The table points one way, the model another
Athletic Club host Real Sociedad at San Mamés on Sunday in La Liga's 11th round, and the pre-match evidence is split in a way that makes the fixture more interesting than the table alone suggests. Real Sociedad sit fifth on 19 points from 10 games; Athletic are eighth on 12 from nine. The visitors have the better recent points haul and the higher league position. Yet Nutmegly's AI read makes Athletic the lean, with a 42.7% home-win probability against 31.0% for Real Sociedad and 26.2% for a draw. That is a plurality, not a majority: the model gives the draw-or-away combination 57.3%. The home tag is a starting point, not a verdict.
The home case rests on the venue split, not on overall superiority. Athletic's home record is two wins, two draws and one defeat, with nine goals for and four against in five league games. Real Sociedad's away record is three wins, no draws and two defeats, with 12 goals for and 11 against in five. The matchup numbers in the data underline the same point: Athletic average 1.80 home goals per game, while Real Sociedad concede 2.20 per away game. That is a home attack against a leaky away defense, and it helps explain why the model leans home even though the table says otherwise.
Athletic's most recent stored result was a 4-0 home win over Getafe on 25 October. That is the standout score in a listed run that otherwise includes a 1-3 defeat at Levante, a 0-0 at Valencia, a 0-0 at home to Alaves and a 1-1 at home to Elche. The run is not a surge: one win, one defeat and three draws, with three clean sheets. It is also thin on goals, with one goal or fewer for Athletic in four of those five matches. The 4-0 is real evidence, but it is the exception in the sample, not the rule.
Real Sociedad's form is the strongest counterargument
Real Sociedad's recent run is the clearest reason to doubt the home lean. The four most recent stored results are all wins: 2-1 at home to Levante, 5-2 away to Malaga, 3-2 at home to Deportivo La Coruna and 3-2 away to Valencia. Before that run, they lost 0-3 at home to Atletico Madrid. The wins are high-scoring and not clean: Real Sociedad scored at least twice in each of them and conceded in all four. Across the listed five, they have no clean sheet. That profile fits a team that can win without controlling the scoreline, and it is exactly the kind of form that makes a 31.0% away probability look more than a formality.
The points-efficiency signals are even more direct. Over the stored eight-match samples, Athletic have taken 1.50 points per game and Real Sociedad 2.38. In the La Liga sample, the gap is wider: 1.20 for Athletic and 2.40 for Real Sociedad. The current table tells the same story: Real Sociedad have 19 points from 10 games, with 19 goals for and 18 against; Athletic have 12 from nine, with 12 for and nine against. If the model were ranking teams by recent league performance alone, the lean would be away. The home read has to overcome that.
There is also a direct warning in the data: an external forecast points in a different direction from the home lean. That does not make the model wrong, but it is a reminder that sample definitions and model inputs can produce different reads of the same fixture. The head-to-head sample is two matches, so it cannot settle the argument either.
The missing pieces: lineups, absentees and rest
The pre-match picture is incomplete in a way that matters. No lineups are confirmed, and the injury report flags six potential absentees. That does not prove either side will be weakened, and it does not identify who might miss out. It does mean the model's inputs could change before kickoff. A home lean built on venue splits and recent scores is more fragile when the actual XIs are unknown. The data's own counterpoints list the same concern: lineups are not fully settled, and the potential absentee count is high enough to add single-match variance.
Rest is not a differentiator in the stored record. Both sides last appeared seven days before kickoff, so there is no rest advantage to cite. The data also warns that missing or longer-than-21-day gaps should not be treated as evidence of extra rest. That is a useful discipline: a gap in the record is not a tactical edge. The same applies to the head-to-head sample, which is too small to carry weight.
The watchpoint: can Athletic's home defense hold?
The match turns on a simple contrast. Athletic have conceded four goals in five home league games; Real Sociedad have scored 12 in five away league games. Something has to give. If Athletic's home defensive record holds, the model's 42.7% home probability has a clear base: a low-conceding home side against an away side that has conceded 11 in five on the road. If Real Sociedad's away scoring carries over, the same matchup looks much more open, and the 31.0% away probability and 26.2% draw become live. The model's over-2.5 probability is 56.8%, which sits closer to the open version than to a controlled home win.
My read is that the home lean is a venue-and-matchup argument, not a form argument. Athletic's listed form is one win in five, with three draws and a 1-3 defeat. Real Sociedad's is four wins in five, with at least two goals in each victory. The model is effectively saying San Mamés and the away defense's leakiness outweigh that form gap. That is defensible, but it is not comfortable. The strongest counterevidence is not the table alone; it is the combination of Real Sociedad's points efficiency, their away scoring and the external forecast disagreement.
The question to watch is whether Athletic's home defense can suppress a Real Sociedad attack that has scored 12 away goals in five. If Athletic take the lead and keep the game in the 1-0 or 2-1 range, the home lean will look like a reasonable read of the venue split. If Real Sociedad take the lead and force Athletic to chase, the model's own 57.3% draw-or-away probability becomes the more relevant number. That is the conditional edge of this fixture: the home case is real, but it depends on the game staying within the home defensive pattern their record suggests.