A promise we kept exactly once
Four of our six soccer leagues could not see a single transfer. We found that out this week, a month after telling you the opposite.
The August post was honest about the problem. Our model learns a club's strength from last season's goals, pulls it halfway back toward the league average, and has no idea a striker moved in July. So we anchored the Premier League to the bookmakers' season-points lines, because those prices contain every signing, and the table stretched out at both ends. All of that happened and all of it holds.
What the post did not say, because nobody had looked, is that the same file for La Liga, Serie A, the Bundesliga and Ligue 1 was sitting there with every club set to null. The EFL Championship had no file at all. You were reading five tables built on last season's goals and nothing else, straight through the busiest window of the year.
What a bookmaker knows that we don't
A season-points line is a hand-typed number, and it barely exists outside England. That is why only one league ever got one. Match prices are different. Every book prices every fixture with the squad that will actually take the field, and football-data.co.uk publishes those prices, from about eight books, one free file per league, updated as matches are played.
So we take the closing price on each match, strip out the bookmaker's cut, and work out how many points the market expects each club to take from the fixtures it has already been priced for. Then we ask our own model the same question about the same fixtures. Where the two disagree, we lean our rating toward theirs.
Both numbers face an identical schedule, which is the part that makes the comparison fair. It also runs twice a day now, so when the January window shuts you will not be relying on anyone here remembering to do anything.
Then we put Arsenal on 84 points
The first version of this was wrong, and it is worth showing you how.
Arsenal had played two league games. Across those two, the market had them taking 2.33 points a game, so we multiplied by 38 and calibrated their rating to a season of 88. Our table came out with Arsenal on 84 points and a 70% title chance, up from 49%.
The same bookmakers who priced those two matches also publish a season line for Arsenal. It says 78. We had taken one number from a market and used it to argue with another number from the same market, which is a reliable sign you have built something backwards.
The multiplication was the mistake. Those two Arsenal fixtures were at home to a promoted club and away at Aston Villa, which is not an average schedule, and 2.33 a game against it was never a season rate. It is a capped lean now. It cannot move any club's rating by more than 12%, however loudly the inputs shout.
One more thing fell out of that. Where a league has a real season line, the line wins. On the Premier League, the only league where we can score both methods against each other, the hand-typed line lands on 79.6 points against a market call of 78, and the match-odds version lands on 73.3. So the Premier League table you see today has not moved at all.
Who the market rates and we didn't
PSG are a 53% shot at Ligue 1, up from 40%. Real Madrid go from 28% to 32% in Spain and gain five projected points. Inter go 49% to 53%. Juventus pick up three points and Barcelona four. Every one of those clubs moved the same direction, because a rating that regresses everyone toward the middle is always going to be shy about the teams at the top.
The Championship is where it stings, though. Wolves and West Ham were both playing Premier League football in May, and relegation is not only a smaller budget. It is a Tuesday night in November, rain coming in sideways, a few hundred in the away end who took a half-day to get there. Our model opened both clubs 16th and 18th in that division before a ball was kicked, because a club arriving with no history in a league gets priced off the relegation zone, and that rule cannot tell the difference between Wrexham coming up and Wolves coming down. A squad that was at Anfield in April, projected to finish below Preston.
The market never had that problem. Wolves move to 11th and West Ham to 13th, and both gain seven points on the projection. If you support either, you were entitled to be annoyed at us last week.
What it means for your call
These are still early numbers. Nine matches priced in Germany, twenty in Italy, thirty in Spain. The lean is deliberately gentle while the sample is thin, and it gets sharper every week as more fixtures get priced.
Every figure here went onto the Ledger the moment it published, stamped and waiting to be graded, and the version we got wrong is written up there too. If you think the market has someone badly placed, that PSG at 53% is generous or that Wolves still deserve better than 11th, the tables are open and you can call the finish yourself.