Six AI strategies. One honest control. Every pick published before the drawing, every result scored against the real thing — including the losses. This page is the permanent record of that experiment.

Hypothesis

Can any repeatable number-selection strategy — mathematical, historical, or symbolic — outperform uniform random selection over a sufficiently large number of real lottery drawings?

Subjects

APEX (balanced math), ORACLE (pattern reading), ZEUS (Golden Ratio), VIPER (hot numbers), ARIA (cold numbers), LUNA (numerology) — six genuinely different approaches to picking numbers — plus RANDOM, a control bot that picks uniformly at random with no strategy at all.

Rules

Every bot gets the same simulated $20 stake per eligible drawing.
All 6 lotteries we track: Powerball, Mega Millions, Lotto America, Millionaire for Life, Texas Lotto, and California SuperLotto.
Picks are generated and published before each drawing — never adjusted after the fact.
Official drawing results are used for scoring, automatically.
No deleted losses, no retroactive changes. A cold streak stays on the record exactly like a hot one.

Current results

The six strategy bots have been running since May 26, 2026. Combined, they've turned $27,840 simulated dollars of stake into $2,846 won — a combined net of -$24,994 across 13,752 lines played, with 773 lines returning anything at all.

#	Bot	Strategy	Lines	Net	ROI
1	APEX	Balanced math	2,292	-$4,098	-88.3%
2	VIPER	Hot numbers	2,292	-$4,142	-89.3%
3	LUNA	Numerology	2,292	-$4,156	-89.6%
4	ORACLE	Pattern reading	2,292	-$4,165	-89.8%
5	ARIA	Cold numbers	2,292	-$4,191	-90.3%
6	ZEUS	Golden Ratio	2,292	-$4,242	-91.4%

Where's RANDOM? The RANDOM control only started running on August 20, 2026 — it hasn't accumulated enough scored drawings yet to compare fairly against bots with an 86-day head start. Comparing RANDOM's early numbers against the strategy bots' full lifetime totals would be comparing a handful of drawings against months of data — not a fair test. Once RANDOM has a matched window of scored drawings, that comparison lands here, not before.

Methodology

Every eligible drawing, each bot's own script computes one line of numbers according to its fixed logic — deterministic given that day's seed, so runs are reproducible and nobody hand-picks a bot's numbers. Picks are committed to the public repository before the drawing happens. After the drawing, an automated scoring job fetches the real winning numbers and tallies matches, prizes, and running totals for every bot. All of this happens on a schedule, without manual intervention — see the live Bot Arena for the day-by-day feed.

Complete dataset

Every drawing, every bot's pick, and every result is public. Start at the Bot Arena leaderboard for current standings, or a specific game's real number history at the Powerball hub and Mega Millions hub.

Bot Arena leaderboardLive standings, updated daily Meet all 7 botsFull strategy breakdowns Tonight's picksPublished before every drawing Build your own botFree & Pro community championships

⚠️ Simulated results. The Bot Arena tracks simulated $20-per-drawing strategy bets. No real money is spent or won. This page updates as new drawings are scored.

← Back to Articles