What international soccer has in common with the March Madness of US basketball
Football stats and philosophy from Kevin Pullein
This is how I try to pick the winner of an international tournament. I make a shortlist of countries that I think are capable of winning any competition if they have a good three or four weeks.
For a European Championship those countries are Spain, Germany, Italy, France, Holland and Portugal. For a World Cup I add Brazil and Argentina. Then I pick one that seems to be friendless in betting markets.
I do not care why they are friendless. I do not mind if the last few results have been bad. Or if the players are said to dislike each other, or the coach, or everybody.
I just think that any squad from those countries has the potential to win any tournament if things go their way. And over seven games the ball is as likely to roll for them as for anybody else.
This strategy nearly worked in 2010 when 12-1 Holland were runners-up at the World Cup, and it did work in 2016 when 22-1 Portugal won the European Championship. Many other times I have felt foolish – for example, last year when I predicted that Argentina would win the World Cup. At big odds, though, you only need a payout now and again.
I may or may not know much about football. I try to act as though I do not. In practice I am behaving like somebody who has a bit of knowledge but not too much. Perhaps by accident I have stumbled on something. Because there is evidence that such people do well with predictions in other sports.
In 1981 Tim Trowbridge set out “once and for all to determine who knows the most about college basketball”. He asked family and friends in Kent, Ohio, to predict the winner of every game in the NCAA March Madness. His was not the first basketball bracket, as these things are called, but it became the longest-running.
The 2019 edition of March Madness started this week. It is a 64-team knockout tournament (68 if you include the preliminary phase) to decide the national college champions. People will be checking brackets in workplace competitions across the United States.
But do they answer the question of who knows most about college basketball? Not according to Tom Adams, author of Improving Your NCAA Bracket with Statistics. He says: “After a few years, the running joke was that the office bracket pool winner was the person in the office who knew the least about college basketball.”
The joke was good enough to get laughs but not quite right.
Psychologist Tina Kiesler asked people to answer a 25-question quiz on basketball. Then she asked them to predict match winners in March Madness. Afterwards she compared performance in the quiz with performance at prediction.
Generally, people who did badly in the quiz also did badly at prediction. But not as badly as the worst predictors of all, who were the people who had scored best on the quiz. Kiesler found the same sort of thing when she turned to other sports.
It should not be surprising. There is compatible research into economics and politics. Philip Tetlock, for example, found that political experts predicted developments no better than “a dart-throwing chimpanzee”.
We can call those basketball predictors experts and novices. The novices knew little or nothing about basketball. The only way they could try to predict match winners was by guessing. The experts knew a lot about basketball, but it did not make them any better at picking winners. Actually it made them worse.
The person in any group who knows most about basketball is likely to do well in a quiz. The person who knows least is likely to do badly. A quiz is probably the best test of knowledge. On any subject, not just basketball. It will examine your awareness of all sorts of things that have already happened. But those things might be worse than useless as a guide to what will happen next. The weird and wonderful ones almost certainly will be.
There were basketball predictors between the novices and experts. They knew something about basketball (more than the novices) but not a lot (less than the experts). And they were better at picking winners in March Madness. The best predictions of all came from people halfway between the novices and experts (though even they would not have beaten bookmakers).
A graph showing the relationship between basketball knowledge and accuracy of basketball predictions would look like a glum person’s mouth. Either side of the middle it would droop downward. On one side, though, the lowest point would be beneath the lowest point on the other side.
The people with intermediate basketball knowledge would know that this college was usually better than that college. Which is why they made better predictions than the novices, who were guessing about everything. But they did not have the arcane knowledge of the experts, who mistakenly thought it would help them to make even better predictions – and ended up with forecasts that were worse than guesses.
The intermediately knowledgeable had no idea that in the last few weeks this team had been cold and that team hot. Or that some other team with a good record overall had a bad record in slow contests, which is what their next opponents would try to give them. Or anything about the mind games of coaches.
The intermediates knew something about basketball but far from everything. Anybody who knows a lot, about basketball or any other sport, can effectively become an intermediate if they accept that much of what they know about the past will tell them nothing about the future.
Follow us on Twitter @racingpostsport
Like us on Facebook RacingPostSport