Level on points, opposite shapes
There are two ways to read the same six points. Everton sit ninth and Ipswich tenth, level on points after four league matches, and that is roughly where the resemblance stops. Everton have won one, drawn three and lost none; Ipswich have won two, lost two and drawn none, with 17 goals arriving across their four games. Saturday at the Hill Dickinson Stadium is therefore a test of a familiar editorial question: do you trust a tidy unbeaten record, or the noisier evidence sitting underneath it? Nutmegly's model has an answer, and several of the form signals supplied alongside it keep arguing back.
Everton's quiet, unbeaten start
Everton's league season so far reads: a 2-0 home win over Crystal Palace on 22 August, a 1-1 draw at Bournemouth, a 2-2 home draw with Manchester United, and a 0-0 draw at Tottenham on 12 September. Five goals scored, three conceded, no defeats. At home specifically they have played twice, won once and drawn once, scoring four and conceding two. Nothing about that is spectacular; everything about it is stable. One caveat on the wider numbers: the stored results also include a 0-1 defeat at Tottenham on 24 May from an earlier period, and across that fuller eight-match stored set Everton's points rate is 1.00 per game. That figure spans more than one season and is not a current-season form line.
Ipswich: goals at both ends, points in bursts
Ipswich's four league games have been considerably louder. A 2-1 home win over Sunderland, a 2-5 defeat at Manchester United, a 0-2 home loss to Liverpool, and a 3-2 win at Crystal Palace on 12 September. Seven scored, ten conceded. Their away record is one win and one defeat from two matches, with five scored and seven conceded β 3.50 goals against per away game in the stored sample, which is two matches and should be read as a small sample rather than a rule. The stored five-match points rate of 1.80 per game includes a 3-0 win over QPR on 2 May from an earlier period, so treat it as a wider sample, not a season summary.
What Nutmegly's model actually says
Generated on 16 September, the AI read leans home: Everton 52.0%, draw 25.4%, Ipswich 22.6%. That is a majority, but only just β 48.0% of the probability mass sits with a draw or an away win combined. The model also places over 2.5 goals at 49.1%, which is worth pausing on, because Ipswich's four league games have averaged 4.25 goals while Everton's have averaged 2.0. The case for the home lean is straightforward enough to state: Everton are unbeaten in four, unbeaten at home, and have conceded twice in two home matches, while Ipswich have shipped seven in two away games in the stored sample and have lost two of four. Forecasts are estimates, not results β the model is pricing a plausible pattern, not promising one.
The counterevidence is not trivial
Most of the signals supplied alongside the prediction point at the visitor. Ipswich carry the better recent points efficiency, 1.80 per stored game against Everton's 1.00. In the Premier League sample available, Ipswich rate 1.50 points per game against Everton's 1.20. The home-versus-away attacking matchup also flags Ipswich, pairing Everton's stored home scoring rate of 1.80 per game β a five-match home sample that is not limited to this season β with the goals Ipswich concede on the road. The rest signal is a thin hook at best, and the same notes warn against reading gaps in the record as rest advantages. Two further gaps matter more: there is no head-to-head sample at all, so that dimension cannot check the read, and the lineups are not confirmed, which the model itself lists as a reason the outcome could shift. An injury feed reporting no absentees is an absence of information, not proof that both squads are at full strength.
The question to carry into Saturday
So the match hinges on which team's pattern the game takes on. If it settles into Everton's register β tight, low-scoring, decided by a single moment at either end β the 52.0% home lean looks reasonable and the draw-heavy record simply continues. If Ipswich drag it into the open, two-way traffic their four league games have produced so far, then that combined 48.0% for draw or away win starts to look thin, and the 49.1% over 2.5 figure becomes the most exposed part of the model's read. The specific thing to watch: whether a defence that has conceded three goals in four matches sets the terms, or whether a side whose league games have averaged 4.25 goals turns a tidy unbeaten start into a shootout. Which version of the same six points shows up on Saturday?