Predictions

Model: low-confidence (early season)

Every forecast the app makes, in one place. Odds come from an Elo-driven Monte Carlo over the real remaining schedule (3,208 games of signal across 802 matches); projections use pace + regression toward tier and career baselines. Each card links to the full detail pages.

🏆Playoff Race

Per-team playoff odds, projected seed and title odds, deterministic clinch / elimination math, and the “what you need” odds-by-final-record curve. Pick a tier:

Team rankings →

⚔️Matchup Predictor

Team vs Team

Elo + roster-skill blend → per-game win probability and the full 4-game series distribution, with the most-likely result.

Compare teams

Player vs Player

Head-to-head on per-game production, rating (OVR), and projected finish.

Compare players

📈Stat Projections

Projected Leaders

Final-season totals (goals, assists, saves…) for every player, paced from current rate and regressed toward tier average, with confidence levels.

Projected leaders

Per-Player / Per-Team

Each player and team page carries its own season projection with an 80% range — open any profile to see current vs pace vs projected.

Browse players

Ratings & Trends

Tier Fit

Who looks over- or under-skilled for their tier — candidates to move up or down.

Tier fit

Rising Players

Multi-season skill trajectory — who's trending up.

Rising players

Rising Teams

Form momentum — early-season vs recent win%.

Rising teams

Player Rankings

Cross-tier OVR and Advanced OVR (boost/speed/demos) leaderboards.

Player rankings

🎯How good is the model?

Predictions are gated on a walk-forward backtest: the model only shows as “enabled” once it beats both a coin flip and a raw win% baseline on out-of-sample games (measured by Brier / log-loss calibration). Its value here is calibration, not point-picking — per-game outcomes in this league are near coin-flips.

Prediction report card Model accuracy & calibration How ratings work

Data source: live (rscna.com) - standings & schedule refresh hourly.