Updated: Monday, Dec 15 01:22:01 AM EST
Merrimack Warriors vs Vermont Catamounts (Final)
59 - 66, (Margin: -7, Total: 125)
Patrick Gymnasium
Second-Half conditional estimates based on a margin of -9.0 points and 51.0 total points (models are continuously adjusted up to Halftime)
------------------------------------------------------------------------------------------------------------------------------------------------------
E[AWAY] E[HOME] E[SP] 68% CI E[TOTAL] 68% CI 2H-A 2H-H 2H-SP 2H-TTL AWAY WIN% HOME WIN% (ODDS HOME)
======================================================================================================================================================
50.9 58.6 -7.7 (-17.8, +2.3) 109.5 (97.3, 121.8) 0.0 0.0 -7.7 109.5 26.5% 73.5% -277
------------------------------------------------------------------------------------------------------------------------------------------------------
Live conditional estimates per 5 Minutes (+/- 1 deviation) (models are continuously adjusted)
------------------------------------------------------------------------------------------------------------------------------------------------------
TYPE 5M 10M 15M 20M 25M 30M 35M 40M
======================================================================================================================================================
E[SP] - - - - - - - -7.5
P[W] - - - - - - - 100.0%
ODDS - - - - - - - -99999
------------------------------------------------------------------------------------------------------------------------------------------------------
Ornstein-Uhlenbeck Process with BIC Model Selection**
TYPE L95% E[X] U95%
E[F|1H] -28.5 -8.4 11.7
E[F|NOW] -7.0 -7.0 -7.0
Deep Learning: Quantile GRU v6.0 (Neural Network with Calibrated Uncertainty)
TYPE L95% L68% E[X] U68% U95%
E[F|1H] -11.3 -9.9 -9.3 -8.7 -7.7
E[F|NOW] -8.5 -7.4 -6.9 -6.5 -5.7
Enhanced Ornstein-Uhlenbeck Process (BIC-optimized AR order, continuous-time Kalman filtering)
MODEL NOTES:
** O-U Process validated on 61,166 games (MAE: 8.50 pts, well-calibrated 68%/95% confidence intervals)
** GRU v6.0 validated on Fall 2025 (MAE: 10.99 pts, 5.2% better than previous LSTM)
CHART LEGEND:
1) Blue line: O-U Improved model trained on first-half data, locked at halftime (68% and 95% CI shown)
2) Green line: O-U Improved model continuously retrained as new plays arrive (68% CI shown)
3) Both lines use the same statistical model; divergence shows how live events update predictions
4) Model automatically selects AR(1) or AR(2) using BIC for optimal fit to irregular scoring events
Updated: Monday, Dec 15 01:22:01 AM EST
Back