Saturday, 5 January 2019

What are the chances?

The first truly daft study of the year has been published, as the Mail reports...

Bookmakers are sending gamblers into debt by advertising only their most addictive products, academics claim

Bookmakers are sending gamblers into debt by advertising only their most addictive products, academics have claimed.

Gambling companies lure football fans into losing big by pushing no-hope bets with enticing odds on adverts, they said.

Research by the Universities of Warwick and Bath said customers are drawn in by the high odds, unaware that these bets will result in huge losses in the long run.

.. Co-author Dr Philip Newall, from the University of Warwick, has proposed a ‘risks warning’ system, similar to the alcohol-by-volume percentage on drinks bottles.

The claim that betting on the correct score in football matches is especially addictive is an invention of the Mail as far as I can see. The authors don't say that in their study. Betting on the correct score is classic recreational betting, usually for small stakes, but it is widely advertised on television so it tends to attract criticism from people who don't like gambling.

The authors' claim is that people are especially likely to lose money on these bets. This is trivially true in the short term as they are longshots by definition (the clue is in the odds), but the authors, who have an article on The Conversation, suggest that they are also a particularly bad bet in the long run and that people should be warned...

You can make a reasonable judgement about responsible drinking by using the percentage alcohol by volume (ABV) information on the label of whichever bottle has been opened. But how can we determine the strength of a football bet?

Gee, I don't know. If only there were some way of crowdsourcing the probability of an event occurring, based on the views of people who have skin in the game. If only this data could be summarised in a simple mathematical form. We could call - I dunno - 'the odds' or something.

The authors have other ideas, however...

In fact, “gambling harm” can also be approximated by a percentage. The “gamblers’ losses” percentage is a measure of the money bet that a gambler will lose in the long term.

Spending money on entertainment is not "harm", but never mind. It is true that we know what the average, long run losses are in the case of fixed-odds games, as the authors point out in their study. Gambling machines generally show the average payout and the house edge on roulette is precisely 2.7%.

This is not possible on skill-based games such as football betting because every player is different, but that's not going to stop the authors trying.  

They went through several seasons of Premiership football to see what would happen if people bet on final scores. Three strategies were modelled using machine learning: least-skilled, most-skilled, and random.

The “least-skilled” strategy chose what might be thought of as the worst case scenario for each match.

In other words, the least likely outcome, eg. Manchester City getting drubbed 4-0 every week.

This mirrors the returns of someone who is not merely unlucky, but is unskilled (and who may benefit from more help and advice).

It doesn't really though, does it? It mirrors a person who is anti-skilled; a person who almost certainly does not exist. The losses of this hypothetical person don't tell us anything about the losses faced by real bettors. Nor does the random scenario, for that matter.

We found that that just randomly selecting correct score bets would hit you with a strong average loss of 34.3%. But the worse case scenario was a whopping average loss of 58.9%, which came when the least skilled strategy picked very high correct scores (such as the away team winning by four goals to nil).

So what?

We believe that the very high differences in product risk across football bets should at least be communicated in some way to consumers. While further research should investigate how best to educate football fans about these different risks, reminders to just “gamble responsibly” won’t cut it.

Consumers need to be told about the risks of football bets with high odds.

Hang on a minute. What about the most-skilled bettors? How did they get on?

You won't find out by reading The Conversation article, nor by reading any of the press coverage, but if you read the study you'll see that they did pretty well.

Betting performance was in each case negative, because betting odds reflect an implied house edge (Cortis 2015). However, in each bet type, it can be seen that the most-skilled prediction performed best, the least-skill strategy performed worst, and Random was in the middle.

Therefore, in each bet type, (positive) betting skill improved performance, while (negative) betting skill worsened performance. Prediction skill mattered most in with correct score bets. With correct score bets, the least-skilled strategy returned -58.9% on average, Random -34.3%, and the most-skilled strategy -3.3%.

3.3% is about as good as it gets in gambling. It certainly beats the National Lottery and bingo, but the authors aren't demanding warnings for those. Instead, it wants to have warnings like this...

Clear as mud, aren't they? Nice skull and crossbones though! Straight out of the anti-smoking playbook. And notice how the most-skilled strategy is curiously absent once again.

Nobody bets randomly and nobody bets on what they believe to be the least likely outcome so these numbers are meaningless. Moreover, since gamblers know that they are going to bet neither randomly nor on what they see as the least likely outcome, they will ignore the warnings.

And rightly so. The odds are a far better guide to probability and risk than the figures generated by this peculiar study.

No comments: