chad g petey told me it only gives numbers that "appear random"Prev Topic Next Topic

New Topic New Poll

BobP

Dump Water Florida
United States
Member #380
June 5, 2002
3,678 Posts
Offline

Jul 26, 2025, 4:12 am

What is the "training data" it's working with?

BobP

. . . or it could be a rat.
hypersoniq

Pennsylvania
United States
Member #1,340
April 6, 2003
2,940 Posts
Offline

Jul 26, 2025, 6:45 am

Quote: Originally posted by BobP on Jul 26, 2025
What is the "training data" it's working with?

BobP

. . . or it could be a rat.

Generally training data would be to take a data set and split it up, maybe the first 70% for training an AI model, and the last 30% for testing the model.

You can change parameters on the model and redo the test data until it becomes more accurate. The error function measures how far off the predictions are off from the test data, then the computer "learns" to reduce the error by applying techniques like gradient descent to find a minimum error level, and adjusts the model (the "weights" applied to each feature) to get better results.

Ignoring the code part, the order is

Linear Algebra (the model)

Statistics (the error function)

And finally

Calculus (the "learning")

Good Times 🙄

Have an EXCELlent day!
dr san

bgonÃ§alves
Brasil
Member #92,558
June 9, 2010
4,155 Posts
Offline

Jul 26, 2025, 8:35 am

hello,

Lottery Pattern Analysis System
Here's a powerful system that focuses on endings, positions, and unusual patterns:

1. Collecting and Organizing Historical Data
The first step is to access a complete database with the results of all previous draws for the lottery you are analyzing.

Source: Official lottery websites, public databases, or APIs (if available).

Structure: For each draw, record:

Drawing number

Drawing date

The numbers drawn (in ascending order for easier analysis).

2. Ending Analysis (Final Digit 0-9)
This analysis focuses on the last digit of each number drawn.

Frequency Count by Ending:

For each draw, count how many times each ending (0-9) appears among the numbers drawn.

Accumulate this count across all historical draws.

Example: If the numbers drawn are 02, 14, 25, 30, 41, 58:

Ending 2: 1 time

Ending 4: 1 time

Ending 5: 1 time

Ending 0: 1 time

Ending 1: 1 time

Ending 8: 1 time

Trend Identification: Observe which endings are more or less frequent in the overall history. Endings that are "late" (have not appeared in many draws) may be considered more likely to appear, although this is only statistical speculation.

Ending Combinations:

Analyze the combinations of endings that often appear together. For example, in 6-number draws, is it common to have a good distribution of endings? Are there draws where all the endings are odd or all the even?

This can help avoid cards with very unbalanced endings.

3. Position Analysis (Positional Places)
This analysis examines the numbers drawn based on their position in the ascending sequence.

Range of Values by Position:

For each draw, record the first number drawn, the second, the third, and so on (always in ascending order).

Calculate the mean, median, and standard deviation of the numbers that appear in each position.

Example: If the lottery draws 6 numbers:

What is the common range of the first number drawn? (Usually low, e.g., 01-10)

What is the common range of the sixth number drawn? (Usually high, e.g., 45-60)

Identifying Anomalies: If the first number of a future draw is very high or the last number is very low, it may be a less common draw.

Evens and Odds by Position:

Observe the frequency of even and odd numbers in each position.

Are there positions where even numbers are more common? Or odd?

4. Unusual Patterns and Delays
This is the most challenging and fascinating part, focused on identifying what hasn't happened or what is statistically rare.

"Late" Numbers (Missing Many Draws):

Monitor the number of draws since the last time each number was drawn.

Numbers that have been "late" for a long time are often targeted by players, but remember that each draw is an independent event.

Filter: You can create a filter to include (or avoid) a certain number of "late" numbers in your game.

Non-Occurring Sequences:

Examine the occurrence of number sequences (e.g., 01, 02, 03).

Long sequences are very rare in lotteries.

Filter: Avoid playing long or very obvious sequences.

Neighboring Numbers and Distance:

Calculate the distance between the numbers drawn in each draw.

Example: For 05, 12, 20, 21, 35, 40, the distances are 7, 8, 1, 14, 5.

Analyze patterns of average distances. It's rare to have all numbers very close together or all very far apart.

Filter: Avoid cards where all the numbers are very close together or very spread out (no neighbors).

Sum of Numbers:

Calculate the sum of all the numbers drawn in each draw.

Observe the common range of the sum. Most sums tend to fall in an intermediate range, neither too low nor too high.

Filter: Discard combinations whose sum is outside the common historical range.

Odd and Even Number Distribution:

Monitor the proportion of odd and even numbers in each draw (e.g., 3 even/3 odd, 4 even/2 odd).

The 3/3 or 4/2 distribution (or vice versa) is the most common.

Filter: Avoid combinations with 6 even or 6 odd numbers, as they are statistically less likely.

Decenary Distribution:

Analyze how many numbers fall in each ten (01-10, 11-20, etc.).

It's common to have numbers spread across several tens, rather than all concentrated in one or two.

Filter: Avoid combinations where all the numbers fall in one or two tens, unless it's a rare pattern you want to explore.

How to Partially Master Randomness (Strategies)
"Partially mastering randomness" in the lottery doesn't mean predicting outcomes, but rather making decisions based on probability and historical patterns to increase the likelihood of selecting less common or more "typical" combinations, which can lead to bigger prizes if you win, since fewer people would be playing the same combinations.
LottoBux

25

Ontario
Canada
Member #109,236
April 9, 2011
61,309 Posts
Offline

Jul 26, 2025, 10:10 am

Quote: Originally posted by sully16 on Jul 26, 2025
This is all above my pay grade, so I say, Put on Pink Floyd and Pour some Wine .

Yep, Same here Sully

Pink Floyd with a few brewskies after work, though.

Wine is only an option if I run out of beer.
sully16

25

Ringleader

Michigan
United States
Member #81,736
October 28, 2009
160,644 Posts
Online

Jul 26, 2025, 8:09 pm

Quote: Originally posted by LottoBux on Jul 26, 2025
Yep, Same here Sully

Pink Floyd with a few brewskies after work, though.

Wine is only an option if I run out of beer.

Beers, It has to be super sweaty hot and must have pretzels or salty snack.
grwurston

100

Same digits. Different games.

United States
Member #90,245
April 24, 2010
16,846 Posts
Offline

Jul 27, 2025, 12:04 am

Quote: Originally posted by LottoBux on Jul 26, 2025
Yep, Same here Sully

Pink Floyd with a few brewskies after work, though.

Wine is only an option if I run out of beer.

Running out of beer is NEVER an option.

List Pick 3 singles by low sums, high sums. Separate each sum row by consecutive pair and non consecutive pair combos. These 3 pair groups. 14 17 47, 25 28 58, 36 39 69 are in 63 combos. 54 combos do not have them. Which sums do the 3 pair groups appear and not appear? Look at root sum groups 147, 258, 369. Play accordingly.
grwurston

100

Same digits. Different games.

United States
Member #90,245
April 24, 2010
16,846 Posts
Offline

Jul 27, 2025, 12:07 am

Quote: Originally posted by sully16 on Jul 26, 2025
This is all above my pay grade, so I say, Put on Pink Floyd and Pour some Wine .

Pink Floyd is best listened to Live.

List Pick 3 singles by low sums, high sums. Separate each sum row by consecutive pair and non consecutive pair combos. These 3 pair groups. 14 17 47, 25 28 58, 36 39 69 are in 63 combos. 54 combos do not have them. Which sums do the 3 pair groups appear and not appear? Look at root sum groups 147, 258, 369. Play accordingly.
rcbbuckeye

100

Texas
United States
Member #55,887
October 23, 2007
22,275 Posts
Offline

Jul 27, 2025, 12:08 am

SHINER BOCK!!!!!

or YENGLING!!!!!

JUST LOOK AT THE ODDS OF ANY JACKPOT GAME, THAT WILL TELL YOU EVERYTHING YOU NEED TO KNOW
Mayte

Mexico
Member #193,539
November 3, 2018
369 Posts
Offline

Jul 27, 2025, 12:34 am

I think it's very naive to think that chatgpt is going to give you the winning lottery numbers, but maybe if you give it the pieces of the puzzle it can help you put all the pieces of the puzzle together.
Brock Lee

us
Member #231,828
May 4, 2023
1,638 Posts
Offline

Jul 27, 2025, 1:45 am

Quote: Originally posted by BobP on Jul 26, 2025
What is the "training data" it's working with?

BobP

. . . or it could be a rat.

this thread: https://www.lotterypost.com/thread/340828
rcbbuckeye

100

Texas
United States
Member #55,887
October 23, 2007
22,275 Posts
Offline

Jul 27, 2025, 9:36 am

Quote: Originally posted by Brock Lee on Jul 27, 2025
this thread: https://www.lotterypost.com/thread/340828

JUST LOOK AT THE ODDS OF ANY JACKPOT GAME, THAT WILL TELL YOU EVERYTHING YOU NEED TO KNOW
Stat$talker

700 light yrs West of Milky Way Galaxy's Center
United States
Member #200,634
September 1, 2019
5,510 Posts
Offline

Jul 27, 2025, 11:23 pm

Quote: Originally posted by rcbbuckeye on Jul 27, 2025

You're in a "Texez 2 Step" cauze,

Yo pseudo-science won't hep

"Cauze YOU don't have a Brain"..

Texez 2 Stepper?, Izz DAT the part You laughin at?

Naaah, drink a YENGLING to THAT

-Stat$talker
BobP

Dump Water Florida
United States
Member #380
June 5, 2002
3,678 Posts
Offline

Aug 2, 2025, 4:36 am

Quote: Originally posted by hypersoniq on Jul 26, 2025
Generally training data would be to take a data set and split it up, maybe the first 70% for training an AI model, and the last 30% for testing the model.

You can change parameters on the model and redo the test data until it becomes more accurate. The error function measures how far off the predictions are off from the test data, then the computer "learns" to reduce the error by applying techniques like gradient descent to find a minimum error level, and adjusts the model (the "weights" applied to each feature) to get better results.

Ignoring the code part, the order is

Linear Algebra (the model)

Statistics (the error function)

And finally

Calculus (the "learning")

Good Times 🙄

Ok that makes sense. If it used the actual draw history attempting to create combinations that feature attributes like combinations that win (much like we do with filters) and achieved a 60/40 hit rate within 10,000 combinations it would put jackpots within the range of Pick-4 Straight more often then not.

BobP

AI isn't what it used to be.