court-vision

# CourtVision -- a multi-sport prediction + decision-support AI with deep per-player intelligence [](https://github.com/neeljshah/court-vision/actions/workflows/proof.yml) *The badge re-runs every analytics module's own `--check` self-verification (`check_all.py`, 52 modules) on a bare clone -- committed-artifact verification only, no private data; it proves the receipts re-verify, not that any claim beats a market. Reproduce it yourself: [REPRODUCE.md](REPRODUCE.md).* **One calibrated AI brain that, across five sports, prices every market from a single coherent engine, drills all the way down to per-player prop distributions, reprices live as the game unfolds, and hunts for edge in the soft pockets where it can actually exist -- proving every claim leak-free and refusing to fabricate a dollar edge it has not earned.** It is *wide*: NBA, MLB, club soccer, the World Cup, and tennis, every one of them speaking the same `predict / to_jd / predict_live` interface, every market (moneyline, totals, spreads, 1X2, BTTS, correct-score, player props, alt-line ladders, same-game-parlays) priced off one calibrated anchor. It is *deep*: under each team number sits a per-player projected distribution, a ~190-feature prop stack, 44 player/team "atlases", playstyle archetypes, a coherent possession-level Monte-Carlo simulator, and a ~100-file edge-intelligence corpus that catalogs -- per sport -- every data source, every market, every beatable pocket, and every modeling lever with an honest ship/reject verdict. And it is *honest*: pregame, the model tracks the devigged closing line within sampling noise on team-strength markets and never beats it (the close's Brier is a few thousandths lower on every row -- see the measured table in the thesis section); in-game -- where conditioning on the realized game state makes the model's own Brier much lower -- the distributional tests read slightly model-favorable at some checkpoints but are UNDERPOWERED/PROVISIONAL, so the honest read is "possibly slightly ahead, not enough to call an edge." **No dollar edge, ROI, or "beat the close" is ever claimed** -- candidate edges are surfaced, tiered by evidence, and only ever proven by forward closing-line value (CLV), never asserted. Built by **[Neel Shah](https://github.com/neeljshah)** -- solo human architect and director of an agentic build pipeline. Engineering judgment, ship/reject decisions, and the validation methodology are mine. Open to **ML / data / quant / founding-engineer** roles -> [neeljshah22@gmail.com](mailto:neeljshah22@gmail.com) --- ### The measured win: in-game conditioning Conditioning on the realized game state is the one calibration result that clears every honesty rail: Brier drops sharply and leak-free, out-of-sample, on real corpora -- **NBA 0.209 -> 0.159, MLB 0.241 -> 0.126** (`edge_claimed: false`). The end-of-Q3 projection heads add a leak-clean **~26% MAE lift over a naive carry-forward baseline** (quoted alone, never paired with the separate endQ3 win-probability Brier, which carries its own Q4 leak). A live book sees the same realized state, so this is forecaster *quality*, not a dollar edge. Full proof and framing: [docs/INGAME_PROOF.md](docs/INGAME_PROOF.md).  --- ## Use it with YOUR Claude (2 minutes, any laptop) This repo ships a working **MCP server** — connect it to Claude Code or Claude Desktop and ask it anything: scouting reports, comparables, matchup previews, verified claim rankings, all fail-closed with receipts. One command downloads the latest descriptive-intelligence data pack (no betting data, no live feeds) and prints your exact Claude config. **Start here: [docs/USE_WITH_CLAUDE.md](docs/USE_WITH_CLAUDE.md)** · server details: [docs/MCP_SERVER.md](docs/MCP_SERVER.md) --- ## Explore the whole system, link by link This README is the front door. **[docs/INDEX.md](docs/INDEX.md) is the full map** -- every tracked document, organized by role and by funnel stage, so you can follow links from the raw data feeds all the way through the models, the simulator, the calibration gates, line-shopping and execution, the live in-game repricer, and the autonomous self-improvement loop. It is deep on purpose: a thorough read is a multi-day tour. - **New here?** -> [docs/INDEX.md](docs/INDEX.md) (the map) and the [GLOSSARY](docs/GLOSSARY.md) (CLV, leak-free, walk-forward, Shin devig, Brier, Kelly, ...) - **Want the honest numbers first?** -> [docs/JOB_EVIDENCE_PACKET.md](docs/JOB_EVIDENCE_PACKET.md) - **What's actually strong (in-game accuracy, calibration-only)?** -> [docs/IN_GAME_SHOWCASE.md](docs/IN_GAME_SHOWCASE.md) - **Want the guided demo path?** -> [docs/PRODUCT_DEMO.md](docs/PRODUCT_DEMO.md) (system health -> a live prediction -> an oracle query with a receipt -> the honesty ledgers) - **How AI built it?** -> [docs/BUILT_WITH_CLAUDE.md](docs/BUILT_WITH_CLAUDE.md) **Deep-dive layer (new)** -> [DAEMONS](docs/DAEMONS.md) - [PLATFORM_HARNESS](docs/PLATFORM_HARNESS.md) - [PAPER_TRADING_STACK](docs/PAPER_TRADING_STACK.md) - [INGEST_PIPELINES](docs/INGEST_PIPELINES.md) - [DATA_DEPTH](docs/DATA_DEPTH.md) - [ASK_SURFACES](docs/ASK_SURFACES.md) - [SPORTS_COVERAGE](docs/SPORTS_COVERAGE.md). The system's depth is now documented subsystem by subsystem, not just at the funnel-stage level above. --- ## How the whole brain works together Everything is one funnel, and every stage feeds the next. A model change in one place propagates coherently to every market through a single seam. ``` DATA keyless, leak-free ingest across 5 sports (ESPN, MLB StatsAPI, Sackmann, | football-data, prediction markets, DFS prop feeds) -> as-of-stamped parquet v SIGNALS leak-safe per-entity features + priors: team ratings (MOV-Elo / EW-Poisson / | surface-Elo / serve-hold), per-player per-exposure rates, 44 atlases, archetypes v MODELS one CALIBRATED win-probability per sport (the anchor) + per-player count | distributions (Poisson / Negative-Binomial, dispersion-calibrated) v ENGINES JointDistribution (coherent score matrix) + the possession Monte-Carlo sim | (emergent teammate correlation) + the live repricer (conditions on realized state) v PREDICTIONS ONE seam -> the full market surface: ML / totals / spreads / 1X2 / BTTS / | correct-score / player-prop ladders / SGP -- pregame and in-game, coherent v INTELLIGENCE the edge-intelligence corpus + the proving spine: every candidate edge tiered hypothesis -> calibration-proven -> CLV-proven; cut where markets are efficient ``` The cohesion is the point: **one win-probability anchors the moneyline, the spread, the total, and the in-game reprice** -- they are coherent reads off the same engine, not four independent models that can disagree. Pregame and in-game agree at tip-off by construction. Adding a sport is an adapter, not a kernel rewrite; adding a market is a read off the same anchor. --- ## What it predicts | Sport | Pregame market surface | Per-player props | In-game reprice | |------------|------------------------------------------------------|-------------------------------------------|----------------------------| | NBA | moneyline, total, spread/margin | pts/reb/ast/3pm/stl/blk + PRA + DD + SGP | end of Q1 / Q2 / Q3 | | MLB | moneyline, run-line, total | K / hits / TB / walks / outs / ER ... | after inning 3 / 5 / 7 | | Club soccer| 1X2, O/U-2.5, BTTS, correct-score, Asian handicap | shots / SOT / fouls / cards / saves ... | half-time | | World Cup | 1X2 (neutral-site), O/U-2.5, BTTS, correct-score | shots / SOT / fouls / fouls-drawn / saves | live minute-by-minute | | Tennis | match-win, games O/U, sets, holds | (aces buildable; feed-gated) | after set 1 | Every output is a calibrated probability or a point forecast with dispersion -- never a recommended wager. The buyer-facing CLI stamps `"edge_claimed": false` on every response. **NBA 2025-26 season coverage.** `games.parquet` carries all 1,156 scheduled games of the 2025-26 season; per-player boxscores now cover the FULL season -- 74 games from the original quarter-level cache plus 1,082 games backfilled from ESPN full-game summaries into the same `quarter_box` cache the pipeline already reads (q0 = full-game totals, q1-q4 = real quarters, zero downstream transform changes). The season's canonical shooter leaderboard claim is produced and independently VERIFIED on this corpus. **Ask it what kind of shooter someone is.** Not a ranking -- a ten-axis trait vector (volume/efficiency/difficulty/gravity/context), each axis citing its own VERIFIED claim, never combined into one score: ```bash python -m scripts.platformkit.intel_query.ask "what kind of shooter is Stephen Curry" # or directly: python -m scripts.platformkit.intel_query.compose_profile "Luka Doncic" ``` **Ask how the paper book is doing.** Streamed off the live paper ledger (never whole-file loaded), every answer paired with its channel's fail-closed greenlight verdict so a units figure never appears without its gate status: ```bash python -m scripts.platformkit.intel_query.paper_analytics "this week by channel" python -m scripts.platformkit.intel_query.paper_analytics "settlement backlog" ``` --- ## Wide in knowledge -- and it knows every little detail **Breadth (wide).** Five domains share one interface and one validation gate, so the brain reasons about a Premier-League match, a World-Cup neutral-site game, an MLB pitching matchup, an ATP set, and an NBA quarter with the same machinery. A keyless, free, idempotent, leak-free data backbone refreshes all five (~93 MB of as-of parquet) -- MLB StatsAPI boxscores, ESPN per-player rosters + athlete season splits, Sackmann tennis, football-data, plus prediction markets (Kalshi / Polymarket) and DFS prop feeds (Underdog / PrizePicks). The opportunity surface is bounded by data breadth, not by model breadth: the same pure best-line / Shin-devig / EV / arbitrage core covers *every* event the feeds return. **The claims scale.** Every derived stat the brain will answer a question from first has to pass an independent validator that marks it `VERIFIED`, `MISMATCH`, or `UNVERIFIABLE` -- **101,089 VERIFIED claim rows** stand behind the ask surfaces today, tallied live from every `*_validation.json` summary under `data/cache/intel_claims/` (71 claim stores, 7 sports). A machine-readable **data census** (`data/frontend/ops/data_census.json`) inventories what's derivable from the corpus on disk per sport -- 67 derivable families across NBA, MLB, soccer, soccer_intl, tennis, WNBA, NPB, KBO, and cross-sport markets -- and ranks every still-`UNBUILT` one by `leverage_rank` into a single cross-sport priority queue the autoloop reads to decide what to build next. The full per-sport breakdown (data on disk, claim counts, biggest gap, model vs close verdict) lives in [docs/SPORTS_COVERAGE.md](docs/SPORTS_COVERAGE.md). **Depth (every little detail).** Under each team number the brain goes all the way down: - **Per-player projected distributions.** Not a point estimate -- a full count distribution per player per stat (per-exposure rate x expected exposure -> Poisson/Negative-Binomial), so it prices the *entire* alt-line ladder ("over 0.5 / 1.5 / 2.5 ... shots", "10 vs 30 points") and the joint structure behind a same-game parlay, not just one line. - **A ~190-feature NBA prop stack** (rolling form, opponent defense, pace, rest/B2B, shot-zone tendencies, on/off, synergy play-type PPP, referee tendencies, schedule hardship, ...), with 23 feature blocks each annotated with the leak-free walk-forward result that killed or kept it. - **44 "atlases"** (28 player + 16 team: usage role, pace fit, matchup splits, vs-scheme splits, rest/B2B splits, spacing gravity, clutch shape, foul-drawing, ...) -- a deep descriptive + correlation asset. - **Playstyle archetypes, never people** -- striker / winger / holding-mid / keeper; power-hitter / contact-hitter / strikeout-pitcher; high-usage creator / 3-and-D wing / rim-runner; big-server / grinder / returner -- so the shrinkage prior knows what a *role* typically does. - **A coherent possession Monte-Carlo simulator** whose teammate correlation EMERGES correct from a shared scoring pie (measured rho ~ -0.10 vs realized) instead of a hand-tuned matrix -- one ~3-4s GPU run prices the whole prop / combo / SGP surface coherently. This is the one thing a marginal prop model structurally cannot do. --- ## Player props + the edge hunt (honest) The deepest, most actionable part of the brain is the **player-prop engine**, and it is where the search for *genuine* edge concentrates -- precisely because that is where edge can plausibly exist. **The thesis (load-bearing), with the measured numbers.** Sharp mainline markets (moneyline / spread / total on liquid sports) are efficient pregame. Leak-free, out-of-sample, vs the Shin-devigged close ([docs/MARKET_EFFICIENCY_PROOF.md](docs/MARKET_EFFICIENCY_PROOF.md)): | Market | n | Model Brier/RMSE | Close | Verdict | |---|---|---|---|---| | NBA moneyline | 372 | 0.1735 | 0.1672 | MATCH (within noise) | | MLB moneyline | 13,992 | 0.2429 | 0.2390 | MATCH (within noise) | | Soccer O/U 2.5 | 7,558 | 0.2465 | 0.2390 | MATCH (within noise) | | ATP match-win | 7,374 | 0.2177 | 0.2028 | BEHIND (freshness gap) | | NBA / MLB totals (RMSE) | 372 / 1,679 | 19.17 / 4.72 | 18.11 / 4.44 | BEHIND (freshness gap) | Nothing beats the close pregame; the close is a few thousandths sharper on every row. That is the expected honest outcome, not a failure. **In-game is different -- and honestly labelled.** Conditioning on the realized game state drops the model's own Brier sharply (NBA 0.209 -> 0.159, MLB 0.241 -> 0.126 vs the pregame-static baseline; calibration, not edge -- a live book sees the score too). Against the in-game *market*, the distributional (CRPS) checkpoint tests currently read slightly model-favorable at some late-game checkpoints but carry UNDERPOWERED / PROVISIONAL verdicts at today's sample sizes -- the honest read is "possibly slightly ahead in-game, not enough to claim an edge." Two things are explicitly **untested**: performance against instantly-updating live lines at real tick latency (all comparisons use captured price series, not live execution), and forward paper excess-vs-close, which at n=36 settled is NOT SIGNIFICANT (CI includes zero). The *beatable* pockets the brain hunts are different in kind: **lazily-priced soft / DFS player props**, **live / in-game lag**, **stale lines on slow books**, **prediction-market vs sportsbook divergence**, and **correlated SGPs a coherent sim can price but a book misprices**. The brain is built to surface and paper-trade exactly those -- and to *cut* effort where it has proven there is no edge. **How a prop edge is found, end to end:** scrape the soft DFS / book line -> build the player's full distribution from leak-free history (empirical-Bayes shrunk to the role archetype, blended with a club-season prior) -> price every rung of the ladder -> compare to the offered line -> rank by EV -- and then **label it by evidence tier and prove it before trusting it**: - **HYPOTHESIS** -- a plausible mispricing, not yet measured (most candidates start here). - **CALIBRATION-PROVEN** -- the model's P(over) is sharper than the line out-of-sample, leak-free (Brier-skill-score > 0). *Genuine, claimable.* - **CLV-PROVEN** -- forward closing-line value accrues on paper at meaningful sample size. *The bar for real money -- and the only thing that ever justifies a dollar claim.* A too-tight distribution would fabricate fake edges, so the brain calibrates dispersion (NB where counts are over-dispersed), demotes any stat it has *measured* to be weak, and ranks proven-stat edges above raw-EV blowups in unproven stats. The honest current state: in-game conditioning is calibration-proven 4/4 sports; goalkeeper Saves is calibration-proven; most prop candidates are HYPOTHESES awaiting CLV. **Paper-only; no dollar edge is claimed anywhere.** That discipline is the product -- an instrument that hunts edge *and* tells you, truthfully, which of its findings are real. --- ## The deep-intelligence layer Beyond the models sits a living **edge-intelligence corpus** (~100 markdown + structured files) that makes the breadth and depth navigable and actionable -- the brain's own map of where to push and where to stop: - **Per sport:** an edge-map (every market tagged beatable vs efficient, with evidence), a data-source ledger (have / missing / how-to-acquire), the full market + prop surface, an inefficiency catalog (each pocket with an in-data detection recipe + a proof method), a model-lever ledger (every lever with a SHIP / REJECT / PENDING verdict), and a prioritized path-to-edge. - **Cross-cutting:** edge theory + the cut-list of efficient markets to stop spending on, proof standards (the leak-free / OOS / CLV bar and the overfit traps that fake edges), per-source scrape specs, per-inefficiency detection recipes, a reusable method library (Poisson-vs-NegBinom, empirical-Bayes shrinkage, Shin devig, isotonic-when, Kelly + correlation sizing, CLV computation, walk-forward leak guards), and a single living edge-ledger of every candidate's evidence tier. Honest scope: the descriptive/atlas intelligence is a deep *scouting + correlation* asset and a predict-time-input the funnel is still wiring in (measured point-accuracy lift on the served model is ~0 today and is reported as such, not oversold). The corpus's value is making the entire search for edge systematic, grounded, and honest. See [docs/INTELLIGENCE.md](docs/INTELLIGENCE.md). --- ## The loop runs itself -- autonomy layer + the answer-engine oracle Beyond the discovery loop (mine residuals -> validate behind the ship gate -> ship or reject) sit the two autonomy stages that used to need a human: **self-shadowing** forward-settles the loop's own not-yet-confirmed verdicts against real outcomes as they land, and **self-proposal** generates and gates new hypotheses on a schedule with zero LLM calls and zero human trigger. A **sentinel layer** watches the watchers -- disk pressure, exception bursts, stalled heartbeats, tamper-evidence hashes on the invariant-enforcing code itself -- and a **one-command system-liveness harness** composes every gate/sentinel/ledger into a single readout that refuses to paper over a down section: a live run this week correctly reported `OVERALL: RED` with the specific failing subsystem named (a stale scraper heartbeat, 8 census-drift entries), not a decorative green. On top of the funnel sits an **answer-engine oracle**: a "what affects what" **effect graph** (555 nodes / 296 edges across NBA, MLB, soccer, and tennis) built entirely from rows the knowledge engine already adjudicated -- zero new statistics computed, only labeled and linked -- plus a **resolver registry** that maps every supported question type to exactly one deterministic source and REFUSES anything unregistered rather than improvising a plausible-sounding answer. The knowledge engine itself is now fully drained across all 4 sports: 197 mechanism hypotheses closed out (89 CONFIRMED_LOCAL, 74 honest NULLs, 34 not locally testable/other) -- every "does X actually happen" question the oracle fields carries a verdict, sample size, p-value, and source file, not folklore. Full account: [docs/PRODUCT_DEMO.md](docs/PRODUCT_DEMO.md). --- ## The thesis in numbers (calibration / sharpness -- NEVER a dollar edge) ### Pregame -- vs the devigged closing line, leak-free OOS Lower Brier / RMSE is sharper. MATCH = within sampling noise of the sharp close. BEHIND = the market's injury / lineup / weather / park / starting-pitcher freshness a public + box-score model cannot see. Source: `vault/_Edge_Maps/_Beat_The_Close.md`, reproduced below verbatim from a live `scripts.platformkit.beat_the_close_scoreboard` run against the real corpora; full framing + the candidate-REJECT table in [docs/MARKET_EFFICIENCY_PROOF.md](docs/MARKET_EFFICIENCY_PROOF.md). | Sport / market | Our model | Close | Verdict | |-------------------|---------------|---------|----------------------------------| | NBA moneyline | Brier 0.1735 | 0.1672 | MATCH (within noise) | | NBA total O/U | RMSE 19.17 | 18.11 | BEHIND (injury/lineup freshness) | | MLB moneyline | Brier 0.2429 | 0.2390 | MATCH (tiny pitcher-blindness) | | MLB total O/U | RMSE 4.72 | 4.44 | BEHIND (park/weather/SP) | | Soccer O/U-2.5 | Brier 0.2465 | 0.2390 | MATCH (pooled Platt) | | Tennis ATP ml | Brier 0.2177 | 0.2028 | BEHIND (ATP closes very tight) | ### In-game -- conditioning on the realized state beats the static pregame line. All 4 WIN. A live book also sees the state, so this is forecaster QUALITY, not a dollar edge. Source: `vault/_Edge_Maps/_Ingame_Scoreboard.md`. | Sport | Static -> Conditional Brier | When | |--------|-----------------------------------------|-------------------------------------| | NBA | 0.209 -> 0.159 | end Q1/Q2/Q3 (rating prior + score) | | MLB | 0.241 -> 0.126 | after inning 3/5/7 | | Soccer | 1X2 0.626 -> 0.502; O/U 0.264 -> 0.176 | half-time | | Tennis | 0.219 -> 0.151 | after set 1 (leak-free leader) | **Reading it.** Pregame MATCHES the devigged close on team-strength win markets and is BEHIND on totals / ATP only by freshness data the market sees and we cannot -- a data-bound gap, not a model defect. The sharpest forecaster FUSES the pregame intelligence prior with the realized state. We never claim a dollar edge, an ROI, or beating the close. --- ## Run it in minutes Slim install -- the predictor needs only a small scientific-Python surface (numpy, pandas, pyarrow, scipy, scikit-learn). The heavy CV / web / daemon dependencies are NOT required. ```bash pip install -r requirements-predictor.txt # or: pip install -e . -> cv-matchup / cv-predict / cv-live ``` ### One matchup -- pregame + in-game in a single JSON read ```bash python -m scripts.platformkit.predict_matchup --sport nba --home BOS --away LAL \ --elapsed 0 --home-score 0 --away-score 0 ``` ```json { "sport": "nba", "home": "BOS", "away": "LAL", "edge_claimed": false, "framing": "Pregame MATCHES the devigged close (calibration/sharpness, not an edge); in-game ADDS the realized state. No $ edge.", "pregame": { "p_home_win": 0.605, "total_mean": 211.3, "margin_home": 3.0 }, "ingame": { "p_home_win": 0.5732, "pregame_p_home": 0.605, "proj_total": 211.3, "proj_margin_home": 4.4 } } ``` Swap `--sport` for `mlb`, `soccer`, `soccer_intl`, or `tennis`. Reproduce the leak-free scoreboards on committed fixtures in under 60s on a fresh clone: ```bash python -m scripts.platformkit.beat_the_close_scoreboard --corpus tests/fixtures/proof python -m scripts.platformkit.ingame_scoreboard --corpus tests/fixtures/proof ``` --- ## Why trust it -- the rigor IS the product - **Leak-free by construction.** Expanding-window walk-forward with assertion-level per-fold leak guards, purge + embargo, and truncation-invariance tests (a feature at time T is byte-identical with or without future events). Cluster-robust Diebold-Mariano significance. The close is a comparison forecaster, never a model input. - **Honest nulls are successes.** "BEHIND by freshness", "MATCH within noise", "this lever is a measured NULL", "this recalibration OVERFITS out-of-sample -> deferred" -- all stated plainly. An efficient market proven efficient is a headline result, not buried. - **It disproves its own hype.** The same harnesses that grade the market were pointed inward and retired a market-follow ROI artifact, a Q4 look-ahead leak, and an L5-proxy ceiling mislabeled as edge. Building the instrument that refutes your own claims is the strongest signal here. - **A units number is never shown without its gate.** The paper-execution stack's **greenlight gate** evaluates every channel against seven pre-registered criteria (sample size in both independent halves, both-halves profitability, CLV significance, after-cost units, trust + eval-gate honesty, excess win rate) and writes a fail-closed RED/AMBER/GREEN verdict nightly -- criteria on trust and eval-gate honesty report RED, never a bare pass, on any missing or stale input. Fills are priced against real captured order-book depth (a VWAP walk of the opposite side's bid ladder, honest partial fills, `fill_quality: no_book` rather than a fabricated fill when the book is stale), and **STUCK detectors** turn a silent settlement stall into a visible alert instead of an unread log line (the incident that motivated it: 63+ silent zero-settle ticks). As of this writing every channel is RED or AMBER -- the gate reporting honestly that no edge has been proven yet is the feature, not a gap to paper over. Full account: [docs/PAPER_TRADING_STACK.md](docs/PAPER_TRADING_STACK.md). **Honesty truth-source:** every number's provenance + every retracted over-claim live in **[docs/JOB_EVIDENCE_PACKET.md](docs/JOB_EVIDENCE_PACKET.md)**. Open gaps: **[docs/KNOWN_LIMITATIONS.md](docs/KNOWN_LIMITATIONS.md)**. We never claim a dollar edge / ROI / beating the close / a CV predictive moat (CV SHAP ~ 0 in production today). **See the receipts, not just the claims.** Every calibration number is published two ways: - **[RECEIPTS.md](RECEIPTS.md)** -- the append-only, honesty-gated ledger. Each row is read verbatim from a machine-readable artifact on disk (path cited per row, nothing hand-typed), verdicts include the losses (`MARKET_SHARPER`, `UNDERPOWERED`) exactly as measured. The same ledger renders as a browsable page, live at **[neeljshah.github.io/court-vision/receipts.html](https://neeljshah.github.io/court-vision/receipts.html)** (static snapshot, as-of date shown on the page). - **Evidence pages** -- one per claim family, each running claim -> receipt -> reproduce: [retraction-story](docs/evidence/retraction-story.md) (the flagship: the instruments that refuted my own headline numbers) - [ingame-conditioning](docs/evidence/ingame-conditioning.md) (the one measured calibration win, market losses stated plainly) - [calibration-decomposition](docs/evidence/calibration-decomposition.md) (my own Brier gap vs the market split into reliability vs resolution, plus the ranked list of the game states my model handles worst -- charts: [reliability](docs/img/reliability_model_vs_market.png), [state-calibration heatmap](docs/img/state_calibration_heatmap.png)) - [devig-stack](docs/evidence/devig-stack.md) (four devig methods; the Shin-devigged close as the honest yardstick) - [leak-instruments](docs/evidence/leak-instruments.md) (walk-forward guard, truncation invariance, multi-corpus gate, ship gate) - [agent-fleet-direction](docs/evidence/agent-fleet-direction.md) (directing the agent fleet under fail-closed gates) - [cv-pipeline](docs/evidence/cv-pipeline.md) (broadcast video to court coordinates, with the not-demonstrated list stated). --- ## Origin / NBA computer-vision lineage (engineering history, not the headline) The platform grew out of **CourtVision**, an NBA broadcast-video CV pipeline (YOLOv8n detection -> SIFT homography -> Kalman + Hungarian tracking -> OSNet re-ID -> EasyOCR -> EventDetector) that turns a raw broadcast into court-coordinate data at ~$0.10-0.13 / full game on one consumer GPU. It is where the validation machinery came from -- but it is lineage, not the product: the CV-derived features carry ~0 measured predictive value today (SHAP ~ 0), and are not sold as an edge. Full audited account: [docs/JOB_EVIDENCE_PACKET.md](docs/JOB_EVIDENCE_PACKET.md) section 2; CV internals: [docs/CV_TRACKING.md](docs/CV_TRACKING.md). --- ## Buyer / reviewer docs | Document | What it covers | |----------|----------------| | [docs/INDEX.md](docs/INDEX.md) | **The full map** -- every document, by role and by funnel stage (start here to go deep) | | [docs/BUILT_WITH_CLAUDE.md](docs/BUILT_WITH_CLAUDE.md) | The agentic build pipeline -- Opus orchestrator + Sonnet executors under hard ship-gates | | [docs/GLOSSARY.md](docs/GLOSSARY.md) | Every term defined once (CLV, leak-free, walk-forward, Shin devig, Brier, Kelly, ...) | | [docs/INTELLIGENCE.md](docs/INTELLIGENCE.md) | The deep-intelligence corpus index -- the brain's map of every edge, source, and lever | | [docs/PREDICTOR_PLATFORM.md](docs/PREDICTOR_PLATFORM.md) | Full platform: thesis, scorecards, architecture, why it sells | | [docs/PROOFS.md](docs/PROOFS.md) | The provability index -- every claim -> the runnable leak-free proof | | [docs/JOB_EVIDENCE_PACKET.md](docs/JOB_EVIDENCE_PACKET.md) | The single honesty truth-source -- every number + the do-not-claim list | | [docs/PRODUCT_DEMO.md](docs/PRODUCT_DEMO.md) | The 15-minute demo path -- system health, a live prediction, an oracle query with a receipt, the composed board, the honesty ledgers | | [docs/KNOWN_LIMITATIONS.md](docs/KNOWN_LIMITATIONS.md) | Open gaps and what is not yet demonstrated | | [docs/PLATFORM.md](docs/PLATFORM.md) | Kernel + adapter multi-sport architecture direction | --- ## Tech stack **ML / data:** Python, NumPy, pandas, pyarrow, scipy, scikit-learn (Isotonic + Platt + NNLS), XGBoost, LightGBM. **Quant / validation:** walk-forward CV (season / era purged), Shin (1992) devig, per-stat isotonic / temperature recalibration, NegBinom dispersion calibration, multi-corpus calibration acceptance gate, cluster-robust Diebold-Mariano, truncation-invariance leak tests, CLV ledger. **CV lineage:** YOLOv8n, OpenCV, SIFT homography, OSNet re-ID, EasyOCR. **Live capture:** MLB GUMBO `feed/live` diffPatch poller at 10s cadence while any game is live (5s politeness floor); an order-book **fill simulator** VWAP-walks captured Kalshi depth ladders so a paper fill is priced against real liquidity, not a snapshot mid. **Serving:** FastAPI, uvicorn, SSE, parquet feature store, compute-once snapshot service. **AI agents:** Claude Code -- Opus orchestrator + parallel Sonnet/Opus executors under hard ship gates (this codebase, including the ~100-file intelligence corpus, was built by that pipeline under human direction). The runtime is **Claude-free** -- classical models + a deterministic self-improve loop, no LLM on the prediction path. **The build harness itself is live**: an autonomous probe-plan-spawn-gate-merge loop (`scripts/platform_harness/`) is 63.9% through its current backlog (53/83 tasks) with zero human required to click "continue." Full account: [docs/BUILT_WITH_CLAUDE.md](docs/BUILT_WITH_CLAUDE.md) - [docs/PLATFORM_HARNESS.md](docs/PLATFORM_HARNESS.md). --- ## Contact Solo-built (human-directed agentic pipeline). Available for senior ML / data / quant / founding-engineer roles. - **Start here:** [docs/JOB_EVIDENCE_PACKET.md](docs/JOB_EVIDENCE_PACKET.md) -- the honest, audited account - **Resume (PDF):** [docs/assets/NeelShahResume.pdf](docs/assets/NeelShahResume.pdf) - **Portfolio:** [neelshahportfolio.netlify.app](https://neelshahportfolio.netlify.app) - **Email:** [neeljshah22@gmail.com](mailto:neeljshah22@gmail.com) --- *All prediction numbers in this README are calibration / sharpness (Brier / RMSE / ECE), never a dollar edge. Candidate prop edges are tiered by evidence (hypothesis -> calibration-proven -> CLV-proven) and proven only by forward CLV, never asserted. The single honesty truth-source is [docs/JOB_EVIDENCE_PACKET.md](docs/JOB_EVIDENCE_PACKET.md); retracted / inflated numbers appear only there and in [docs/KNOWN_LIMITATIONS.md](docs/KNOWN_LIMITATIONS.md), in explicit retraction context, and never here.*