HOU @ BOS
Houston Rockets · Boston Celtics · Tuesday, 12 January 2027 at 01:30
UpcomingUpdatedAs of 24 Sept
What Forecast expects · Even
50%BOSchance to win
BOS · 50.3%49.7% · HOU
- A toss-up for the model: BOS 50.3% · HOU 49.7%.
- In games: out of 10 like this one, BOS would win 5 and HOU 5.
Confidence
low
- The model has it close to 50-50: either result would be normal.
- No game forecast has been checked against a result yet this season: confidence cannot be high until there are some.
- The model does not know this game's injuries or starting five.
Why
- Home court is worth 7.2 probability points to BOS.
- HOU's projected roster is stronger: 4th in the league versus BOS's 9th. That moves 6.9 points toward HOU.
- Who carries BOS: Payton Pritchard · Jayson Tatum · Derrick White.
- BOS has 4,342 season minutes unassigned (22%); the model counts them at replacement level.
- Who carries HOU: Amen Thompson · Alperen Sengun · Kevin Durant.
- HOU has 3,353 season minutes unassigned (17%); the model counts them at replacement level.
What weighs most
Home court+7.2 toward BOS
Projected roster+6.9 toward HOU
probability points
What could flip it
- Without Payton Pritchard (BOS) the favorite would flip: the model would give HOU 55.1%.↔ HOU 55%
- Without Amen Thompson (HOU), with his minutes at replacement level, the model would give BOS 59.2%.BOS 59%
- What the model does not look at and this number does not capture: rest and travel, the day's injury report, the starting five and recent form.
Uncertainty
- The model gives a number, not a band: its reliability is measured by comparing forecasts with results, not by drawing a margin.
- Over 10,749 games in 2017-2025 it did better than home court alone (Brier error 0.227 versus 0.247; 0 would be perfect).
- Rosters are not settled: the forecast is recomputed daily with that day's roster.
- It does not know injuries, rest, travel or the starting five: any of them can move this game without the number changing.
What changed since the last snapshot
Loading…
After the game
- Tips off on 2027-01-12. Afterwards, expected versus actual will be compared and recorded here.
Is it calibrated?
10,749
historical games
2017-2025
0.227
error (Brier)
home court only 0.247
- No forecasts have been settled this season yet: this model's calibration starts being measured with the first games (2026-10-20). Until then, its track record is historical: 10,749 games from 2017-2025.
How it is measured and limits
Model
SportiveWorld's game model, frozen before the season: it reads each roster's projected strength and home-court advantage. It does not read the day's injury report.
Data is up to date.
Sample · No settled games yet to measure it against.
Injuries · The model does not read the day's injury report: the rosters it uses are the season projections.
- · Rosters are not settled yet: the number can move until the season starts.
As of 24 Sept · SportiveWorld model
Verified against the source · 14/14
- ✓ expected competition and game
- ✓ home and away
- ✓ probability intact
- ✓ favourite = higher probability
- ✓ sealed and signed snapshot
- ✓ registered model
- ✓ sealed before tip-off
- ✓ consistent official result
- ✓ core verified by the server
- ✓ verified components
- ✓ numbers add up
- ✓ consistent scenarios
- ✓ freshness without contradiction
- ✓ consistent track record
This is reading the data, not a bet: we work with no bookmakers and show no odds. And a game turns on an injury, a hot night or a bad call.