Gameweek 6 Predictions: Our AI Has No Draws, the BBC Panel Has 12

Forecasts as at Friday morning, 9 October 2026 · Matches Saturday 10 to Monday 12 October

Our model has called five home wins, five away wins and no draws for Gameweek 6. Yet the draw probabilities across the ten fixtures add up to 275%, which means the same model expects roughly 2.75 draws this weekend. Both things are true at once, and it says something about how a single outcome call works. We explain it below.

The BBC's three forecasters, by contrast, have gone for a draw in 12 of their 30 selections. In three fixtures (Aston Villa v Brentford, Ipswich v Fulham and Hull v Everton) none of them agrees with us.

All 10 predictions are on the fixtures page, and every probability below is for 90 minutes only.

The Gameweek 6 predictions at a glance

Standout call: Liverpool v Manchester City

We have Manchester City at 48% to win at Anfield, with Liverpool on 28% and the draw on 24%. City's five straight league wins and two wins from two away games provide the form context.

That away record is a small sample, so treat it as a trend rather than a rule. The BBC panel is split three ways: the BBC's AI also calls City, Chris Sutton says draw and the guest forecaster says Liverpool. It is one of only two fixtures where all three gave different answers. The other is Palace v Forest.

Biggest predicted upset: Aston Villa v Brentford

Our "upset" test is simple: a fixture where we back the away side and none of the three BBC forecasters does. Only one fixture passes it. We have Brentford at 46% against Villa on 28%, with the draw on 26%. The BBC's AI says draw, and Sutton and the guest both say a Villa win.

Too close to call: Chelsea v Bournemouth

This is the tightest gap of the week. Chelsea are on 38%, the draw on 36% and Bournemouth on 26%, only two points between our top two outcomes. We have Chelsea as the call, but the model is saying very little here.

All three BBC forecasters go for a Chelsea win. That makes it one of only two fixtures with a unanimous BBC call. The other is Manchester United v Tottenham. Crystal Palace v Nottingham Forest is the next closest, on a four-point gap.

The rest of the gameweek, match by match

Arsenal v Leeds United (58 / 24 / 18). Our joint-highest home probability this week. The BBC's AI and Sutton agree, and the guest forecaster goes for a draw.

Ipswich Town v Fulham (48 / 29 / 23). We back the home side. The BBC's AI and Sutton both say draw, and the guest forecaster goes for Fulham, so nobody on the panel agrees with us.

Sunderland v Brighton & Hove Albion (18 / 24 / 58). Our joint-highest probability of the weekend and the biggest single-outcome number for an away side. Brighton's recent results include a 3-0 win over Arsenal and a 5-0 win at Coventry. Only the BBC's AI agrees with us, while Sutton and the guest say draw.

Manchester United v Tottenham Hotspur (58 / 24 / 18). The BBC panel is unanimous for United. Tottenham are winless in their last five league matches and without an away win this season. Both are short runs, so be careful about reading too much into them.

Crystal Palace v Nottingham Forest (28 / 34 / 38). A narrow Forest edge. Palace have no home wins and have conceded seven goals in two home games — 3.5 per game at Selhurst Park. That is a tiny sample, not a reliable rule for this fixture. Sutton is the only BBC forecaster with Forest.

Hull City v Everton (48 / 28 / 24). A home call with a 20-point lead over the draw. The BBC's AI and Sutton say draw, and the guest says Everton, so the panel offers no support for the home win.

Coventry City v Newcastle United (28 / 26 / 46). The Monday-night game. Newcastle are our call. The BBC's AI agrees with us, while Sutton and the guest say draw.

The honest angle: why "no draws" is a warning, not a boast

An outcome call is a single forecast, not the full probability distribution. In this round, every fixture has a home or away win probability above its draw probability. That does not mean the draw probability is zero, or that choosing no draws proves the model is working well. Our published record scores the forecast outcome, using the predicted scoreline where one is recorded; this preview shows only the home, draw and away probabilities. For these ten forecasts, those outcome calls also match the highest probabilities. Our 10 draw probabilities total 275%, so we expect nearly three draws, and we have named none of them.

There are two other things to hold in mind:

• Our own numbers imply about 4.9 correct calls out of 10 (the ten top probabilities add up to 486%). Across Gameweeks 1 to 5 we got 18 right from 47, or 38.3%. On 47 matches that is too small a sample to say whether the model is overconfident or simply unlucky, but it is a gap we will keep tracking.
• We trail the BBC forecasters so far. On the AI vs BBC page, through Gameweek 5, the BBC's AI is on 24 from 50 (48%), Chris Sutton on 24 from 50 (48%) and the BBC guest on 20 from 50 (40%). We are on 18 from 47. The 47-versus-50 difference is three Gameweek 1 fixtures with no prediction recorded: Everton v Crystal Palace, Ipswich Town v Sunderland and Nottingham Forest v Leeds United. They have not been backfilled and are excluded from our scored sample. The missing predictions are confirmed; the original cause is not established by the current records. The BBC totals cover all 50 fixtures, so the samples are not identical.

Nothing in this preview is a result. It is a record of what we think before a ball is kicked, so you can hold us to it.

What to watch

We will review this gameweek after Monday night's game, comparing every call here with the result and with the BBC panel. These figures are a publication-time snapshot, not a claim that all ten forecasts are already locked. Check each match page for the latest numbers and the AI vs BBC page for the recorded calls. Track the running season record on our accuracy page, and see the full BBC comparison on the AI vs BBC page.

BBC predictions are factual attributions of publicly published forecasts. Source: BBC Sport — Chris Sutton, BBC AI and Starsailor frontman James Walsh, Gameweek 6.

FAQ

What are the Premier League Gameweek 6 predictions?

Our AI forecasts Arsenal, Chelsea, Ipswich, Manchester United and Hull winning at home, and Brentford, Brighton, Forest, Manchester City and Newcastle winning away. Every home, draw and away percentage is in the table above.

Who does the AI favour in Liverpool v Manchester City?

Manchester City, at 48%, against 28% for Liverpool and 24% for a draw.

Which Gameweek 6 match is the hardest to call?

Chelsea v Bournemouth: 38% Chelsea, 36% draw, 26% Bournemouth.

How accurate is the FootyPulse AI this season?

18 correct outcomes from 47 scored matches (38.3%) across Gameweeks 1 to 5, as at publication. Three Gameweek 1 fixtures have no recorded prediction and are excluded. The live figure is on the accuracy page.