calendar days past the start of a sampling period
skewed_cos(freq, phase, skew, x, iter) = cos(freq * x + phase) when iter == 0
skewed_cos(freq, phase, skew, x, iter) = cos(freq * x + phase + skew * skewed_cos(freq, phase, skew, x, iter - 1)) when iter > 0
The parameters whose values are genetically-optimized are based on how good the fit to prices was, the current phase of the cyclic part of the functions, and the relationships between qualities of the two functions (period and current phase in the example). These same genetically-optimized parameters apply to more than one asset, but the functions for any asset depend on that asset's prices and would be different than the functions for another asset. And, the functions get recalculated daily to use the most recent sampling period.
As far as determining resources to allocate to a trade, I try to keep individual trades to have similar risk level as in
this post.
