Yes, I know that it is impossible to predict lottery results. But if the results would be predictable? Of course, I know they are not, but just hypothetically, for a play with Perceptron Neural Network and prediction algorithms, I want to show you that in 250 lines of code, you can play with prediction thanks to the Encog 3.3 library. I recently bought 2 of Jeff Heaton’s books about neural networks and C# and his library. That inspired me to play with prediction, forecast, trends, classifications, regression, and more… I do not want to bring you too many details and share those 250 lines of code. To make it work in Visual Studio, you have to just download by NuGet encoding-dotnet-core 3.3 library and include it in the console application with that code. But please remember, predicted values are correct in math, but it is impossible to predict truly random lottery results, which are Polish Lotto. One more thing: good results can take some time, sometimes a few hours, so be patient when you run it. Thanks for reading!
namespace LottoPredictor
{
using System;
using System.Collections.Generic;
using System.IO;
using System.Net;
using Encog.Engine.Network.Activation;
using Encog.ML.Data.Basic;
using Encog.Neural.Networks;
using Encog.Neural.Networks.Layers;
using Encog.Neural.Networks.Training.Propagation.Resilient;
class LottoResult
{
public int V1 { get; private set; }
public int V2 { get; private set; }
public int V3 { get; private set; }
public int V4 { get; private set; }
public int V5 { get; private set; }
public int V6 { get; private set; }
public LottoResult(int v1, int v2, int v3, int v4, int v5, int v6)
{
V1 = v1;
V2 = v2;
V3 = v3;
V4 = v4;
V5 = v5;
V6 = v6;
}
public LottoResult(double[] values)
{
V1 = (int)Math.Round(values[0]);
V2 = (int)Math.Round(values[1]);
V3 = (int)Math.Round(values[2]);
V4 = (int)Math.Round(values[3]);
V5 = (int)Math.Round(values[4]);
V6 = (int)Math.Round(values[5]);
}
public bool IsValid()
{
return
V1 >= 1 && V1 <= 49 &&
V2 >= 1 && V2 <= 49 &&
V3 >= 1 && V3 <= 49 &&
V4 >= 1 && V4 <= 49 &&
V5 >= 1 && V5 <= 49 &&
V6 >= 1 && V6 <= 49 &&
V1 != V2 &&
V1 != V3 &&
V1 != V4 &&
V1 != V5 &&
V1 != V6 &&
V2 != V3 &&
V2 != V4 &&
V2 != V5 &&
V2 != V6 &&
V3 != V4 &&
V3 != V5 &&
V3 != V6 &&
V4 != V5 &&
V4 != V6 &&
V5 != V6;
}
public bool IsOut()
{
return
!(
V1 >= 1 && V1 <= 49 &&
V2 >= 1 && V2 <= 49 &&
V3 >= 1 && V3 <= 49 &&
V4 >= 1 && V4 <= 49 &&
V5 >= 1 && V5 <= 49 &&
V6 >= 1 && V6 <= 49);
}
public override string ToString()
{
return string.Format(
"{0},{1},{2},{3},{4},{5}",
V1, V2, V3, V4, V5, V6);
}
}
class LottoListResults : List<LottoResult> { }
class Program
{
static void Main(string[] args)
{
var wc = new WebClient();
var fileDB = Path.GetTempFileName();
try
{
wc.DownloadFile("http://www.mbnet.com.pl/dl.txt", fileDB);
LottoListResults dbl = null;
if (CreateDatabase(fileDB, out dbl))
{
var deep = 20;
var network = new BasicNetwork();
network.AddLayer(
new BasicLayer(null, true, 6 * deep));
network.AddLayer(
new BasicLayer(
new ActivationSigmoid(), true, 5 * 6 * deep));
network.AddLayer(
new BasicLayer(
new ActivationSigmoid(), true, 5 * 6 * deep));
network.AddLayer(
new BasicLayer(
new ActivationLinear(), true, 6));
network.Structure.FinalizeStructure();
var learningInput = new double[deep][];
for (int i = 0; i < deep; ++i)
{
learningInput[i] = new double[deep * 6];
for (int j = 0, k = 0; j < deep; ++j)
{
var idx = 2 * deep - i - j;
var data = dbl[idx];
learningInput[i][k++] = (double)data.V1;
learningInput[i][k++] = (double)data.V2;
learningInput[i][k++] = (double)data.V3;
learningInput[i][k++] = (double)data.V4;
learningInput[i][k++] = (double)data.V5;
learningInput[i][k++] = (double)data.V6;
}
}
var learningOutput = new double[deep][];
for (int i = 0; i < deep; ++i)
{
var idx = deep - 1 - i;
var data = dbl[idx];
learningOutput[i] = new double[6]
{
(double)data.V1,
(double)data.V2,
(double)data.V3,
(double)data.V4,
(double)data.V5,
(double)data.V6
};
}
var trainingSet = new BasicMLDataSet(
learningInput,
learningOutput
);
var train = new ResilientPropagation(
network, trainingSet);
train.NumThreads = Environment.ProcessorCount;
START:
network.Reset();
RETRY:
var step = 0;
do
{
train.Iteration();
Console.WriteLine("Train Error: {0}", train.Error);
++step;
}
while (train.Error > 0.001 && step < 20);
var passedCount = 0;
for (var i = 0; i < deep; ++i)
{
var should =
new LottoResult(learningOutput[i]);
var inputn = new BasicMLData(6 * deep);
Array.Copy(
learningInput[i],
inputn.Data,
inputn.Data.Length);
var comput =
new LottoResult(
((BasicMLData)network.
Compute(inputn)).Data);
var passed = should.ToString() == comput.ToString();
if (passed)
{
Console.ForegroundColor = ConsoleColor.Green;
++passedCount;
}
else
{
Console.ForegroundColor = ConsoleColor.Red;
}
Console.WriteLine("{0} {1} {2} {3}",
should.ToString().PadLeft(17, ' '),
passed ? "==" : "!=",
comput.ToString().PadRight(17, ' '),
passed ? "PASS" : "FAIL");
Console.ResetColor();
}
var input = new BasicMLData(6 * deep);
for (int i = 0, k = 0; i < deep; ++i)
{
var idx = deep - 1 - i;
var data = dbl[idx];
input.Data[k++] = (double)data.V1;
input.Data[k++] = (double)data.V2;
input.Data[k++] = (double)data.V3;
input.Data[k++] = (double)data.V4;
input.Data[k++] = (double)data.V5;
input.Data[k++] = (double)data.V6;
}
var perfect = dbl[0];
var predict = new LottoResult(
((BasicMLData)network.Compute(input)).Data);
Console.ForegroundColor = ConsoleColor.Yellow;
Console.WriteLine("Predict: {0}", predict);
Console.ResetColor();
if (predict.IsOut())
goto START;
if ((double)passedCount < (deep * (double)9 / (double)10) ||
!predict.IsValid())
goto RETRY;
Console.WriteLine("Press any key for close...");
Console.ReadKey(true);
}
}
catch (Exception exception)
{
Console.WriteLine(exception.ToString());
}
finally
{
File.Delete(fileDB);
}
}
static bool CreateDatabase(
string fileDB,
out LottoListResults dbl)
{
dbl = new LottoListResults();
using (var reader = File.OpenText(fileDB))
{
var line = string.Empty;
while ((line = reader.ReadLine()) != null)
{
var values = line.Split(' ')[2].Split(',');
var res = new LottoResult(
int.Parse(values[0]),
int.Parse(values[1]),
int.Parse(values[2]),
int.Parse(values[3]),
int.Parse(values[4]),
int.Parse(values[5])
);
dbl.Add(res);
}
}
dbl.Reverse();
return true;
}
}
}
p ;).
Thank you for you post, it looks interesting and just what i have been looking to play with.
I want try try break the lotto numbers into patterns and see what the neural network predicts. I have had some what success doing it manually be in only small wins but the idea is more the fun of it.
Where is the data set that you are using as learning set for the neural network ? Also the test set data ?
HI , I just download a csv from a lotto website and imported it into my application.
I have been playing with the different activation to see which one works best. I also want to try have it run through a combination of configs to see which out work the best. will let you know the results.
Great!
Do you think the output could be different? Do you think we could show all the 49 numbers with each one his “predictability”?
Something like…
#1: 0.0749239248
#2: 1.4320832473
#3: 0.4383729751
… and so on to #49.
So we could make more than one selection with the 7 best numbers (for example).
Thank you!
How can i download the program
you don’t
comp[ile
how would I change the database file for past results ?
Hi Thomas,
I have created a SQL database with past results and changed the code to get data from this SQL table rather than text file. It is easier to update this table with latest results, works beautifully.
Sergei
Create new project and just copy this code into your new .cs file as Program.cs.
Thanks for sharing code. my question is the lottery data in .txt file should be ascending
or descending from draw date standpoint.
another question is do we have a upgraded version of this code?
Thanks
The file named dl.txt should have following data format:
1. 27.01.1957 8,12,31,39,43,45
2. 03.02.1957 5,10,11,22,25,27
3. 10.02.1957 18,19,20,26,45,49
4. 17.02.1957 2,11,14,37,40,45
5. 24.02.1957 8,10,15,35,39,49
6. 03.03.1957 24,26,31,35,43,47
7. 10.03.1957 13,20,23,29,38,44
…
To use it: Prepare new Console Project in C# and copy the entire code to Program.cs and add from NuGet the ENCOG library. Compile and you have it. The code in the line 94 has the “deep” parameter. You may increase that but when you make it too big like 200 it will be much slower. Hope it answers your question. Thanks!
Yes, there is newer code from Piotr:
https://iblog.isowa.io/2018/05/04/eurojackpot-with-encog-ann-and-stats/
i made few changes to the code.
from 6 to 5 numbers with maximum of 55.
i don’t understand the meaning of training error value, the red match and the green match.
what about the prediction, someone can explain what kind of prediction is it? thanks all
john