private StarterInfo CreateStarterInfo(Horse horse, Dictionary<Horse, StarterInfo> map)
{
if (horse.Scratched)
{
return null;
}
var dummyStarterInfo = new StarterInfo();
dummyStarterInfo.RacingDate = horse.Parent.Parent.Date;
dummyStarterInfo.Distance = horse.Parent.CorrespondingBrisRace.DistanceInYards;
StarterInfo currentStarterInfo = dummyStarterInfo;
foreach (var pp in GetPastPerformancesWithValidGoldenFigures(horse))
{
StarterInfo si = StarterInfo.CreateFromBrisPastPerformance(pp);
currentStarterInfo.Previous = si;
si.Next = currentStarterInfo;
currentStarterInfo = si;
}
if (dummyStarterInfo.CountPreceding < 3)
return null;
map.Add(horse, dummyStarterInfo);
return dummyStarterInfo;
}
public double GetOpinion(StarterInfo si)
{
if (si.CountPreceding < _numberOfPastPerformances)
return 0;
var v1 = StarterInfo.NormalizedInputForMemoryLearning(si, _numberOfPastPerformances);
_patterns.ForEach(p=>p.CalculateEuclidianDistances(v1));
_patterns.Sort(CompareEuclidianDistances);
double positive = 0, negative =0;
int desiredOutcomeIndex = _patterns[0].X.Count - 1;
for(int i = 0; i < 2; ++i)
{
if(_patterns[i].X[desiredOutcomeIndex] == 1)
{
++positive;
}
else
{
++negative;
}
}
return positive >0 ? 1.0 : 0.0;
}
internal static List<double> NormalizedInputForMemoryLearning(StarterInfo si, int ppCount)
{
var input = new List<double>();
if (si.CountPreceding >= ppCount)
{
si.GetPreviousPerformances(ppCount).ToList().ForEach(si2 => input.Add(si2.GoldenFigure));
double min = input.Min();
double max = input.Max();
for (int i = 0; i < input.Count; ++i)
{
input[i] = (input[i] - min) / max;
}
PopulateNormalizedDataForMemoryLearning(si, input, ppCount);
input.Add(NormalizeDaysOff(si));
input.Add(NormalizePostPosition(si));
input.Add(NormalizeDistance(si));
input.Add(NormalizeBoolean(si.RestrictedToFemalesOnly));
input.Add(NormalizeBoolean(si.RestrictedToStateBredsOnly));
input.Add(NormalizeBoolean(si.WasClaimed));
input.Add(NormalizeBoolean(si.BlinkersOff));
input.Add(NormalizeBoolean(si.FirstTimeLasix));
input.Add(NormalizeBoolean(si.RestrictedToTwoYearsOnly));
input.Add(NormalizeBoolean(si.RestrictedToThreeYearsOnly));
double desiredValue = si.GoldenFigure > max ? 1.0 : 0.0;
input.Add(desiredValue);
}
return input;
}
public static bool CheckIfStarterWillImprove(StarterInfo si)
{
if(null == _singleton)
{
_singleton = new MemoryBasedNeuronNetwork(4);
_singleton.AddTrainingPatterns(StarterInfo.LoadStartersFromFile(@"C:\Users\John\Desktop\neural_network.csv", 4));
}
return _singleton.GetOpinion(si) > 0;
}
internal static List<double> NormalizedInputForImprovementDetection(StarterInfo si, int ppCount)
{
var input = new List<double>();
var gf = new List<double>();
si.GetPreviousPerformances(ppCount).ToList().ForEach(si2 => gf.Add(si2.GoldenFigure));
double min = gf.Min();
double max = gf.Max();
PopulateNormalizedData(si, input, min, max, ppCount);
input.Add(NormalizeDaysOff(si));
input.Add(NormalizePostPosition(si));
input.Add(NormalizeDistance(si));
return input;
}
internal static StarterInfo CreateFromBrisPastPerformance(BrisPastPerformance pp)
{
if (null == pp || pp.GoldenFigureForThisHorse == -999)
return null;
var si = new StarterInfo { Previous = null, Next = null };
si.HorseName = pp.Parent.CorrespondingHorse.Name;
si.RaceId = -9999;
si.TrackCode = pp.Parent.CorrespondingHorse.Parent.Parent.TrackCode;
si.RacingDate = Utilities.GetDateInYYYYMMDD(pp.Date);
si.FinishPosition = Convert.ToInt32(pp.FinalPosition);
si.Odds = pp.OddsAsDouble;
si.GoldenFigure = (int)pp.GoldenFigureForThisHorse;
if (si.GoldenFigure < 0)
{
si.GoldenFigure = 1;
}
si.WinnersGoldenFigure = (int)pp.GoldenFigureForTheWinner;
si.TrackCondition = pp.TrackCondition;
si.Distance = pp.DistanceInYards;
int scp;
int.TryParse(pp.SecondCallPosition, out scp);
si.SecondCallPosition = scp;
return si;
}
private StarterInfo[] GetPreviousPerformances(int count)
{
var array = new StarterInfo[count];
var si = Previous;
for (int i = 0; i < count; ++i)
{
array[i] = si;
if (null != si)
{
si = si.Previous;
}
}
return array;
}
private static void PopulateNormalizedDataForMemoryLearning(StarterInfo starterInfo, List<double> input, int count)
{
starterInfo.GetPreviousPerformances(count).ToList().ForEach(
si =>
{
input.Add(NormalizeTrackCondition(si));
input.Add(NormalizeOdds(si));
input.Add(NormalizeDistance(si));
input.Add(NormalizePosition(si.FinishPosition));
input.Add(NormalizePosition(si.SecondCallPosition));
input.Add(NormalizeDaysOff(si));
input.Add(NormalizeRaceType(si));
input.Add(NormalizePostPosition(si));
input.Add(NormalizeBoolean(si.RestrictedToFemalesOnly));
input.Add(NormalizeBoolean(si.RestrictedToStateBredsOnly));
input.Add(NormalizeBoolean(si.WasClaimed));
input.Add(NormalizeBoolean(si.BlinkersOff));
input.Add(NormalizeBoolean(si.FirstTimeLasix));
input.Add(NormalizeBoolean(si.RestrictedToTwoYearsOnly));
input.Add(NormalizeBoolean(si.RestrictedToThreeYearsOnly));
});
}
internal static List<double> NormalizedInputToChooseBetweenFirstAndSecondFavorite(StarterInfo favorite, StarterInfo secondFavorite, StarterInfo thirdFavorite, int ppCount)
{
var input = new List<double>();
var gf = new List<double>();
favorite.GetPreviousPerformances(ppCount).ToList().ForEach(si => gf.Add(si.GoldenFigure));
// favorite.GetPreviousPerformances(ppCount).ToList().ForEach(si => gf.Add(si.WinnersGoldenFigure));
secondFavorite.GetPreviousPerformances(ppCount).ToList().ForEach(si => gf.Add(si.GoldenFigure));
//secondFavorite.GetPreviousPerformances(ppCount).ToList().ForEach(si => gf.Add(si.WinnersGoldenFigure));
thirdFavorite.GetPreviousPerformances(ppCount).ToList().ForEach(si => gf.Add(si.GoldenFigure));
double min = gf.Min();
double max = gf.Max();
PopulateNormalizedData(favorite, input, min, max, gf.Average(),ppCount);
PopulateNormalizedData(secondFavorite, input, min, max, gf.Average(), ppCount);
PopulateNormalizedData(thirdFavorite, input, min, max, gf.Average(), ppCount);
input.Add(NormalizeDaysOff(favorite));
input.Add(NormalizeDaysOff(secondFavorite));
input.Add(NormalizeDaysOff(thirdFavorite));
//input.Add(NormalizeOdds(favorite));
//input.Add(NormalizeOdds(secondFavorite));
//input.Add(NormalizeOdds(thirdFavorite));
//input.Add(NormalizePostPosition(favorite));
//input.Add(NormalizePostPosition(secondFavorite));
//input.Add(NormalizePostPosition(thirdFavorite));
//input.Add(NormalizeDistance(favorite));
return input;
}
private static double NormalizeTrackCondition(StarterInfo si)
{
return si.TrackCondition.Trim().ToUpper() == "FT" ? 0.3 : 0.7;
}
static double NormalizeRaceType(StarterInfo si)
{
switch(si.RaceType)
{
case RaceType.GRADE_1:
case RaceType.GRADE_2:
case RaceType.GRADE_3:
case RaceType.STAKES:
return 0.9;
case RaceType.ALW:
return 0.67;
case RaceType.MSW:
return 0.52;
case RaceType.CLM:
return 0.17;
case RaceType.MCL:
return 0.02;
default:
return 0.02;
}
}
static double NormalizePostPosition(StarterInfo si)
{
if(si.PostPosition == 1)
{
return 0.8;
}
else if (si.PostPosition == 2)
{
return 0.7;
}
else if (si.PostPosition >= 10)
{
return 0.1;
}
else if (si.PostPosition >= 8)
{
return 0.2;
}
else
{
return 0.3;
}
}
private static double NormalizeOdds(StarterInfo si)
{
if (si.WasTheFavorite)
return 0.9;
if (si.Odds <= 3.0)
{
return 0.6;
}
else if (si.Odds <= 5.0)
{
return 0.2;
}
else
{
return 0.02;
}
}
private static double NormalizeDistance(StarterInfo si)
{
return si.Distance < 1780 ? 0.3 : 0.7;
}
private static double NormalizeDaysOff(StarterInfo si)
{
int layoff = si.LayoffInDays;
if (layoff == 0)
return 0.05;
else if (layoff < 10)
return 0.8;
else if (layoff < 30)
return 0.5;
else if (layoff < 80)
return 0.25;
else if (layoff < 120)
return 0.15;
else
return 0.10;
}
private static StarterInfo CreateFromString(string[] s, int stringArrayIndex, StarterInfo previous)
{
if (stringArrayIndex >= s.Count())
return null;
var si = new StarterInfo {Previous = previous, Next = null};
si.HorseName = (null == previous) ? s[stringArrayIndex++] : previous.HorseName ;
si.RaceId = Convert.ToInt32(s[stringArrayIndex++]);
si.TrackCode = s[stringArrayIndex++];
si.RacingDate = s[stringArrayIndex++];
si.FinishPosition = Convert.ToInt32(s[stringArrayIndex++]);
si.Odds = Convert.ToDouble(s[stringArrayIndex++]);
si.GoldenFigure = Convert.ToInt32(s[stringArrayIndex++]);
if (si.GoldenFigure < 0)
{
si.GoldenFigure = 1;
}
si.WinnersGoldenFigure = Convert.ToInt32(s[stringArrayIndex++]);
si.TrackCondition = s[stringArrayIndex++];
si.Distance = Convert.ToInt32(s[stringArrayIndex++]);
si.SecondCallPosition = Convert.ToInt32(s[stringArrayIndex++]);
si.RaceType = DeserializeRaceType(s[stringArrayIndex++]);
si.PostPosition = Convert.ToInt32(s[stringArrayIndex++]);
DeserializeRestrictions(si, s[stringArrayIndex++]);
si.FieldSize = Convert.ToInt32(s[stringArrayIndex++]);
si.WasTheFavorite = Convert.ToInt32(s[stringArrayIndex++]) == 1;
var next = CreateFromString(s, stringArrayIndex, si);
si.Next = next;
return si;
}
private static void DeserializeRestrictions(StarterInfo si, string s)
{
s = s.Trim();
if (s.Length == 7)
{
si.RestrictedToFemalesOnly = (s[0] == '1');
si.RestrictedToStateBredsOnly = (s[1] == '1');
si.WasClaimed = (s[2] == '1');
si.BlinkersOff = (s[3] == '1');
si.FirstTimeLasix = (s[4] == '1');
si.RestrictedToTwoYearsOnly = (s[5] == '1');
si.RestrictedToThreeYearsOnly = (s[6] == '1');
}
}
private static void PopulateNormalizedData(StarterInfo starterInfo, List<double> normalizedData, double min, double max, double average, int count)
{
starterInfo.GetPreviousPerformances(count).ToList().ForEach(
si =>
{
normalizedData.Add(NormalizeGoldenFigure(si.GoldenFigure, min, max, average));
// normalizedData.Add(NormalizeGoldenFigure(si.WinnersGoldenFigure, min, max, average));
normalizedData.Add(NormalizeTrackCondition(si));
// normalizedData.Add(NormalizeOdds(si));
normalizedData.Add(NormalizeDistance(si));
//normalizedData.Add(NormalizePosition(si.FinishPosition));
/*
normalizedData.Add(NormalizePosition(si.SecondCallPosition));
normalizedData.Add(NormalizeDaysOff(si));
normalizedData.Add(NormalizeRaceType(si));
normalizedData.Add(NormalizePostPosition(si));
normalizedData.Add(NormalizeBoolean(si.RestrictedToFemalesOnly));
normalizedData.Add(NormalizeBoolean(si.RestrictedToStateBredsOnly));
normalizedData.Add(NormalizeBoolean(si.WasClaimed));
normalizedData.Add(NormalizeBoolean(si.BlinkersOff));
normalizedData.Add(NormalizeBoolean(si.FirstTimeLasix));
normalizedData.Add(NormalizeBoolean(si.RestrictedToTwoYearsOnly));
normalizedData.Add(NormalizeBoolean(si.RestrictedToThreeYearsOnly));
*/
});
}