Where the mistake comes from
A bookmaker is not stupid and neither is the market. But both respond, largely, to two things: recent results and what the public expects.
At the end of August, with two or three matches played, recent results are the worst information available. A team that has won twice 1-0 while conceding three expected goals sits in the table alongside a team that won on merit. The odds for the next match, though, tend to treat them the same.
That is where price parts company with value.
The measure we use here is expected points. You take every match, look at the chances both sides created, work out how often in a hundred that performance would have produced a win, a draw or a defeat, and turn it into an average number of points. Then you add up the season.
If a team has six points and deserved two, the difference is neither merit nor failure. It is noise, and noise eventually goes away.
Who has taken far more than they earned
Figures computed on the matches played through Sunday night.
| Team | P | Points | Expected points | Gap | |---|---|---|---|---| | Hull (Premier League) | 2 | 6 | 1.81 | +4.19 | | Lazio (Serie A) | 2 | 6 | 2.86 | +3.14 | | Atletico Madrid (La Liga) | 3 | 7 | 4.17 | +2.83 | | Sevilla (La Liga) | 3 | 6 | 3.37 | +2.63 | | Real Betis (La Liga) | 3 | 6 | 3.54 | +2.46 | | Newcastle (Premier League) | 2 | 4 | 1.72 | +2.28 | | Everton (Premier League) | 2 | 4 | 1.75 | +2.25 | | Strasbourg (Ligue 1) | 2 | 3 | 0.79 | +2.21 | | Lille (Ligue 1) | 2 | 4 | 1.81 | +2.19 | | Monaco (Ligue 1) | 2 | 6 | 3.83 | +2.17 | | Udinese (Serie A) | 2 | 4 | 1.94 | +2.06 |
The extreme case is Hull. Six points from six, a perfect record, and a level of performance worth less than two on average. How did they manage it? From both ends at once: they scored three goals from 1.96 expected, and above all they conceded nothing at all while giving up 2.96 expected goals to their opponents.
Three goals' worth of chances conceded, none of them taken. That is the number that does not repeat.
Lazio are the same story in Serie A. Six points, two goals scored, none conceded, and 2.81 expected goals given up. Two 1-0 wins built on a clean sheet the numbers do not explain.
And then Strasbourg, on three points from performances worth 0.79, having conceded 5.30 expected goals in two matches.
Who has taken far less
| Team | P | Points | Expected points | Gap | |---|---|---|---|---| | Brest (Ligue 1) | 2 | 2 | 5.35 | -3.35 | | Villarreal (La Liga) | 3 | 2 | 5.15 | -3.15 | | Monza (Serie A) | 2 | 0 | 3.15 | -3.15 | | Nottingham Forest (Premier League) | 2 | 1 | 3.98 | -2.98 | | Valencia (La Liga) | 3 | 1 | 3.79 | -2.79 | | Coventry (Premier League) | 2 | 0 | 2.46 | -2.46 | | Crystal Palace (Premier League) | 2 | 0 | 2.38 | -2.38 | | Torino (Serie A) | 2 | 0 | 2.26 | -2.26 | | Racing Santander (La Liga) | 3 | 4 | 6.15 | -2.15 | | Lens (Ligue 1) | 2 | 3 | 5.02 | -2.02 |
Brest are the clearest picture of what it means to be unlucky at both ends. They created 6.28 expected goals in two matches and scored four. They conceded 2.08 and let in four. Both halves of the pitch went against them, and the table says two points.
Monza have no points with 3.11 expected goals created. The problem is not the attack: they have conceded seven goals from 2.56 expected goals against. Seven from two and a half. That is not a defence collapsing, it is a defence everything went past.
Lens have produced 8.07 expected goals in two matches, the highest figure on this list, and have three points.
What it really means, and what it does not
Precision matters here, because this is the point where this kind of analysis usually gets bent out of shape.
It does not mean Brest will beat anyone next weekend. Over a single match the noise is still bigger than the signal, always.
It does not mean Hull are a bad team either. Two matches do not tell you what a side is worth, and a goalkeeper saving everything can keep saving everything.
It means something narrower and more useful: right now the table is not a good estimate of what those twenty-four teams are worth. And because odds chase the table, for a few weeks the prices on these teams will be pulled in the wrong direction.
Those who have won above their performances will be priced shorter than they deserve. Those who have lost below their performances will be priced longer.
A note on transparency, and a promise
Two things we want to say plainly.
The first. You will not find odds in this article, and the reason is that we do not currently have active access to bookmaker prices. So we cannot tell you whether a single price is good value: we can only tell you where the market historically tends to be wrong. We would rather say that than pretend to have data we do not have.
The second. This analysis is published before the next round is played, and it stays here with the date on it. In a few weeks we will come back to these two lists and check, without excuses, whether the teams at the top have slowed down and the ones at the bottom have recovered. If we were wrong we will write it as prominently as we are writing this.
A statistics site that does not keep count of its own mistakes is an opinion site with charts.
See the numbers in full. In the Meritometro you will find the verdict on every match based on chances rather than the scoreline, and in the Nerd Zone you can rebuild the expected points of any team in the big five leagues.
This article analyses the relationship between performances and results. It is not an invitation to bet. If you gamble, only stake what you can afford to lose.