Nick Koutras
Member
First draft for the procedure Announcers
based on ideas by Dennis Bassboss and Beaker, and implemented by
Nick Koutras.
Purpose: To select and display n numbers that will constitute the
prediction for the next draw of a Lotto game based on the
below described algorithm.
Description
Accepted Inputs:
Start Draw
Duration
Backward draws to include as predictors
Quantity of predicted numbers.
Trials
Step 1. Train the neural network
The backward draws are analyzed and the top x numbers with frequency
of appearance >1 are used. The quantity of x varies according to
draws specified. Always the top numbers of the frequency are used.
Following a FIFO (first in first out) priority queue, we collect these
results for up to 80% of the draws specified by Duration.
Step 2. Formulate the equations
From the data collected we design the neural network to be used
as predictor.
Step 3. Test the NN
We test the predictive ability of the net vs the remaining 20% of the draws
using the above neural network.
Step 4. Display statistical results.
The collected data from step 3 are displayed on the screen complete
with the ability factor of the neural net, R² value.
Step 5. Display the n numbers that the network will select
as predictions for the next draw.
Step 5 will be repeated as many times as Trials indicate.
Please note that is backward test algorithm. That means that we can test
our predictions on previous draws based on the above algorithm.
First trial will be ready this week.
If you have any recommentations please let me know.
Nick
based on ideas by Dennis Bassboss and Beaker, and implemented by
Nick Koutras.
Purpose: To select and display n numbers that will constitute the
prediction for the next draw of a Lotto game based on the
below described algorithm.
Description
Accepted Inputs:
Start Draw
Duration
Backward draws to include as predictors
Quantity of predicted numbers.
Trials
Step 1. Train the neural network
The backward draws are analyzed and the top x numbers with frequency
of appearance >1 are used. The quantity of x varies according to
draws specified. Always the top numbers of the frequency are used.
Following a FIFO (first in first out) priority queue, we collect these
results for up to 80% of the draws specified by Duration.
Step 2. Formulate the equations
From the data collected we design the neural network to be used
as predictor.
Step 3. Test the NN
We test the predictive ability of the net vs the remaining 20% of the draws
using the above neural network.
Step 4. Display statistical results.
The collected data from step 3 are displayed on the screen complete
with the ability factor of the neural net, R² value.
Step 5. Display the n numbers that the network will select
as predictions for the next draw.
Step 5 will be repeated as many times as Trials indicate.
Please note that is backward test algorithm. That means that we can test
our predictions on previous draws based on the above algorithm.
First trial will be ready this week.
If you have any recommentations please let me know.
Nick
