Search

Following up on the (now locked) Neural Net Lottery Picker thread — the backtest problem
Was reading back through the old Neural Net Lottery Picker thread and it got locked before the most interesting part really got going. The MCMC/Markov discussion in there was the best thread I've found on here, so starting a fresh one to keep it alive. The point that stuck with me: someone noted they found no real advantage running Markov on draw history because the independence of each draw keeps failing to reject it's just a discrete uniform distribution. That matches what I keep hitting -
Jun 23, 2026, 1:00 pm - NumberCruncher2 - Lottery Discussion Forum

markov pick3Reply
ARIMA is an AutoRegressive Integrated Moving Average. This does not help with lottery draws. For an example, I have looked at using ARIMA in time series analysis on the pick 3... the predictions center around 4.5, you can't play 4.5. It stays the same regardless of the number of epochs used or the training/validation data split, and that was using LSTM (Long Shot Term Memory) from 1 to 3 layers deep. A Markov model is closer to the reality, as it only depends on the current state to choo
Nov 21, 2025, 9:11 am - hypersoniq - Lottery Discussion Forum

My take on math and ball lotteryReply
What I got from that video was a way to find the factor (in that case 120) relating the combinations to the permutations. The next video on the youtube feed was a really good 25 minute dive into the origins and apllications of Markov chains, from predicting neutron behavior in the first atomic bomb to the google search algorithm...
Dec 19, 2025, 9:46 am - hypersoniq - Mathematics Forum

markov pick3
To create learning curves for Pick 3 lottery series, the process involves measuring the performance of a predictive model as the amount of training data used in time series models increases. Steps to create learning curves for Pick 3 1. Data Preparation: o Gather the Pick 3 lottery history. o Divide it into progressively increasing training and test sets (example: use the first 100 draws to train, then 200, 300, etc.). o For each set, extract the features and the target (lottery r
Nov 21, 2025, 7:06 am - dr san - Lottery Discussion Forum

Following up on the (now locked) Neural Net Lottery Picker thread — the backtest problemReply
This is exactly the kind of rigor I was hoping someone here had actually done. The chi-square goodness-of-fit across sample sizes is the part most people skip- they eyeball a frequency chart and call something hot without ever testing it against a uniform null. The fact that it kept reading as discrete uniform from 10 draws up to full history is the whole answer, really. Your IQR refinement is clever - adding the count since last drawn to deal with stale hots is the kind of thing that sounds
Jun 27, 2026, 2:46 pm - NumberCruncher2 - Lottery Discussion Forum

Markov chain generatorReply
That does sound like an interesting Python project. Aside from input draws and output, there is the possibility of capturing the internal list of probabilities of state transitions in a nested dictionary data structure. Similar to how Markov chain generators work in natural language processing (like auto complete in a browser). Being able to display what it learned can help understand (and adjust) the inner variables for better results. Good luck!
Sep 26, 2025, 9:20 am - hypersoniq - Lottery Discussion Forum

Michigan: 8/1 - 8/31/2025Reply
Markov top 15 1 131 2 171 3 431 4 134 5 137 6 136 7 138 8 471 9 174 10 177 11 931 12 434 13 437 14 151 15 176 Thirty more, still above the median joint-probability line 178 631 731 436 438 971 474 477 934 937 461 467 468 498 497 178 361 367 368 398 397 492 496 498 998 997 994 993 991 989
Aug 2, 2025, 8:49 am - Duckhunter - Pick 3 Forum

Neural Net Lottery PickerReply
I am working with the concept of Markov Chains when it comes to the transition probabilities from one draw to the next using the historical collection of followers for each digit in each position. So far I have noticed no advantage to using such a system because the independent nature of each draw is failing to reject the hypothesis that the underlying history data is anything but a discrete uniform distribution of independent draws. It does not deter my attempts and realize what I need is a
Feb 12, 2026, 11:06 am - hypersoniq - Mathematics Forum

Looking Back - Looking ForwardReply
Any new plans or new strategies going into 2026? Moving the two python scripts (classifying hot/cold/neutral and first order Markov chain of followers) into a fully functional GUI for windows and android. Any lottery lessons learnt in 2025? System development is more fun than playing Biggest single win in 2025? $500 on pick 3, once on each system.
Dec 31, 2025, 11:59 pm - hypersoniq - Lottery Discussion Forum

Lottery Analysis Tools .Reply
It is not too much of a challenge to generate all sorts of statistics, the trick is interpreting them. Perhaps the next step is to let the computer figure out what stats are important. I think that a Markov Decision Process, based on rewards and constraints, like a form of reinforcement learning might be a path forward. Create an AI agent that will traverse draw history, feeding each draw result with every feature (high/low, hot/cold, odd/even, frequency, etc) and let IT decide which sta
Sep 7, 2025, 1:08 pm - hypersoniq - Lottery Discussion Forum