Projected scoreboard
A home-away score, spread, and combined total built from recent efficiency, not a betting line.
Prediction Engine Summary
Project game winners with a projected score, win chance, and a plain-English read of why the lean goes that way.
Win probabilityScore projection
Sample matchup
62% win prob
Duke
Home
North Carolina
Away
74-69
Score
+5.1
Margin
Net Diff
Top edge
Net Rating Diff+0.18
Home Court+0.11
Pace Matchup-0.06
Win probability
A 0-100% lean. 62% is a real edge. 52% is a coin flip the model barely prefers.
Matchup DNA
Last-10 net, eFG%, turnovers, and the glass: the pillars that decide most college games.
Why it leans
Blueprint ranks which gaps actually moved the lean: net rating, shooting, and the glass, in basketball language.
Start a matchup
Lock two teams. Then ask for a lean.
Home and away are not labels for show. Location, recent home/away records, and conference quality all count. Pick sides the way the game will actually be played.
- 1Choose home
The team with the building, the crowd, and the home-court edge.
- 2Choose away
A different school. The model compares each roster’s own rolling form.
- 3Run the model
Optional: bench players first so form is rebuilt from games they missed.
Sample lean
Duke
Home · 74
VS
North Carolina
Away · 69
Home 62%Win chanceAway 38%
Spread
Home −5.1
Total
143.0
Read
Slight lean
62% is not a lock. Treat anything under ~55% as a toss-up and read why it leans before you trust the favorite.
What goes into a lean
The basketball the engine weighs
Each side is built from that school’s own recent games (last 5 and last 10). Away production is not “what home allowed.” It is what the visitor has been doing on their own film.
Dean Oliver’s Four Factors
Effective FG%, turnover rate, offensive rebound %, and free-throw rate. The four ways a possession ends that still explain most college scoring.
Efficiency and net rating
Points per 100 possessions on both ends, plus net. A +8 net over the last 10 travels farther than a hot 3-point night.
Form windows, not season noise
Last 5 catches the current rotation. Last 10 is the sturdier DNA. Shooting, assists, steals, and blocks ride in those same windows.
Place, quality, and conference
Home flag, location win%, conference game, conference-tier gap, and strength of schedule so a mid-major road trip is not treated like a home Quad 1.
Lock. Project. Explain.
Same three beats every time: two real sides, a scoreboard lean, and a readable reason it went that way.
Lock
Home and away, the way the game will be played. Optionally leave players out so recent form matches the rotation you expect.
Project
Win chance, projected score, spread, and total, built from recent efficiency, shooting, and the glass.
Explain
The biggest edges that pushed the favorite (net rating, home court, pace), ranked so you can argue with the lean, not guess at it.
Read it like a scout, not a lock ticket
The why-it-leans board shows which gaps mattered for this pair. A big net-rating edge is the model saying the efficiency gap is doing real work. A tiny home-court bump is flavor, not the whole story.
Net rating gap
Primary driver
Home court
Location
Pace matchup
Pulls back
Missing a starter?
Exclude players before you run. The engine drops games those names appeared in and rebuilds last-5 and last-10 form from the remaining boxes, closer to the version of the team that will actually take the floor.