Correction

A long term study of online cards and casino gambling shows that, for these games of true “chance”, unless they cheat, about the only way gamblers can win is to get lucky and then stop gambling.

That is, the house, which takes a cut of all transactions, always wins and the gamblers all revert to the negative (residual) mean. The longer they play the closer they get to the negative mean.

In a two year study in the US about 10% of gamblers at casinos actually made money – but they didn’t bet for the whole two years. They came in to the casino, got lucky and then took their winnings and ran. Everyone that stayed at the table eventually lost money.

On the other hand, a five-year Taiwanese study shows that 15% of day-traders on the stock markets consistently made money, whereas 85% did not.

This implies that the stock market is not a game of chance for some. There is enough predictability for 15% of these day traders to predict what is going, on average.

Horses, dogs, trots and the like are a sort of hybrid between a game of chance and a system with predictability built in.

The races are weighted by favouritism but with odds adjusted for same. The end result is supposedly a game of “chance”.

However it’s a charade. There is residual predictability due to two factors. First, no matter what the odds are, some runners are just better. Secondly, and probably more importantly, we have race fixing and adjusting – those activities, small and large, that shift the true “hidden” odds away from what the book says.

That small number of professional punters that consistently make money at the races simply have more access to this hidden information. It’s all about information asymmetry related to the delta between the true odds and the quoted odds.

I bet it’s well less than 15% of the punters. more like 5%, at the races that revert to a positive mean on their gambling.

So while the ATO says that financial spread betting isn’t a game of chance, they are probably partially correct for 15% of the players. But the same rule should apply to the 5% or so of track punters that have a positive mean. I still think the taxation of gamblers should be according to their mean return is and not what they are gambling on.

More interesting is the idea that there is a relationship between the degree of predictability in a gambling medium and the percentage of people that have a positive mean return on gambling in that medium, over the long term.

Just think how much information those successful punters are ignoring. They must have figured out that only certain data inputs are important but that the rest is ‘fluff’ designed to confuse your average mug punter.

At the races there has been generations of professional punters and you can imagine that there is an apprenticeship of sorts, passing down the know-how of which data to take notice of – these are outcomes of the human learning algorithm.

I can imagine developing learning algorithms to compete with professional punters but the problem is that we might not have access to the key data. They keep this stuff very much to themselves and we cannot assume that all the key input data is publicly available, even for day trading and financial spread betting.

Then there is a second order approach. Rather than trying to simulate the professional punter, all you have to do is keep an eye on their trades. Let them simply be living and breathing algorithms doing your work for you. The best way to capture their efforts is through an online betting platform.

However, one would be well placed not to over-use such information. By capturing the punts of the professionals and sharing this information broadly you would pervert the market and ruin the odds on the winners. This would force the market into a period of destabilisation until some punters figured out a new equilibrium where they could get a stable return.

By which time they will have figured out that they should avoid your online platform.