Home win
65.50%
Arsenal
ENG1 · Premier League · Match Analysis
Arsenal vs
Leeds United
Model-based match probabilities derived from multi-season xG data, team strength, venue performance, recent form, Elo ratings and league context.
Model output
Match probabilities.
Probabilities are model estimates, not certainties or betting guarantees.
Draw
21.99%
X
Away win
12.51%
Leeds United
Score model
Most likely scorelines.
Exact-score probabilities from the Poisson + Dixon-Coles goal model.
Additional scorelines
3:16.85%
0:06.53%
4:04.33%
2:24.06%
1:23.87%
Team context
Strength and recent form.
Venue-adjusted team strength, Elo rating and recent xG performance.
Home
Arsenal
Elo1655.4
Overall attack1.73
Overall defense0.81
Home attack1.83
Home defense0.67
Form signal+0.789
Last 5 xGF / xGA1.71 / 0.95
Last 10 xGF / xGA1.74 / 0.89
Last 10 xGD+0.85
League statusSTABLE · from ENG1 2025/26
Away
Leeds United
Elo1515.1
Overall attack1.45
Overall defense1.46
Away attack1.37
Away defense1.69
Form signal+0.222
Last 5 xGF / xGA1.61 / 1.28
Last 10 xGF / xGA1.49 / 1.48
Last 10 xGD+0.02
League statusSTABLE · from ENG1 2025/26
Elo difference
+140.3
Elo is one input among several and does not determine the final 1X2 probability on its own.
Head-to-head
Historical matchup context.
H2H values provide context and are not used as a standalone prediction.
Goal distribution
Total goals profile.
The model's probability distribution for the total number of goals.
06.53%
113.20%
224.11%
322.39%
416.42%
59.63%
64.71%
7+3.02%
Technical detail
How the model reached the estimate.
Detailed diagnostics are kept below the headline probabilities so the public report remains readable.
League baseline and model setup
LeagueENG1
Loaded teams27
Loaded historical matches1570
Target season2026/27
Previous seasons in history4
Season window2022/23 → 2023/24 → 2024/25 → 2025/26 → 2026/27
League tierTier 1
Historical league matches1570
xG matches1570 / 1570
Effective xG matches791.97
Observed home xG1.60
Observed away xG1.32
Model home xG baseline1.60
Model away xG baseline1.32
Total league xG baseline2.91
Home-advantage weight1.000
Model configdata/models/ENG1.json
Team attack half-life240.0 days
Team defense half-life240.0 days
Attack strength weight1.000
Defense strength weight1.000
Venue prior matches0.00
DistributionPoisson + Dixon-Coles
Dixon-Coles rho-0.130 (effective -0.130)
Dixon-Coles decay0.000
Goal overdispersion0.000
Implied goal variance H / A2.10 / 0.83
Relative strengths and form diagnostics
Arsenal
Home attack strength1.108
Home defense weakness0.633
Base lambda home1.84
Attack form xGF level1.720
Defense form xGA level (lower is better)0.930
Opponent-adjusted attack residual+0.249
Opponent-adjusted defense residual (higher is better)+0.511
Opponent-adjustment weight0.000
Model attack signal+0.263
Model defense signal+0.526
Opponent blend0.000
Form matches 5 / 105 / 10
Last 5 xGF / xGA / xGD1.71 / 0.95 / +0.76
Last 10 xGF / xGA / xGD1.74 / 0.89 / +0.85
Leeds United
Away attack strength1.026
Away defense weakness1.041
Base lambda away0.86
Attack form xGF level1.570
Defense form xGA level (lower is better)1.348
Opponent-adjusted attack residual+0.067
Opponent-adjusted defense residual (higher is better)+0.064
Opponent-adjustment weight0.000
Model attack signal+0.113
Model defense signal+0.108
Opponent blend0.000
Form matches 5 / 105 / 10
Last 5 xGF / xGA / xGD1.61 / 1.28 / +0.33
Last 10 xGF / xGA / xGD1.49 / 1.48 / +0.02
Historical data coverage
Arsenal
Total matches157
Home / away matches78 / 79
xG matches157 / 157
xG home / away coverage78 / 78 · 79 / 79
Effective total30.94
Effective attack / defense30.94 / 30.94
Effective home matches15.04
Effective home attack / defense15.04 / 15.04
Leeds United
Total matches81
Home / away matches41 / 40
xG matches81 / 81
xG home / away coverage41 / 41 · 40 / 40
Effective total22.73
Effective attack / defense22.73 / 22.73
Effective away matches10.92
Effective away attack / defense10.92 / 10.92
Historical data coverage
64.15%
This measures usable historical evidence. It is not the probability that the prediction is correct.
Validation
Historical model performance.
Leakage-safe backtesting compares the fitted model with a league baseline over 1550 historical matches.
Positive improvement means the model performed better than the league baseline.
Total goals log-loss vs baseline+0.34%
1X2 log-loss vs baseline+8.36%
BTTS Brier vs baseline+0.33%
Over 2.5 Brier vs baseline+1.37%
Full backtest metrics
Model
MAE total goals1.3257
Exact score log-loss2.9853
Total goals Brier0.8276
Total goals log-loss1.9024
1X2 Brier0.5853
1X2 log-loss0.9827
BTTS Brier0.2452
Over 1.5 Brier0.1577
Over 2.5 Brier0.2410
Over 3.5 Brier0.2225
League baseline
MAE total goals1.3179
Exact score log-loss3.1008
Total goals Brier0.8295
Total goals log-loss1.9089
1X2 Brier0.6487
1X2 log-loss1.0724
BTTS Brier0.2461
Over 1.5 Brier0.1580
Over 2.5 Brier0.2443
Over 3.5 Brier0.2250
Calibration diagnostics
Home win · ECE 2.47%
0- 10% n=14 avg= 8.3% actual= 7.1% 10- 20% n=95 avg= 16.3% actual= 11.6% 20- 30% n=247 avg= 25.8% actual= 22.7% 30- 40% n=346 avg= 35.2% actual= 39.0% 40- 50% n=327 avg= 44.8% actual= 45.0% 50- 60% n=258 avg= 55.1% actual= 56.2% 60- 70% n=165 avg= 64.5% actual= 68.5% 70- 80% n=83 avg= 74.1% actual= 75.9% 80- 90% n=15 avg= 84.1% actual= 93.3%
Draw · ECE 1.36%
0- 10% n=3 avg= 9.4% actual= 0.0% 10- 20% n=139 avg= 17.0% actual= 14.4% 20- 30% n=1267 avg= 25.9% actual= 25.6% 30- 40% n=141 avg= 31.1% actual= 22.0%
Away win · ECE 3.47%
0- 10% n=82 avg= 7.6% actual= 7.3% 10- 20% n=310 avg= 15.6% actual= 15.5% 20- 30% n=394 avg= 25.0% actual= 21.1% 30- 40% n=351 avg= 34.5% actual= 38.7% 40- 50% n=236 avg= 44.3% actual= 41.5% 50- 60% n=119 avg= 54.3% actual= 63.9% 60- 70% n=43 avg= 64.5% actual= 74.4% 70- 80% n=12 avg= 73.2% actual= 75.0% 80- 90% n=3 avg= 81.8% actual= 66.7%
BTTS · ECE 2.61%
30- 40% n=3 avg= 35.9% actual= 33.3% 40- 50% n=202 avg= 47.4% actual= 49.0% 50- 60% n=862 avg= 55.3% actual= 57.0% 60- 70% n=463 avg= 63.5% actual= 59.0% 70- 80% n=20 avg= 71.1% actual= 60.0%
Over 2.5 · ECE 2.48%
20- 30% n=1 avg= 26.1% actual= 0.0% 30- 40% n=32 avg= 37.7% actual= 43.8% 40- 50% n=346 avg= 46.1% actual= 49.4% 50- 60% n=661 avg= 55.1% actual= 57.0% 60- 70% n=434 avg= 64.3% actual= 62.7% 70- 80% n=75 avg= 72.9% actual= 66.7% 80- 90% n=1 avg= 80.1% actual=100.0%
Walk-forward - latest periods
| Period | N | Model 1X2 LL | Baseline 1X2 LL | Improvement |
|---|---|---|---|---|
| 2025-08 | 30 | 0.9775 | 1.0463 | +6.58% |
| 2025-09 | 30 | 0.9920 | 1.0984 | +9.68% |
| 2025-10 | 30 | 0.9667 | 0.9993 | +3.26% |
| 2025-11 | 40 | 0.9613 | 1.0369 | +7.29% |
| 2025-12 | 56 | 0.9987 | 1.0989 | +9.11% |
| 2026-01 | 49 | 1.1387 | 1.1126 | -2.34% |
| 2026-02 | 42 | 1.0318 | 1.1382 | +9.34% |
| 2026-03 | 32 | 1.1246 | 1.0900 | -3.17% |
| 2026-04 | 30 | 1.0012 | 1.0588 | +5.44% |
| 2026-05 | 41 | 1.0345 | 1.0728 | +3.57% |
| 2026-08 | 20 | 0.9484 | 1.0473 | +9.44% |
| 2026-09 | 30 | 1.0955 | 1.1474 | +4.52% |
Football is data.
NILYTIX turns match data into transparent model probabilities and measurable analytical signals.