About ProbWin

ProbWin is an AI-powered sports predictions service founded in 2024. Our proprietary machine learning technology analyzes millions of sports data points to generate predictions on NBA, NHL, Soccer, Tennis, and MLB. Our models cover Spreads, Totals, Moneyline, Over/Under 2.5, and 1X2 markets. Since launch, we maintain a verified track record: 59.6% win rate on 2,540 NHL picks, 61.2% on 477 NBA picks, and 64.1% on 2,314 MLB picks. All our results are public and timestamped on our results page.

My background in software and project management gave me the framework: build something robust, measurable, iterative. After nearly 5 years of tests and prototypes, then 2 years of intensive development, ProbWin was born — not from a "tipster", but from serious work combining data, modeling and automation.

Less talk, more proof. ProbWin is built around three principles:

Our ML models are strictly independent from bookmakers. No pick selection, no recommendation, no analysis depends on any operator — our predictions are produced from public data and our proprietary AI pipeline. Our main revenue comes from subscriptions: we only earn when our analyses are useful to you.

Unlike the majority of prediction services that derive up to 70% of their revenue from bookmaker affiliations — and often adapt their picks accordingly — our business model is aligned with your interest. On certain markets where access to card payment is limited (notably French-speaking Africa), we additionally provide links to a partner operator, transparently labelled as such. These links are never displayed to users located in France or in countries subject to the ANJ.

Behind ProbWin, there's a real processing pipeline:

• 1 Data collection & acquisition, cleaning, consolidation — the real invisible work.
• 2 Specialized AI models per market: one market = one dedicated model (NHL Totals ≠ NHL ML ≠ NBA Spreads…). This is not a GPT generating predictions on the fly — each model is an AI trained to detect patterns in historical data to estimate probabilities.
• 3 Continuous retraining: models adapt when conditions change (seasons, trends, contexts). They regularly learn from the latest results.
• 4 Value/edge filter: we only publish an official prediction when the detected advantage is significant. These thresholds are defined according to patterns discovered during our training runs.

On the MLB side, we're talking hundreds of thousands of games and tens of millions of data points, with advanced metrics (xERA, FIP, WHIP, ballparks…). In NHL, we integrate expected goals (xG), goalie stats and back-to-back context. In NBA and NFL, we analyze pace, offensive/defensive ratings, rest days, and much more.

ProbWin is available as Freemium, but the subscription funds what actually costs money:

• → Data (often paid) — historical data for training, but also daily: players, games, odds multiple times a day.
• → Servers and maintenance — the model must run, update, stay stable.
• → Ongoing development — improving quality, traceability, the experience. Other markets are in preparation.

The goal remains to be as accessible as possible, without selling dreams: we build a statistical edge over the long term, not a promise of guaranteed gains.