equine-edge.horse — Horse Racing
UK horse racing predictor built on 245K Class 1–4 races (2015–2025). A Race Profitability Scorer classifies races into Tier 1 Focus / Value / Avoid based on class, prize money, course tier, and field size. An XGBoost classifier (ROC-AUC 0.69) stacked with LightGBM and ExtraTrees predicts win probability per horse, with fractional odds conversion and bookmaker comparison to flag value bets; daily predictions refreshed automatically via GitHub Actions.
pitch-oracle.com — Premier League
Premier League match predictor with a four-model soft-voting ensemble (XGBoost, Random Forest, Gradient Boosting, Logistic Regression) scoring 3.5% above single-model XGBoost, plus optional PyTorch neural network and LSTM time-series models (+4.9% vs baseline). A Statistics tab profiles referee disciplinary tendencies, manager win rates, and team form; upcoming fixtures are fetched from ESPN API with kickoff times in Eastern Time.
gridiron-oracle.com — NFL
Three specialized XGBoost models for spread (91.9% win rate at ≥54% confidence, 75.5% ROI), moneyline (59.5% win rate, 65.4% ROI), and over/under, plus an XGBoost + LightGBM ensemble for DraftKings Pick 6 player props (Passing, Rushing, Receiving yards, TDs). Trained on 50+ leak-free temporal features from NFLverse play-by-play (2020–2024) with F1-score threshold optimization; nightly GitHub Actions pipeline with Kelly Criterion bankroll management and in-app Elite/Strong bet notifications.
gridlocked.racing — Formula 1
F1 race-finish predictor with 86+ engineered features (practice deltas, qualifying gaps, teammate deltas, circuit type, altitude, weather). Supports XGBoost, LightGBM, CatBoost, Ensemble, and a Position-Group model routing predictions between podium, points, and backmarker sub-models — MAE improved from 1.94 to ≈1.48 in the March 2026 rewrite. Data from F1DB JSON and FastF1 (safety cars, flags); Monte Carlo + RFE + Boruta + SHAP feature selection; email notifications deliver per-race predictions.
bracket-oracle.com — March Madness
March Madness predictor trained on 15,961 complete games (2016–2025) with 11 features from CBBD, KenPom (NetRtg, ORtg, DRtg, AdjT, Luck, SOS), and BartTorvik (AdjOE, AdjDE) — improving spread MAE by 8.4% and moneyline accuracy by 2.7 points versus the 3-feature baseline. XGBoost, Random Forest, and Logistic Regression ensemble with Kelly Criterion sizing, underdog value detection, and automated team-name canonicalization across all sources (99.7% coverage across 364 teams).
oracle-on-ice.com — NHL
NHL analytics platform pulling real-time data from the NHL Stats REST API and ESPN odds. An Expected Goals model blends raw goals-per-game with NHL xGF/xGA data (50/50), with goalie adjustments via GSAA, HD SV%, and SV% weighting. Eleven pages cover today's matchups, advanced team metrics (xGF%, CF%, FF%), Value Finder, per-60 player props, shooting-% regression watch, goalie comparisons, injury impact scoring, line movement with sharp-action signals, and model backtesting.
fairway-oracle.com — PGA Golf
PGA tournament winner predictor using an XGBoost ensemble trained on OWGR rankings and strokes-gained features (2018–2025). Scrapes PGA Tour, ESPN, and OWGR data with no paid APIs required; weather from Open-Meteo, market odds from The Odds API. Supports LSTM/Transformer sequence models for player time-series, course embedding clustering, fractional-Kelly walk-forward backtesting, live tournament re-ranking, and a value-bet tab highlighting model edge above market-implied probability.
breakpoint-oracle.com — Tennis
Comprehensive tennis predictor using ELO ratings, surface-specific form, serve/return stats, and H2H counts built from TennisMyLife (2020–present) and tennis-data.co.uk odds with 81.5% name-match coverage. Live pre-match odds from Matchstat RapidAPI with a 500-call/month budget guard. Three app tabs — Today's Matches (model vs. market edge highlighted in green), Match Explorer, and ELO Rankings — rebuilt daily via GitHub Actions.
betting-cleanup.com — MLB Baseball
Seven-page MLB analytics platform (Today, Stats, Matchup Analysis, Models, Performance, Pick 6, Info) with XGBoost and LightGBM ensemble models for moneyline, run line, and totals. Integrates MLB Stats API, PyBaseball, Retrosheet, and FanGraphs park factors + Guts! weights; in-game context factors include umpire run environment, bullpen workload, platoon advantage, and park factor. Live odds from ESPN and The Odds API with 30-minute caching; Kelly Criterion bankroll calculator and nightly GitHub Actions pipeline.
betting-baseline.com — NBA
Streamlit NBA analytics platform with win probability, spread, totals, and DraftKings Pick 6 player-prop models built on XGBoost and LightGBM. Pulls from nba_api, sportsdataverse hoopR (PBP-derived team features), Basketball Reference advanced metrics, and The Odds API. Referee assignments from nba_official with ESPN fallback, injury monitoring with positional impact scoring, and automated GitHub Actions pipelines for nightly prediction caching, hoopR refresh, and referee data.
tailgate-edge.com — College Football
Full-stack college football analytics app with five seasons of CFBD data, XGBoost and Ridge models covering win probability, spread, and totals. Features weekly predictions, value bets, team explorer pages, and Kelly-criterion bankroll sizing for FBS matchups.
mls-oracle.com — MLS
MLS match predictor built around the structural quirks that generic soccer models miss — salary-cap parity, cross-country travel fatigue, artificial turf, and Designated Player availability. Uses an XGBoost/Random Forest/Gradient Boosting/Logistic Regression ensemble with American Soccer Analysis xG data, great-circle travel distance, and playoff-race flags to predict Home Win, Draw, and Away Win probabilities with risk scoring and betting tips.
bet-bundesliga.com — Bundesliga
Bundesliga match predictor built on the same architecture as the Betting Oracle soccer suite. Uses a soft-voting ensemble (XGBoost, Random Forest, Gradient Boosting, Logistic Regression) trained on multiple seasons of historical results, with rolling form averages, vig-removed implied probabilities, and Shannon entropy risk scoring. Pages cover upcoming predictions, fixtures & standings, team deep dives, bookmaker markets, and value best-bet alerts.
laliga-linea.com — La Liga
La Liga (Primera División) predictor using 10 seasons of football-data.co.uk results and a weighted soft-voting ensemble (XGBoost ×2, Random Forest ×1.5, Gradient Boosting ×1, Logistic Regression ×0.5). Features xG rankings from FBref, last-5 form, Copa del Rey congestion flags, H2H analyzer, bookmaker market odds, and a Best Bets tab that surfaces value plays where model edge exceeds 4% versus the market.
