QUICK ANSWER

Who will win Arsenal vs Chelsea?

Arsenal 21 Chelsea. Before kick-off the model gave Arsenal 57%, the draw 25% and Chelsea 18%, with a score pick of 21. Every prediction is published before the match and never revised.

— Onside (onsidearena.com), predictions graded in public.

Every scoreline, priced

ARS goals down, CHE goals across

ARSCHE 0CHE 1CHE 2CHE 3CHE 4CHE 5
ARS 05.4%3.3%2.2%0.7%0.2%0.0%
ARS 18.1%10.4%4.7%1.6%0.4%0.1%
ARS 29.8%9.9%5.0%1.7%0.4%0.1%
ARS 36.9%7.0%3.6%1.2%0.3%0.1%
ARS 43.7%3.7%1.9%0.6%0.2%0.0%
ARS 51.6%1.6%0.8%0.3%0.1%0.0%

The engine’s scoreline probabilities — the joint chance of each exact score from the same expected-goals model behind the win call above. The outlined cell is the single most probable score (11); scores with either side on 6+ goals sum to 2.2%.

Full time · GW3 · model pl-market-v2

Arsenal v Chelsea21

Onside's model gives Arsenal a 57% chance of beating Chelsea in Gameweek 3, with 25% on the draw and 18% on an away win. The score prediction is Arsenal 21 Chelsea, from expected goals of 2.13 to 1.01. Every Onside prediction is recorded before kickoff and graded in public.

Arsenal
57%
Draw
25%
Chelsea
18%
Model pick: 21 · most probable 11
OUR RECORD — MODEL RECEIPTS →
Was the model right?
Model pick · locked pre-kickoff
21
Final score
21
Graded
+5 · exact score

Recorded before kickoff, graded by the same 5/3/2/0 scoring as every player in the prediction league. Full season record: model receipts.

SCORELINE DISTRIBUTION · ONSIDE MODEL

Top 5 likely scorelines · model receipts

EXPECTED GOALS
2.131.01
★ MODE
11
10.4%
21
9.9%
20
9.8%
10
8.1%
31
7.0%
OVER 2.5 GOALS
61%
BOTH TEAMS TO SCORE
57%

Model updated from 30 played matches. Scorelines computed from Onside's match model. The mode is the single most likely result; the full grid sums to 100%.

Why the model leans this way
Arsenal
Attack ranks #1 and defence #1 of 20 on Onside's xG-based ratings, with home advantage worth ×1.10 on goal rate.
Chelsea
Attack ranks #3, defence #12. Expected goals in this fixture: 2.131.01.

Ratings are fitted on three seasons of expected goals with time decay, refreshed after every match, and blended with the betting market where prices exist (pl-market-v2). Availability of key attackers adjusts goal rates before kickoff.

Think you can beat this call?

Lock your score above before kick-off — every pick is graded 5/3/2/0 against the model and your mates, first results from GW1 (21 Aug).

All predictionsEvery Premier League match, called by the modelWin probabilities, scorelines and expected goals for the full gameweek — graded in public.VIEW ALL →
More GW3 predictions