Simulations and Penalties...

A belated Happy New Year to everyone. So as avid readers of this blog will know, I have been working on simulating tournaments or parts of them in the past few weeks and I'm very close to publishing this. I ran several 10000 iteration simulations given Ed's hints he gave on this page: http://www.football-rankings.info/2020/12/simulation-of-scheduled-football-matches.html over the weekend and the results are interesting. One thing that wasn't mentioned in the calculations but was mentioned in the comments below it was what to do with penalty shoot-outs where these are needed. Confusingly the conversation was between Ed and Ed. So before reading these comments I started thinking that I wanted to work out a better model than what I had assumed initially, that the higher-ranked team at the time of the match (working out changes in points for each match rather than using the FIFA ranking from before the tournament) would also win the match on penalties. This of course renders a winner, but certainly isn't totally representative because as "Ed" (not the Ed who ran the site) pointed out, this skews the results perhaps by as much as 5%. Actually, yesterday and today I analyzed all penalty shoot-out matches and came up with the following.

Considering all matches for which we have a FIFA ranking (from July 1993) which have required penalty shootouts, 508 to be exact, 53.5% were won by the higher-ranked team. This increases to 60.2% (123 matches) if we actually only look at matches following the 2018 FIFA ranking revamp which started using the more credible ELO ratings. While these numbers perhaps aren't enough to base statistical models on, they do show that there is a significant likelihood that the "better" team will win a penalty shootout. Ed (the one who ran the site) worked out that 56.9% were won by the better team.

I also wanted to look at whether removing friendly matches or matches played in very minor tournaments like the King's Cup or Carlsberg Cup made a difference. For this we have 355 matches with 53.0% being won by the higher-ranked team, and 56.9% being won by the higher-ranked team using only matches played following the 2018 revamp (102 matches).

Then I also looked at whether there was any difference between teams playing at home or when the matches was played at a neutral venue, because I think if we looked at that as well it would make for a more accurate model. So for neutral matches (including the friendly and minor tournaments matches), the numbers are 57.9% since 1993, and 62.0% since 2018.

The most interesting numbers came from the matches where a home team was involved. For these, the higher-ranked team only won 48.2% considering all matches, with the lower-ranked team therefore winning 51.8% of the time. That's not significant in itself, but differs quite substantially considering all or just the neutral matches. Perhaps this can be attributed to the added pressure of performing in front of your home crowd which makes missing even more significant. Those were numbers though for all matches since 1993, a time period when the rankings weren't considered to be very good, so we should take that with a pinch of salt. For the numbers for matches since the 2018 ranking took effect, the results are 56.8% for the higher-ranked home team, 43.2% for the lower-ranked home team.

So in the next few days I'll consider what I'll do for the matches requiring penalties, and then I should be able publish my first simulations, probably first for the UEFA Nations League quarter finals which I currently use to simulate with, and the CONCACAF Nations League.

So watch this space!