Kneeslide
EXPLORATION 010 / PREMIER LEAGUE · 2026-27

Simulator

One match. Ten thousand possibilities.

Results through 2026-11-08
AFC Bournemouth
Nottingham Forest
Home win56.5%
Draw22.6%
Away win21.0%
0123456+
6+<0.1%0–6+2 of 10,000 simulations<0.1%1–6+2 of 10,000 simulations<0.1%2–6+3 of 10,000 simulations3–6+0 of 10,000 simulations4–6+0 of 10,000 simulations5–6+0 of 10,000 simulations6+–6+0 of 10,000 simulations
5<0.1%0–57 of 10,000 simulations0.1%1–56 of 10,000 simulations<0.1%2–57 of 10,000 simulations<0.1%3–57 of 10,000 simulations<0.1%4–51 of 10,000 simulations<0.1%5–51 of 10,000 simulations6+–50 of 10,000 simulations
40.3%0–424 of 10,000 simulations0.5%1–456 of 10,000 simulations0.5%2–443 of 10,000 simulations0.3%3–430 of 10,000 simulations0.1%4–414 of 10,000 simulations<0.1%5–41 of 10,000 simulations<0.1%6+–41 of 10,000 simulations
31.0%0–3108 of 10,000 simulations1.9%1–3193 of 10,000 simulations1.8%2–3176 of 10,000 simulations1.1%3–3124 of 10,000 simulations0.5%4–341 of 10,000 simulations0.2%5–320 of 10,000 simulations<0.1%6+–38 of 10,000 simulations
23.0%0–2305 of 10,000 simulations5.5%1–2558 of 10,000 simulations5.1%2–2474 of 10,000 simulations3.2%3–2332 of 10,000 simulations1.5%4–2136 of 10,000 simulations0.5%5–247 of 10,000 simulations0.2%6+–223 of 10,000 simulations
15.7%0–1578 of 10,000 simulations10.7%1–11,059 of 10,000 simulations9.9%2–1990 of 10,000 simulations6.1%3–1628 of 10,000 simulations2.8%4–1279 of 10,000 simulations1.1%5–1104 of 10,000 simulations0.4%6+–134 of 10,000 simulations
05.5%0–0560 of 10,000 simulations10.2%1–01,064 of 10,000 simulations9.5%2–0951 of 10,000 simulations5.9%3–0575 of 10,000 simulations2.7%4–0284 of 10,000 simulations1.0%5–0104 of 10,000 simulations0.4%6+–040 of 10,000 simulations
Elo
1587
1505
Proj. points
1.92
0.85
Proj. goals
1.86
1.04

Simulated match

Generating match…

OPEN DATA. OPEN METHODS.

A forecast, not a promise.

Results through 2026-11-08. A dated snapshot, not a live feed. 2026-27 only. Model 2.0.0. Each season samples plausible team strengths, then simulates match scores. It cannot anticipate every injury, transfer or tactical change.

Model, validation & qualification assumptions

How KElo measures team strength

KElo (Kneeslide Elo) is a results-based rating of a club’s strength. Beating a strong opponent earns more rating points than beating a weak one; an unexpected defeat costs more than an expected defeat. A higher number means a stronger team within this system. These are rating points, not league points, and 1500 is a starting reference rather than a guaranteed league average.

Where the ratings start

Ratings are reconstructed by replaying Premier League results in date order from 2014–15. Every club starts at 1500 when it first appears in that history. Clubs already in the history retain their previous rating when they return, subject to the season-start adjustment below. A newly seen promoted club starts at 1500; its lower-division results are not used. Domestic cups and European matches do not enter KElo.

How each match changes KElo

Before a match, the two ratings determine an expected result between 0 and 1. The actual result is 1 for a win, ½ for a draw and 0 for a loss. The rating moves according to the difference between the actual and expected result:

Rating change = 20 × (actual result − expected result).

The multiplier, K=20, controls how quickly the rating reacts. If equally rated teams have no venue adjustment, the expected result is 0.5: a win adds 10 points, a loss removes 10, and a draw leaves both ratings unchanged. If a team’s expected result is 0.75, a win earns only 5 points; a draw costs 5, and a defeat costs 15. The opponent gains exactly what the other team loses.

For the home team, the expected result is 1 ÷ (1 + 10(away rating − home rating − home adjustment) / 400). The away team’s expected result is the remainder to 1. This is an expected result score, with draws counting as half, rather than a standalone win probability. A 400-point adjusted rating advantage gives an expected result of about 0.91.

A 1–0 win and a 5–0 win produce the same KElo change against the same opponent at the same pre-match ratings. Goal difference, xG, injuries, transfers and player availability do not directly change KElo. Recent results matter because they update the rating after every match; older results remain embedded in the starting rating for the next match.

What happens between seasons

At the start of each included season, participating clubs move 10% of their distance from 1500 back toward that reference. The new rating is 1500 + 0.90 × (previous rating − 1500). For example, 1700 becomes 1680, and 1300 becomes 1320. This reduces how much of a previous season’s advantage or disadvantage carries forward. It does not mean a club loses 10% of its total rating or league points.

Home advantage and the forecast

This model uses a 0-point home adjustment when updating KElo. That means venue does not change the expected result used in the rating update. It does not mean home advantage is absent from the forecast. The match and season forecasts separately use different home and away scoring rates, alongside the KElo gap and recency-weighted goals and xG. The displayed probabilities therefore come from the combined scoring model, not KElo alone.

How to read the rating

Compare clubs within KElo and at the same snapshot date. The rating uses only results available through that snapshot’s cutoff. It is our own calculation, not a ClubElo rating, and the two scales are not interchangeable. Strength of schedule uses separately sourced ClubElo ratings; the table predictor and match simulator use KElo.

How the settings were chosen

The Elo settings and scoring weights were selected from 4,800 combinations using 1,520 rolling pre-match predictions in 2019–23, after a 2014–19 warmup. A separate retrospective audit covers 1,140 matches in 2023–26. Those seasons were excluded from this selection, but had already been inspected under the previous model: they are not a fresh blind test.

Audit outcome log loss: 0.98387 versus 0.98406 for the previous model (lower is better). Multiclass Brier score: 0.58683. The difference is small, varies by season, and does not establish a reliable accuracy gain. Home adjustment and regression remain uncertain estimates, not proven constants.

Season projections were tested before the season and after 90, 190 and 280 matches. A persistent team-strength uncertainty of 70 Elo points was selected on training-season 80% interval scores, which penalize both missed outcomes and unnecessarily wide ranges. Across the three audit seasons and four cutoffs, mean final-points error was 6.86 points; nominal 80% intervals covered 77.5% of team totals. Repeated cutoffs are correlated, and three seasons are a small sample. Historical snapshots are reconstructions, not forecasts published at the time.

xG providers and coverage vary by season. Fixed recency decay, goal-rate priors and initialization of unseen clubs remain modeling assumptions. Full sensitivity results, expanding-window checks and limitations are in the backtest download.

Scorelines use Poisson probabilities with the low-score correction chosen in training. Finite score support extends to 20 goals; probabilities are normalized. Season standings include actual results, then simulated scores. Exact ties use head-to-head points and away goals; residual ties use a modeled neutral playoff, with a mini-tournament approximation for three or more teams.

European columns show league-position probabilities: top four for Champions League and top five for Europe overall. No extra performance spot or cup-place redistribution is assumed. Cup winners and European titleholders are not simulated. Top-four and top-five chances remain available in the team inspector.

Sources: OpenFootball results, Premier League fixtures, Understat historical xG, xgEdge recent xG, European qualification rules. The fixture snapshot is checked against every completed score and all 380 home/away pairings. Dates can change.

Data sources & downloads