Three straight wins isn’t enough for us to conclude that a team is brilliant
In an article last week I used Bayes theorem to argue that a rather short run of losses without a win, around about three, should be enough to convince you that your favourite team is going to be crap this season.
I’ve had some great feedback on the article and I thought it worth updating after the weekend’s football.
It seems that some people are convinced that Chelsea, Manchester United and Manchester City are all ‘Brilliant’, having each secured three wins in a row. I would prefer that my article from last week does not contribute to any such hysteria.
While three wins in a row is certainly good, I feel obliged to point out, that according to my model, we still can’t conclude that these teams are Brilliant. The probability that a team is Brilliant after three straight wins is:
Pretty good, but not up to the 95% level that scientists traditionally accept as a mark of statistical significance.
How many matches does it take before we can be reasonably certain of a team’s Brilliance? We can solve this problem by increasing the number of wins one week at a time.
So, it isn’t until around six straight wins that we can safely conclude that a team is Brilliant. Last season, Manchester City won their first five matches and were firmly established as the bookies favourites to win the Premier League. By the end of the season it looked very different for City.
The reason it takes six wins to convince us that a team is Brilliant and only three losses to convince us that a team is Crap is that losses by good teams should be rare. Seeing a sequence of events we believe to be common, i.e. wins by a good team, provides less useful information, than seeing a sequence of rare events, i.e. losses by a team we believed were good.