Strategy Lab

Planned

Manage and compare strategies side by side, tune ensemble weights, and view backtest performance versus a random baseline. The strategy plugins below are already implemented and used by Prediction Studio.

This page will include
  • Side-by-side comparison of up to four strategies
  • Hit distribution, average matches, variance and confidence intervals
  • Random-baseline percentile and stability across historical windows
  • Per-strategy enable/disable and ensemble weight sliders

Core engine (stats, strategies, ensemble scorer, walk-forward backtest) is already implemented in lib/ and unit-tested — these pages surface it in UI.

Frequency Analysis
Descriptive
frequency

Scores numbers by how often they have appeared, using shrinkage toward the expected uniform frequency so small windows are not over-trusted.

Descriptive only. Past frequency does not change the probability of the next independent draw.

Hot Numbers
Descriptive
frequency

Favours numbers whose recent frequency runs above their expected value.

Descriptive. "Hot" is not evidence a number will stay hot; draws are independent.

Cold Numbers
Descriptive
frequency

Favours numbers whose frequency runs below their expected value.

Descriptive. Being "cold" does not make a number more likely next draw.

Gap / Overdue
Descriptive
gap

Scores numbers by how far their current gap exceeds their historical average gap (gap percentile).

Being "overdue" creates no mathematical obligation to appear. Draws are independent.

Recency-Weighted
Descriptive
frequency

Applies exponential decay so recent draws weigh more heavily than older ones.

Descriptive weighting only; not predictive of independent draws.

Odd/Even Balance
Descriptive
balance

Soft-scores a line by how typical its odd/even split is versus the historical distribution. Unusual splits are down-weighted, not rejected.

A more typical split does not raise the probability of being drawn.

Low/High Balance
Descriptive
balance

Soft-scores the low/high split versus the historical distribution (boundary from the rule era).

Descriptive shaping only; does not change draw odds.

Sum Range
Descriptive
balance

Scores a line higher when its sum sits near the centre of the historical sum distribution, with configurable tail exposure.

Central sums are more common historically but every valid line is equally likely.

Spread
Descriptive
diversity

Rewards lines whose numbers are well spread across the range (max−min, internal gaps) rather than clustered.

Cosmetic diversity; no effect on draw probability.

Entropy
Descriptive
diversity

Scores how evenly the line covers the number-space buckets (higher entropy = more even coverage).

Even distribution is aesthetic; does not improve odds.

Anti-Crowd
Descriptive
diversity

Estimates how "human-popular" a line looks (birthday clustering, sequences, multiples of five, all ≤31). Higher score = LOWER estimated sharing risk.

Does NOT change the probability of being drawn. It only estimates how likely a jackpot would be shared if the line won.

Prisoner Cycle Heuristic
Experimental
experimental

Experimental — no proven lottery advantage. Builds a deterministic permutation of the number universe, decomposes it into cycles, and prefers lines whose numbers span multiple cycles (a diversity device inspired by the 100-prisoners problem).

The success probability of the prisoner puzzle does NOT transfer to independent lottery draws. This is a diversity heuristic only.