forked from mnbvc188199/Warrior_EA
e9c562b39f
- Introduced `FeatureScale.mqh` with `FeatSquash` function for stateless feature scaling. - Added `RegimeMath.mqh` class for regime arithmetic, including efficiency and variance calculations. - Documented the Mind trading logic in `MIND.md`, detailing the trading process and modes. - Created `VOLNORM_PLAN.md` and `VOLNORM_RESULTS.md` for tick-volume normalization testing. - Implemented `read_book.py` for analyzing trade book data and correlations. - Developed `volnorm.py` for testing tick-volume normalization with new and old methods.
53 lines
2.4 KiB
MQL5
53 lines
2.4 KiB
MQL5
//+------------------------------------------------------------------+
| |
//| BetSizer.mqh |
| |
//| AnimateDread |
| |
//| |
| |
//| FROM A PROBABILITY TO A SIZE (AFML ch. 10). |
| |
//| |
| |
//| With p = P(this bet wins), the statistic z = (p - 1/2) / |
| |
//| sqrt(p (1 - p)) is the test of "is this better than a coin", and |
| |
//| m = 2 Phi(z) - 1 is the bet size it earns: 0 at p = 0.5, rising |
| |
//| smoothly to 1. Size follows conviction continuously instead of on |
| |
//| a cliff, and a coin-flip gets no capital at all. |
| |
//| |
| |
//| HOW THIS BOOK USES IT. The base risk (RISK_0_25) is what the |
| |
//| validated book earns at its AVERAGE win rate p0. A setup the |
| |
//| journal rates below average gets m(p) / m(p0) of that risk - so |
| |
//| the sizer can only take risk OFF, never add it. That is deliberate |
| |
//| on a prop account: the 5% limit is a ceiling, and a model that is |
| |
//| wrong about a setup should be wrong on the safe side. Sizes are |
| |
//| stepped to 0.1 so a fluctuating estimate does not create a |
| |
//| different lot every time. |
| |
//+------------------------------------------------------------------+
| |
#ifndef WARRIOR_BETSIZER_MQH
| |
#define WARRIOR_BETSIZER_MQH
| |
| |
#include "EdgeStats.mqh"
| |
| |
class CBetSizer
| |
{
| |
public:
| |
//--- m(p), AFML eq. 10.1. 0 for p <= 1/2 (no edge, no bet).
| |
static double Size(const double p)
| |
{
| |
if(p <= 0.5)
| |
return 0.0;
| |
if(p >= 0.999)
| |
return 1.0;
| |
const double z = (p - 0.5) / MathSqrt(p * (1.0 - p));
| |
return 2.0 * CEdgeStats::Phi(z) - 1.0;
| |
}
| |
//--- The multiple of the base risk for a setup rated p, given the book's average p0.
| |
//--- Never above 1. 0 means "the journal rates this a coin flip or worse".
| |
static double RiskScale(const double p, const double p0)
| |
{
| |
const double base = Size(p0);
| |
if(base <= 0.0)
| |
return 1.0; // the book has no measurable edge to scale against
| |
if(p >= p0)
| |
return 1.0;
| |
const double m = Size(p) / base;
| |
return MathFloor(MathMin(m, 1.0) * 10.0 + 0.5) / 10.0;
| |
}
| |
};
| |
#endif // WARRIOR_BETSIZER_MQH
|