Historical frequency does not change the mathematical odds of a valid combination in a fair random lottery.

Enter five main numbers to compare their structure with the selected historical sample.

This tool compares historical characteristics only. It does not estimate a combination’s probability of winning.

Each historical test point uses only draws that occurred before the target draw. This helps compare model behavior without allowing the model to see the answer in advance.

Backtest matches describe historical model behavior only. They do not establish improved future lottery odds.

These are descriptive comparisons of the tested historical window. PatternPicks does not treat the highest historical rate as proof that a model will outperform in future random draws.

Compare the analytical models with deterministic pseudo-random valid combinations over the same historical test targets. This control helps show whether an apparent backtest result is meaningfully different from ordinary random matching.

The random baseline is a research control, not a betting or lottery strategy. Short historical samples can vary substantially by chance.

See the complete 0–5 main-number match distribution for every tested model and the random control. Showing the full distribution prevents isolated higher-match outcomes from hiding frequent misses.

Freeze a projected research set with its model version, verified data window and creation time. Once frozen, the original record is preserved on this device for later comparison with subsequent verified draws.

Ledger v1 is stored in this browser. Cloud/account persistence is planned before paid-subscriber launch.

Aggregate results from projections that were frozen before a later verified draw became available. Pending records are excluded from performance rates.

Forward observations are descriptive. Small samples can vary substantially by chance and do not establish future predictive advantage.

Forward rates are shown with 95% Wilson confidence intervals so small samples are not presented with false precision. Wider intervals mean greater uncertainty.

Confidence intervals quantify sampling uncertainty in observed rates; they do not convert historical results into future winning probabilities.

Compare genuinely different research hypotheses on the same historical targets: long-term frequency, trailing 90-day momentum, target-weekday history, structure balance, combined model and random control.

Each target uses only information dated before that target. Model leadership in a historical sample is descriptive and is not evidence of future winning odds.

A separate analytical engine for ordered-digit games. Select Pick 3 or Pick 4 above to analyze positional frequency, repeated-digit structures, sums, parity and a reproducible research projection.

Digit patterns describe verified historical draws. They do not change the mathematical probability of a future fair drawing.

Test digit models chronologically using only draws available before each target. Compare long-term positional frequency, trailing 90-day positional momentum and a deterministic random-digit control.

Straight and box hits are historical test outcomes, not future winning probabilities. Complete misses remain in the record.

Profile draw-session behavior and structural patterns including repeated digits, digit sums and odd/even composition. These modules use the currently selected verified sample.

Structure frequencies are descriptive historical statistics and do not change the odds of a future fair drawing.

Rank ordered-digit candidates using independent historical signals rather than a single frequency table. The model combines long-term position frequency, trailing 90-day position momentum, target-session history and structural fit.

Combined-model scores measure historical fit only. They are not probabilities of winning.

Compare several ordered-digit candidates generated from different historical signals. Scores are normalized historical-fit scores, not winning probabilities.

Loading comparison…

Sample Structure Index summarizes characteristics of the selected historical dataset. It is not a candidate-combination score, a probability of winning, or a prediction of a future draw.