Skip to main content

A quantitative research terminal for lottery analytics

Not a crystal ball. A 16-agent analytical swarm over 54,603 real draws, a Bloomberg-style analysis terminal, and honest math that shows its work.

54,603 draws28 games11 countries1955→today16 agents

The agent swarm

Sixteen specialised agents each compute a weighted vote from the historical warehouse, in parallel. A coordinator merges their votes into a single consensus distribution, then a sharing-risk penalty de-weights combinations other players over-pick — so a win is less likely to be split. See it run →

Frequency

χ² · 30/100/1000

Multi-window frequency analysis with chi-square testing.

Gap

Poisson process

Poisson-process gap analysis on overdue numbers.

Markov

P(nₜ | nₜ₋₁)

First-order transition probabilities from history.

Bayesian

Beta-Binomial · CDM

Beta-Binomial posterior with a Dirichlet prior (Cover 1989).

Ensemble

geo-mean stack

Geometric-mean stack of short/long/exponential-decay estimators.

Pair Correlation

co-occurrence lift

Per-number max lift with any pair, above chance.

Hot/Cold

rolling momentum

Rolling-window momentum over recent draws.

Sharing Risk

archetype Monte-Carlo

Player-archetype Monte Carlo to avoid split jackpots.

Plus a Spiritual Guide agent and a Coordinator that applies the sharing-risk penalty (γ) — the full roster is 16 agents across 10 selectable strategies.

The /explore terminal

A Bloomberg-style research surface. Every analytic is real statistics on real data, with the methodology one click away — and an honest “degrees-of-freedom” watermark that warns you when you’re over-fitting by tweaking parameters.

Open the terminal
  • Residual frequency lollipop with Bonferroni correction
  • Gap / recurrence beeswarm + geometric CDF
  • Markov transition heatmaps
  • CDM-Bayesian posterior smoothing plots
  • Strategy forest plots + efficient frontier
  • Calendar heatmaps, sum & joint distributions
  • Backtesting + calibration scorecards
  • Compare mode — two games side by side

Spiritual Lab

An optional add-on layering 11 tradition systems — numerology, Western & Vedic astrology, BaZi, Tzolk’in, tarot, I Ching, runes, magic squares, sacred symbols, and lucky-day calendars — blended with the statistical pick for players who want a personal lens.

Real data warehouse

54,603 draws · 307,867 numbers · 28 games across 11 countries, from 1955 to today. The same warehouse feeds the swarm, the terminal charts, and the backtests — no synthetic fixtures in production. Predictions are a paid feature: paid plans include a daily allowance, with credit top-up packs when you need more.

The honest edge

Lottery draws are random. No system changes the odds of any single ticket, and PatternSight never claims otherwise. What it gives you is systematic, transparent number selection with every calculation shown, plus sharing-risk reduction so a win is less likely to be split. That’s a real, defensible edge — and we’d rather state it plainly than sell magic.

Build on PatternSight

Programmatic access to lottery analytics — the same engine as the app, over a REST API and an MCP server. It returns descriptive statistics about past draws and the swarm’s mathematical picks; it is not a prediction API, and no outcome is guaranteed.

REST API

JSON over HTTPS. Draw histories, frequency and gap statistics, and — on the Ultra tier — swarm picks and walk-forward backtests.

curl https://api.patternsight.app/v1/games/powerball/frequency \
  -H "Authorization: Bearer $PATTERNSIGHT_API_KEY"

MCP server

rolling out

The same surface as eight Model Context Protocol tools — so an AI assistant will be able to call PatternSight directly in a conversation once the server is live.

{
  "mcpServers": {
    "patternsight": {
      "url": "https://api.patternsight.app/mcp",
      "headers": { "Authorization": "Bearer $PATTERNSIGHT_API_KEY" }
    }
  }
}

See the swarm work

Watch 16 agents turn decades of draw data into one systematic pick — then explore the terminal yourself.