By Tony Attwood
Two stories in the media caught my attention today. One is the reporting of Understat’s xG table which apparently shows that in the 22 league games played under Mr Arteta, Arsenal has gained (the implication is “luckily got”) nine points more than the data would expect. The newspaper didn’t quite go back to “Lucky Arsenal” as we used to hear in olden times, but did add, “That suggests a measure of fortune and that a regression to the mean is likely.”
In short Arsenal have been riding their luck and shouldn’t be in the top section of the league table, and their comeuppance will arrive in any day now. Comments of “Lucky Arsenal” should be used.
The article (which comes from the Guardian) does at least have the good sense to add that “But efficiency is something data analysis still struggles fully to elucidate.”
And yes, of course it is just possible that Arsenal have been lucky all the way through those 22 league games, although that seems to be pushing “luck” a bit far. After all, it is possible that in a lucky draw of seven numbers, the same seven numbers could come out of the bag in the same order seven times running. It is possible, but unlikely. Luck holding through 22 games? Also very unlikely.
And in Arsenal’s case the manager hasn’t just been “lucky” in 22 league games because during that spell we won six consecutive FA Cup games, and the fact that these were excluded from the calculations give us a sense that only statistics that suit the thesis are being chosen.
We had no draws in the FA Cup, four of the six games were against Premier League sides, only one was at home and two of the games were against two of the top four clubs: Chelsea and Manchester City. It was not an easy ride, we won six in a row without replays and out games included beating two of the top clubs in the land.
But you could also say under Mr Arteta Arsenal won only three of their first ten games, one being against a Championship team and the other against a team that was heading towards relegation.
And this is the real point: you can’t prove anything you want with statistics, that is certainly a fallacy, but quite often one can select statistics in order to match what you want to prove. So for example, Mr Arteta won just one of his first eight league games with Arsenal. Totally true. But also true is the fact that Mr Arteta lost just one of his first 14 league and FA cup games with Arsenal.
Here are the games.
|21 Dec 2019||Everton v Arsenal||D||0-0||Premier League|
|26 Dec 2019||AFC Bournemouth v Arsenal||D||1-1||Premier League|
|29 Dec 2019||Arsenal v Chelsea||L||1-2||Premier League|
|01 Jan 2020||Arsenal v Manchester United||W||2-0||Premier League|
|06 Jan 2020||Arsenal v Leeds United||W||1-0||FA Cup|
|11 Jan 2020||Crystal Palace v Arsenal||D||1-1||Premier League|
|18 Jan 2020||Arsenal v Sheffield United||D||1-1||Premier League|
|21 Jan 2020||Chelsea v Arsenal||D||2-2||Premier League|
|27 Jan 2020||AFC Bournemouth v Arsenal||W||1-2||FA Cup|
|02 Feb 2020||Burnley v Arsenal||D||0-0||Premier League|
|16 Feb 2020||Arsenal v Newcastle United||W||4-0||Premier League|
|23 Feb 2020||Arsenal v Everton||W||3-2||Premier League|
|02 Mar 2020||Portsmouth v Arsenal||W||0-2||FA Cup|
|07 Mar 2020||Arsenal v West Ham United||W||1-0||Premier League|
The newspaper columnists who write these things are not stupid, they know they are just picking numbers to suit themselves and they know they are doing it to wind up Arsenal supporters.
Which is one reason why this blog so often finds itself printing a list of results; just to prove we are not making it up or taking results out of context.
This season’s results look promising:
|29 Aug 2020||Arsenal v Liverpool||W||1-1 (5-4)||FA Community Shield|
|08 Sep 2020||Ipswich Town v Arsenal||W||1-2||Football League Trophy|
|12 Sep 2020||Fulham v Arsenal||W||0-3||Premier League|
|19 Sep 2020||Arsenal v West Ham United||W||2-1||Premier League|
|23 Sep 2020||Leicester City v Arsenal||W||0-2||League Cup|
You might want to take out the League Trophy match, but still the rest look quite good. But they only have a real context if you compare like with like. So in our next game we have Liverpool: let’s see how they have been doing:
|29 Aug 2020||Arsenal v Liverpool||L||1-1 (5-4)||FA Community Shield|
|12 Sep 2020||Liverpool v Leeds United||W||4-3||Premier League|
|20 Sep 2020||Chelsea v Liverpool||W||0-2||Premier League|
|22 Sep 2020||Wigan Athletic v Liverpool||L||6-1||Football League Trophy|
|24 Sep 2020||Lincoln City v Liverpool||W||2-7||League Cup|
If we keep in that League Trophy game we can see we did much better than them. And in looking at the league cup we might note we played the team who came fifth last season, while Liverpool played a League One club.
So yes you can play with statistics and they can prove lots of things. But as long as you compare like with like, they certainly can be valuable.
- Arsenal’s 25 players list is over-full – we must lose some before we can buy
- Infantino speaks at Fifa’s 70th Congress, but don’t tell the UK media
- Why Premier League spending should be much more carefully controlled
- The home and away scandal: ignorance, or cover up?
- The reason why Liverpool and Man C are ahead of Arsenal.
- How which referee a club gets has a major impact on the result of each game
- The statistical evidence that shows PGMO are biased against Arsenal
- How European football has taken up the fight against clubs breaking FFP