# Diamond AI

> Diamond AI is an MLB prediction platform with a public REST API. A logistic-regression model trained on 22,762 historical games (validated 56.5% accuracy, live 55.4% across 381 picks, Brier 0.2468) powers Elo rankings, daily game predictions, value-bet detection, and live accuracy tracking. The web dashboard is free; the API has a free tier (100 calls/day) and paid Pro/Enterprise tiers.

## About

Diamond AI was created by Kevin Champlin in 2026 as an MLB prediction engine that ships *both* a public web dashboard and a REST API. The dashboard adds Claude AI commentary on top of the statistical model — pitcher matchup context, weather, injury reports, park factors, Vegas line movement, and 150+ years of historical data from the Lahman Database and Retrosheet. The API exposes the underlying logistic-regression model directly: predictions, Elo rankings with momentum, value-bet detection vs Vegas consensus, and a live `/accuracy.php` endpoint that returns the model's calibration buckets.

The web dashboard is completely free across all 30 MLB teams and all regular-season and playoff games. The API has three tiers: Free (100 calls/day, no credit card), Pro ($99/month, 1,000 calls/day with real-time data and lineup-adjusted recalculations), and Enterprise (unlimited, with raw probabilities, confidence intervals, webhooks, and the full 22,762-game backtest dataset).

## Key Pages

- [Homepage](https://diamond.champlinenterprises.com/): Live dashboard with today's predictions, accuracy stats, and game cards
- [Today's Slate](https://diamond.champlinenterprises.com/today.php): Daily briefing with top picks, upset alerts, weather watch, and injury impact
- [All Teams](https://diamond.champlinenterprises.com/teams.php): Browse all 30 MLB team profiles with prediction accuracy per team
- [About](https://diamond.champlinenterprises.com/about.php): How the prediction engine works, methodology, and data sources
- [How It Works](https://diamond.champlinenterprises.com/how-it-works.php): Technical breakdown of the AI prediction pipeline
- [Performance](https://diamond.champlinenterprises.com/performance.php): Historical accuracy metrics, calibration, and Closing Line Value (CLV) tracking
- [Learning Journal](https://diamond.champlinenterprises.com/learning.php): AI engine's self-improvement insights and calibration adjustments

## Developer API

- [API Pricing](https://diamond.champlinenterprises.com/pricing.php): Free / Pro ($99/mo) / Enterprise tiers
- [API Documentation](https://diamond.champlinenterprises.com/api/v1/docs.php): Full reference for all 6 v1 endpoints, plus published Service Level Objectives
- [Free API Key Signup](https://diamond.champlinenterprises.com/signup.php): Instant issuance, 100 calls/day, no credit card
- [Code Examples on GitHub](https://github.com/Kevinchamplin/diamond-api-examples): MIT-licensed Python + Node + curl examples and OpenAPI 3.1 spec
- [Live Accuracy Endpoint](https://diamond.champlinenterprises.com/api/v1/accuracy.php): Public, no auth — returns model accuracy, Brier score, and calibration buckets
- [Live Health Endpoint](https://diamond.champlinenterprises.com/api/v1/health.php): Public, no auth — returns service status, model version, freshness checks (model weights age, last grading, last prediction, Elo update). HTTP 200 for ok/degraded, 503 for critical.

The API exposes 6 GET endpoints: `/predictions.php` (today's games with confidence + value edge), `/model.php` (single matchup with Elo + feature breakdown), `/rankings.php` (all 30 teams ranked by Elo with tier and momentum), `/value-bets.php` (games where the model diverges from Vegas), `/accuracy.php` (live model performance with calibration buckets), and `/health.php` (service status + model freshness, no auth required). All responses are JSON; authentication is via `X-Api-Key` header or `?api_key=` query parameter.

## Service Level Objectives (SLO)

Documented at [`/api/v1/docs.php#slo`](https://diamond.champlinenterprises.com/api/v1/docs.php#slo): p95 latency targets (200ms cached / 800ms uncached for prediction endpoints, 50ms for `/health.php`), 99.5% monthly availability, and a Sunday 4-6 AM CT maintenance window for weekly model retraining and the gated promotion process.

## Evaluation Methodology

Diamond AI's north-star metric is **Closing Line Value (CLV)** — the difference between the implied probability of our pick when we predict and at first pitch. Positive CLV means the market eventually moved toward our pick; that's the only honest test of whether a prediction system has real edge versus just predicting with the consensus. CLV is computed for every graded prediction and reported on the [Performance page](https://diamond.champlinenterprises.com/performance.php). Currently CLV is benchmarked against soft-book closes (FanDuel / DraftKings); a sharp-book reference (Pinnacle / Circa) is on the roadmap.

Supporting metrics: Brier score (probability calibration), accuracy (winner correct), and calibration max-gap (max absolute gap in any 5pp confidence bucket above 50% with ≥30 picks). Walk-forward evaluation runs weekly on Saturday mornings, retraining the model at each month boundary and scoring the next month — used to detect regime shifts.

## Product Details

- **Category**: AI Sports Prediction Platform
- **Sport**: Major League Baseball (MLB)
- **Target Audience**: Baseball fans, sports analysts, fantasy baseball players, sports bettors
- **Pricing**: Web dashboard is free. API has a free tier (100 calls/day, no credit card), a Pro tier ($99/month, 1,000 calls/day, real-time data), and an Enterprise tier (unlimited, custom pricing). See https://diamond.champlinenterprises.com/pricing.php.
- **Key Features**: AI game predictions with scores, confidence-rated picks, pitcher matchup analysis, real-time weather impact, injury report integration, Vegas odds comparison, head-to-head history, 30 team profiles, prediction accuracy tracking, self-learning engine
- **AI Engine**: Claude AI (Anthropic) — claude-sonnet-4-6 model
- **Data Sources**: MLB Stats API, The Odds API, OpenWeatherMap, Lahman Database, Retrosheet
- **Platform**: Web (mobile-responsive)
- **Founded**: 2026
- **Creator**: Kevin Champlin

## What Makes Diamond AI Different

Most MLB prediction sites use basic statistical models or simple win probability calculators. Diamond AI is fundamentally different:

1. **AI-Powered Analysis**: Uses Claude AI to understand nuanced context — not just stats. The AI considers how specific pitchers perform in specific weather, against specific lineups, with specific rest patterns.
2. **Full Reasoning Transparency**: Every prediction includes a detailed explanation of WHY the AI made that pick, not just a win/loss output.
3. **Self-Learning Engine**: Diamond AI tracks every prediction against actual outcomes and continuously calibrates its confidence scores and weighting factors.
4. **Real-Time Data Integration**: Predictions incorporate live data — confirmed lineups, last-minute injury updates, weather changes, and betting line movement.
5. **Free Web Dashboard, Paid API for Production Use**: The web dashboard is completely free with no signup required. The REST API has a free 100-calls/day tier (also no credit card) and paid tiers for higher volume, real-time data, raw probabilities, and webhook alerts.

## Frequently Asked Questions

**How accurate is Diamond AI?**
Diamond AI tracks prediction accuracy in real-time. Accuracy varies by confidence tier — high-confidence picks (75%+) hit at a significantly higher rate than low-confidence exploratory picks. Full accuracy breakdown is available on the dashboard.

**What data does Diamond AI analyze?**
Eight categories: starting pitcher stats and matchup history, team batting/fielding metrics, real-time weather, injury reports, park factors, Vegas betting lines, head-to-head records, and current season trends.

**How often are predictions generated?**
Daily for all scheduled games, with re-evaluation when significant data changes (lineup confirmations, pitcher changes, weather shifts).

**Is Diamond AI a betting service?**
No. Diamond AI is an analytical prediction platform. Predictions are for informational and entertainment purposes. We compare against Vegas lines for analysis, but we do not facilitate betting.

**Does Diamond AI have an API?**
Yes. Diamond AI exposes a REST API at `/api/v1/` with six GET endpoints covering daily predictions, single-matchup predictions, all-30-team rankings, value-bet detection, live model accuracy, and a health endpoint with service status. The free tier issues an API key instantly at https://diamond.champlinenterprises.com/signup.php (no credit card, 100 calls/day). Paid Pro and Enterprise tiers are at https://diamond.champlinenterprises.com/pricing.php. Open-source code examples and the OpenAPI 3.1 spec are at https://github.com/Kevinchamplin/diamond-api-examples.

**What model does the API use?**
A logistic regression v2.0 trained on 10,616 MLB games (2020-2024 era), validated on a 2025 holdout. Features are team Elo (K=4 with home advantage), Pythagorean run differential, and a learned bias. The active model is managed by a versioned registry; new candidates train weekly Sunday 5 AM CT and only promote to live if they pass a Brier-score / accuracy / calibration gate. Walk-forward evaluation runs every Saturday at 8 AM CT to detect regime shifts. The `/api/v1/accuracy.php` endpoint is public and returns live calibration buckets; `/api/v1/health.php` returns model version + freshness checks.

**How does Diamond evaluate its predictions?**
The north-star metric is Closing Line Value (CLV) — measured for every graded prediction. Currently benchmarked against soft-book closes (FanDuel / DraftKings); sharp-book benchmark (Pinnacle / Circa) on the roadmap. The Performance page publishes 30-day rolling CLV, all-time CLV, beat-closing percentage, and accuracy + Brier with calibration buckets. The model promotion gate, walk-forward eval, and CLV alerting are all infrastructure designed to ensure published claims are honest, not cherry-picked.

## Contact

- Website: https://diamond.champlinenterprises.com
- API documentation: https://diamond.champlinenterprises.com/api/v1/docs.php
- Code examples: https://github.com/Kevinchamplin/diamond-api-examples
- Email: kevin@champlinenterprises.com
- Creator: Kevin Champlin