The Betfair Prof: "Forget the ear plugs! There's some noise you can't get rid of at the Grand Prix!"

The Betfair Prof / Leighton Vaughan Williams / 27 May 2008 / Leave a Comment

The Betfair Prof, Leighton Vaughan-Williams, tells us why it's all about the noise when it comes to prediction markets...

It's about 40 years since a computer program hit the headlines by claiming to be able to work out what would have happened if Muhammad Ali had taken on Rocky Marciano when both were at their peak, and just seven years since the 'Virtual' Champion Hurdle, Champion Chase and Gold Cup was all that was available to replace the thrill of the real but abandoned 2001 Cheltenham Festival.

If it isn't going to happen in real life, there's something appealing about knowing what would have happened if only it had! A sort of sophisticated 'Pools Panel', one might say, with a bit of technical wizardry thrown in. For the record, Rocky knocked out 'The Greatest' in the 13th round of a bruising encounter, while Istabrag, Tiutchev and Legal Right showed what happens when hard logic is showered with a sprinkling of stardust.

All good fun, but one of the problems with the idea lies with something which may be worth considering in implementing any betting strategy. It boils down to what economists call 'noise', or the tendency of any closed system to be afflicted by random factors which cannot be predicted in any systematic fashion by a model. Over the long run, noise tends to even out, so that a good model will overestimate and underestimate performance indictors in equal measure. The predictions will be unbiased. In the short run, however, the more noise in the model, the more risky will be any bet based on the prediction.

If there is no noise at all, i.e. no random variations, then any mistakes in predicting the outcome of an event are the sole fault of the forecasting system. An example is the effect of gravity on an apple. A model of the way the world behaves will predict that if you drop the apple it will fall. There is no noise in the prediction apart from the minor effects of air turbulence and the like. However, take a race like the Monaco Grand Prix, and the influence of random factors reduces the confidence in any estimate, however sophisticated the forecasting system. It may get it right in the long run, but the long run can be a very long time, and certainly a lot longer than the time it takes to run the race.

Followers of forecasting systems must, therefore, take account of the amount of noise in the system before deciding how much to stake, and how much confidence to invest in any individual prediction.

Full consideration of all the noise at the Grand Prix might even have saved one Betfair trader from offering [200.0] about Lewis Hamilton winning after he got too close to a crash barrier! After all, what's one early flirtation with a crash barrier when you're racing in the rain at Monaco?

In a controlled computer environment, I guess that Adrian Sutil might also have clinched fourth place in his Force India car (at least if you turned the program on with a handful of laps left), but then again what do computers know about being shunted up the backside in the closing stages by the defending Formula One Champion?

What indeed?

Professor Leighton Vaughan Williams is the Director of the Political Forecasting Unit and Betting Research Unit of Nottingham Business School, Nottingham Trent University, and Editor of the Journal of Prediction Markets

Click here for the Betfair Politics Zone