On Thursday, closing the piece on the 26,726 simulated bets, we wrote something precise: simple markets are priced well, and if a margin exists it sits where a team's perception has drifted away from what that team produces. Then we named nine teams.
One matchday later, five of those teams have played. It is time to look at the score, before someone forgets.
The score, name by name
Monza was the most undervalued team in Italy: 1 point in four matchdays against 7.74 expected points. On Friday night it beat Sassuolo 2-1. The result arrived, the performance a little less so, given that the expected goals for the match read 1.61 against 2.27 for the visitors. Settled, with reservations.
Villarreal was the most extreme case in La Liga: 2 points from 7.47 expected. It won 3-1 away at Malaga producing 4.25 expected goals against 0.36. Here it settled, full stop.
Bologna was the biggest hole: 1 point from 6.94 expected. It drew 1-1 with Torino, moving its points total but playing its poorest match of the season, 0.70 expected goals. Half a point in its favour: the direction is right, the explanation is not.
Betis was among the teams we called inflated, 12 points from 6.29 expected. It won 1-0 against Getafe. But look at how: 2.61 expected goals against 0.33, meaning a performance worthy of a wide win. Wrong prediction, and the reason it is wrong is interesting.
Cagliari was the other name on the inflated list, 9 points from 5.29. It won 1-0 at Udine entering the box once in ninety minutes. Wrong prediction, and in style.
Two settled, one half, two wrong. Lecce, Atalanta, Lille and Brest have not played yet.
Why two out of five is not a failure
Here is the part worth understanding, and it is the reason this article exists.
Regression to the mean is not a prediction about a single match. Saying that Cagliari has 6 points more than its performances are worth means saying that, over twenty or thirty matches, that gap will tend to close. It says nothing about what will happen on Saturday afternoon in Udine.
Over a single match the noise is bigger than the signal, always. A team that deserves 1.2 points per game gets three or zero, never 1.2: this is why anyone promising to know who wins on Sunday is selling something.
The correct way to use a gap like this is to accumulate many similar situations and wait. The wrong way is to check it week by week and conclude, after five matches, that it works or it does not. Five matches do not answer, and anyone who read Thursday's piece knows that our serious test used 26,726 of them.
Cagliari, and the limit of the reasoning
That said, a case like Cagliari's deserves to be looked in the face instead of filed away as noise.
After five matchdays it has 12 points with 5.77 expected points. The gap, +6.23, is the largest across the five major leagues right now. It has conceded 2 goals from 8.77 expected goals conceded, and it has entered the box 13 times all season, allowing 51.
There are two possible explanations, and both are legitimate. The first: something is happening that the model does not see, for example a defence that forces shots from consistently difficult positions and a goalkeeper in a state of grace. The second: it is the normal tail of the distribution, that part of the season when everything goes someone's way, and the model is not wrong, it simply does not know when it will end.
Betis is a different and more useful case. We flagged it as inflated by looking at the points, and yet its production numbers were already growing: 12.15 expected goals produced in six matches. Against Getafe it played like a strong team. That is the lesson we take home this week: the gap between points and expected points is a crude indicator, and it must always be cross-checked with what the team creates and concedes, not only with what it collects.
The open positions, updated
After the latest matchday the list has moved. In Italy the biggest hole belongs to Bologna, 2 points from 8.02 expected, followed by Venezia, zero points from 5.48, and Monza, 4 from 8.68 despite the win.
On the opposite side, beyond Cagliari, Lecce remains with 6 points from 2.06 and Atalanta with 6 from 2.33, both still to play.
In the Premier League the new name is Tottenham: 2 points in five matches from 5.45 expected, with 2 goals scored from 5.82 expected goals produced. In La Liga Villarreal stays at -5.44 despite the win, and keeps producing like nobody else, 98 box entries in six matchdays.
The honest point
First, what we are not doing. We are not giving betting tips and we are not attaching an odds price to any of these names: the odds we collect stop at last spring, so for this season we do not even know how much the market is already pricing these divergences. Without the price, an idea is not a bet.
Second, the margin, when it exists, is small. The test we published on Thursday says that betting blindly on simple markets loses 6.14% from the bookmaker's margin alone, and that the market price gets probabilities right within two percentage points. Any edge built above that level is thin and requires hundreds of bets to show itself.
Third, the most important thing. Betting remains gambling. This column exists to tell the story of how to read a season's numbers and to keep the score in public even when it goes badly, not to promise earnings.
T.L.
Want to see how much each team truly deserves its league position? The Meritometro compares actual points and expected points. For those who want the models and not the headlines, there is the Nerd Zone.