This is a piece I’ve purposefully waited to recommend – as the knockout stages of the European Championship are going to be the perfect example of this phenomenon. JLikens has some cool and abstract thoughts and I’m sure I’ll be recommending some of his other posts in the future. For now I’m looking at this one (link).
I think most people will accept that there are occasions where two teams play a game and the result isn’t a fair reflection what transpired. In other words the best team doesn’t always win. And that’s in a single game. As we increase the number of games that are played we also increase the odds of this happening at least once. The World Cup and European Championships both require a team to win three straight games in order to claim the title, meaning there are three chances for an upset to occur. Therefore, whilst the best team are still the favourites to win the competition, nothing is guaranteed.
Whilst the linked post looks at a regular NHL season and the Stanley Cup playoffs as opposed to football it provides a perfect demonstration of randomness.
The methodology is simple but brilliant and I love those:
Identify and build a model to represent the spread of talent within the league
Validate the model using some known values to make sure it works
Simulate a huge number of seasons to give a representative result
Analyse
Basically I think this is the gold standard we should all be striving for.
This, to me, is one of the most telling quotes from the post:
“The best team does very well in general, but the range in outcomes is considerable. It wins its division a majority of the time yet still manages to finish dead last every now and then (about once every 200 seasons)”
I’m not saying this would happen in the Premiership, I suspect the spread of talent is much wider due to the lack of a salary cap, but it serves as a timely reminder that random happens. I certainly doubt that the best team wins the Premiership title every year though.
As a final aside the comments section for that post is hilarious. Some people just don’t get it.
James' Blog 
Just some quick calculations for Group A using UEFA’s site stats …
If we take TSR as the indicator of quality of a team then Group A looks like this after the group games …
CzechRep : 0.467
Greece : 0.292
Russia : 0.576
Poland : 0.638
Of course this is all after the fact and what you would want is to see how this then can be translated into the upcoming knockout rounds and possible predictions … furthermore, it is over 3 games only against teams in the same group, hardly a good indicator of quality, but a good source of data to determine pre-tournament quality of the teams (excluding friendlies) is not really available.
There a numerous other factors not taken into account but hey … too many to note here.
Germany comes out with a TSR of 0.514 from Group B (arguably a more difficult group thus one could assume they faced better opposition, so not fair to compare between the two groups, but I’m sure the two TSRs could be adjusted to reflect individual group quality somehow) so based on basic comparisons could be seen as favourites for their game v Greece. Here’s hoping for an upset. 🙂
This is relatively obvious, would be more interesting to compare two closer quality teams.
Very basic and not very exact, but fun none the less. 🙂 Of course you can’t post this kind of stuff as a blog entry cuz you’ll be vilified by your readers for not using proper procedures and making conclusions based on hot air. hehe
Yer I considered attempting adjusting TSR during the group stage but I’m too lazy. I’d rather correct for score effects over a Premiership season if I’m lucky enough to get the data.
Greece are a truly terrible team though and even worse than their TSR suggests. If we break down their games into game state we see this (numbers are minutes)
Ahead 45 Tied 104 Behind 121
We know that:
1) Teams take a larger proportion of shots when they’re behind as they have more of an incentive to shoot
2) Greece were behind far more than they were ahead
As a result Greece’s overall TSR (0.292) is very likely to be artificially high compared to the number we should be interested in – TSR when the game is tied
For context the worst TSR over a Premiership season is 0.362
Hi,
Well … here’s Greece’s TSR while the game has been tied …
0.258.
v Poland: 6 shots of a total of 16 for both teams … includes a missed penalty.
v CzechRep.: 0 shots of 1 in total … goal came in 3rd minute of game.
v Russia: 2 shots of 15 for both teams.
(TSR: 8 from 31 = 0.250.)
There could be errors by me cuz I had to read the match commentaries … goals that broke tie situations were counted in the shots … it is all fairly basic stuff. 🙂
I think these numbers are correct, they include blocked shots of which the Greeks had a ton during the first half of the Russia game. I checked the next 45 mins and I think UEFA calculates them under their “Total attempts” stat, which I used to get the TSR in the above post.
Further adjusting of the TSR to account for group strength could be done by using FIFA World Rankings (shudder) grades of the team and their opponents … not sure how to do that weighted calculation …
Russia 13 (zone 9)
Greece 15 (zone 11)
CzechRep. 27 (zone 16)
Poland 62 (zone 32)
(rank on June 6th 2012 – zone rank only includes UEFA teams.)
This would allow for inter-group comparisons. I don’t really like the FIFA World Rankings as it also includes friendlies where teams can make more substitutions than normal or might experiment with regards to players and tactics. I guess other ranking systems could be looked at (e.g. ELO) but I don’t know the finer details to how points/positions are calculated for each system.
So … yup Greece have been terrible. 🙂 I’m still hoping for an upset, love cheering for the underdog. 🙂 Too bad Greece are missing their most “explosive” player, ah well.
woops … that’s a TSR of 0.250!!
8 of 32 – missed one shot after double checking the data.
🙂