Gameweek 10 · · Stadium of Light
Medium: the top outcome is at least 45%, or 15+ points clear of the next.
Chelsea's higher strength rating (Elo 1646 vs 1490) drives most of this prediction towards a Chelsea win.
Most likely scorelines
- 11-112.2%
- 21-29.4%
- 30-18.6%
- 40-27.6%
- 52-17.4%
| SUN ↓ CHE → | 0 | 1 | 2 | 3 | 4 | 5 |
|---|---|---|---|---|---|---|
| 0 | 7 | 9 | 8 | 4 | 2 | |
| 1 | 7 | 12 | 9 | 5 | 2 | |
| 2 | 5 | 7 | 5 | 3 | 1 | |
| 3 | 2 | 3 | 2 | 1 | ||
| 4 | ||||||
| 5 |
From the Dixon-Coles goals model, rescaled so the home / draw / away totals match the final prediction. Expected goals: SUN 1.23, CHE 1.58.
Model vs bookmakers
| Outcome | Model | Market | Gap |
|---|---|---|---|
| SUN win | 29% | 29% | +1 pts |
| Draw | 26% | 27% | -1 pts |
| CHE win | 45% | 44% | +0 pts |
DraftKings (via ESPN): decimal odds 3.25 / 3.50 / 2.10, bookmaker margin 7.0% removed proportionally. ◆ marks gaps of 10+ points. Bookmakers have historically been more accurate than this model (see Performance), so a big gap is more often the model missing something (injuries, team news) than a mispriced market.
Recent form
League position
| SUN | CHE | |
|---|---|---|
| Position | 20th | 4th |
| Points | 5 | 16 |
| Played | 9 | 9 |
| W-D-L | 1-2-6 | 5-1-3 |
| Goal difference | -10 | +4 |
| Points per game | 0.56 | 1.78 |
SUN at home · CHE away
| This season W-D-L | GF–GA | PPG | Last season | |
|---|---|---|---|---|
| SUN at home | 1-1-2 | 1–4 | 1.00 | 9-6-4 · 1.74 ppg |
| CHE away | 2-0-2 | 6–7 | 1.50 | 7-5-7 · 1.37 ppg |
Underlying numbers (this season, per game)
| SUN | CHE | |
|---|---|---|
| xG for | 1.42 | 1.23 |
| xG against | 1.88 | 1.42 |
| Goals for | 0.78 | 2.11 |
| Goals against | 1.89 | 1.67 |
| Shots for | 12.0 | 13.4 |
| Shots against | 14.1 | 12.8 |
| On target for | 4.0 | 5.4 |
| On target against | 3.8 | 5.6 |
| Clean sheets (total) | 2 | 2 |
Premier League matches only. xG from football-data.co.uk (8 and 8 matches with xG).
Rest since last match (all competitions)
Includes domestic cups and European games from ESPN's schedule data.
Head to head (last 10)
- WMay 24, 2026Sunderland 2–1 Chelsea
- WOct 25, 2025Chelsea 1–2 Sunderland
- LMay 21, 2017Chelsea 5–1 Sunderland
- LDec 14, 2016Sunderland 0–1 Chelsea
- WMay 7, 2016Sunderland 3–2 Chelsea
- LDec 19, 2015Chelsea 3–1 Sunderland
- LMay 24, 2015Chelsea 3–1 Sunderland
- DNov 29, 2014Sunderland 0–0 Chelsea
- WApr 19, 2014Chelsea 1–2 Sunderland
- LDec 4, 2013Sunderland 3–4 Chelsea
W/D/L from Sunderland's point of view. League meetings only.
What moves this prediction
- Strength ratings (Elo + goals model)-29 pts
- League table-1 pts
- Home / away form+1 pts
- Recent form+0 pts
- Rest & fixture congestion+0 pts
Each bar shows how the gap between home-win and away-win probability changes when the two teams are made equal on that group of inputs. It explains the model, not football causality.
Model components
- Dixon-Coles goals modelweight 50%29%27%44%
- Gradient boostingweight 50%30%24%46%
- Logistic regressionweight 0%27%30%43%
- Eloweight 0%23%26%51%
The final prediction is the weighted average shown. Weights were chosen on the 2016-17 to 2020-21 validation seasons. Model version v20261029-e9c638.