SE
Week 20 · Divisional Round

Model predictions, tracked transparently.

Ensemble ML picks across NFL moneyline, spread, and totals markets. Every result published, regardless of outcome.

Last updated Jan 18, 2026 · 9:42 AM EST
Season Record
124–87
58.8% win rate · 211 picks
Moneyline Accuracy
76.9%
Week 8 performance
Return on Investment
+14.2%
Season to date
Model Status
Active
Stack ensemble · v3.2

This week's picks

Click any row to expand the full matchup breakdown
Matchup Model Pick Spread Odds Confidence Result
Kansas City Chiefs vs Houston Texans
Full market board
Matchup breakdown
Houston Texans
Kansas City Chiefs
Divisional Round · Jan 18, 2026 · 4:30 PM EST
Spread: KC −8.5 · Total: 43.5 · FanDuel
Scoring (Points/Game)
Offense PPG22.8#18
Defense PPG Allowed21.2#13
Total Yards / Game
Offense Yards328.4#19
Defense Yards Allowed312.8#14
Rushing Yards / Game
Offense Rush Yds108.2#24
Defense Rush Yds Allowed118.9#17
Passing Yards / Game
Offense Pass Yds220.2#14
Defense Pass Yds Allowed193.9#9
3rd Down Efficiency
Offense 3D Conv %38.1#19
Defense 3D Conv %37.8#11
Scoring (Points/Game)
Offense PPG27.5#6
Defense PPG Allowed18.4#3
Total Yards / Game
Offense Yards358.9#8
Defense Yards Allowed289.1#5
Rushing Yards / Game
Offense Rush Yds121.4#14
Defense Rush Yds Allowed102.3#6
Passing Yards / Game
Offense Pass Yds237.5#5
Defense Pass Yds Allowed186.8#4
3rd Down Efficiency
Offense 3D Conv %46.2#4
Defense 3D Conv %34.5#4
Regular Season Matchup History — HOU @ KC (Last 3)
DateVisitorHomeResult (Home)Total
2024-12-21HOU 19KC 27+8.046.0
2023-12-24HOU 16KC 24+8.040.0
2022-12-18HOU 24KC 30+6.054.0
Average total (last 3): 46.7 KC is 3–0 in last 3 meetings
Buffalo Bills vs Baltimore Ravens
Philadelphia Eagles vs Los Angeles Rams
Detroit Lions vs Washington Commanders
Methodology

How the models work

Every prediction is the output of a reproducible pipeline. No hot takes, no gut feel — just tracked, versioned models running over structured data.

01

Stack Ensemble

Three gradient-boosted classifiers (XGBoost, LightGBM, CatBoost) combined via a logistic-regression meta-learner. Each component trained on a held-out fold to reduce overfitting.

XGBoostLightGBMCatBoost
02

PostgreSQL Game Database

Historical play-by-play, team-week aggregates, and opponent-adjusted efficiency metrics. Schema designed for fast feature generation at inference time.

PostgreSQLPython
03

Automated Data Pipeline

OCR extraction from SportsLine paired with The Odds API for FanDuel lines. Selenium-driven refreshes, versioned CSVs, reproducible from raw source to prediction.

SeleniumOCROdds API
04

Line Movement & ROI Tracking

Opening-to-close line deltas logged per game. Unit-based ROI calculated weekly against the model's stated confidence, published without revision.

Unit trackingCLV
§
Transparency: Picks are logged before kickoff and never retroactively edited. Model version, feature set, and confidence are stamped with each prediction. Past performance does not guarantee future results — please wager responsibly.