Wisdom of the Crowd

In this article, Jack Houghton discusses the Wisdom of the Crowd. A societal concept that is highly relevant to Betfair.

We think you’ll love this discussion so read more below.

His book takes as its source inspiration a 1907 article by Francis Galton, submitted to the journal Nature, which describes how nearly 800 participants, when asked to guess the weight of a slaughtered and dressed ox, were accurate to within 0.8% of the real weight, when considering the “middlemost” (median) number.

Surowiecki goes on to provide many examples of where crowd wisdom has trumped the perceived insight of experts, including the use of markets within the betting exchange to predict the outcomes of elections, which he claims have outperformed traditional polling methods in their prescience.

Surowiecki’s book captured a surge of interest with Galton’s theory, which has seen academics and policy makers researching ways to apply the wisdom of crowds to challenges as diverse as the movement of pedestrians to the prevention of genocide.

As punters, fully understanding the observed phenomena of crowd wisdom is crucial. After all, betting markets, more than most other areas of life, frequently gather together a diverse group of people to predict the outcome of something. When a tentative Betfair punter puts the first offer of a bet up in a new market, what follows are a series of counter-estimates, until a liquid market is produced that provides Galton’s “middlemost” number – the median of all punters’ prediction as to the likely chance of any outcome.

And as reported in our article on favourite-longshot bias, data suggests that Betfair punters are every bit as good as Edwardian weight-guessing Ox-fanciers. In one small study, carried out in 2005, which looked at 65,000 results for British racing, horses seemed to win at a rate predicted by their odds. $1.50-shots won two-thirds of their races, $2.00-shots won half, and $100.00 won one per cent of their races.

This could put some off betting altogether. After all, if Betfair markets are near-perfect indicators of chance, what’s the point of betting?  How can you ever identify a value bet – where the odds you are getting represent a percentage chance of winning that is less that the actual chance of winning – if the market is so goddam efficient?

The answer lies in Surowiecki’s required conditions for collective wisdom to exist. Among other conditions, he claims that the crowd requires diversity, independence, and decentralisation. In other words, in markets where there are few people betting, and where those few are familiar with each other’s opinions, the accuracy of predictions will go down.

Whilst Betfair markets will often offer the conditions that Surowiecki requires for crowd wisdom, and so will be highly predictive when taken as a collective over a long period, they will be significantly less accurate when it comes to any individual market and outcome. Successful punters can identify these markets and exploit the short-term variation.

Nonetheless, the long-term prescience of Betfair markets should be sobering. When betting, we are taking on a large collective of other individuals, and we need to be confident that our selection methods and staking strategies are good enough to outperform most of them.

Next time we’ll look at the wisdom of crowds in more detail, including how to combat it when using Betfair.

In Part 1 we looked at the concept of crowd wisdom and how it applies to Betfair markets, being sobered by their likely accuracy and pondering how we can combat that accuracy when we bet.

For all the occasions where crowds have demonstrated to be brilliantly predictive, punters should find hope in all the times the crowd has been wrong.

In the run-up to the 1992 UK General Election, every major polling company predicted a win for Neil Kinnock’s Labour party.  Even considering the margin-for-error, all those polls were significantly inaccurate, with John Major’s Conservative Party winning a surprise parliamentary majority.

Questionable polling seems to be on the rise, with Britain’s vote to leave the European Union, the United States’ election of Donald Trump, and numerous Senate and House races in the US in the 2014 and 2016, not universally precited by pollsters.

It’s important not to overly state the failures of such polls – most had the UK referendum as a close-contest and, likewise, polls in the US predicted some scenarios where Trump could win – but nonetheless many were poor at predicting voter turnout, which led to feeble forecasting.

In his 2004 book, The Wisdom of Crowds, Surowiecki examines the failings of some polling, claiming that Iowa Electronic Markets (IEM), a tool used in the US by academics to give students real-world experience of markets which operates in a similar way to Betfair, had consistently outperformed polling companies since its inception.  IEM, though, didn’t do so well in the 2016 elections either: just two weeks before election day it viewed a Clinton win as a 93% near-certainty.

The failings of polls – and of markets that try to outperform polls – can be explained by revisiting the conditions that Surowiecki claims are necessary for a crowd to be wise: diversity, independence, and decentralisation.

Political polls are based on relatively small samples – often around 1,000 people.  Whilst polling companies attempt to ensure the diversity of those polled, this diversity is limited because only the type of people prepared to answer polling questions are ever considered.  Some have argued that voting to leave the European Union and voting for Donald Trump were, for many, dark secrets, that those polled would not want to share with pollsters.

Betting markets around elections typically have far more participants and, unlike polls, the participants are expressing an opinion about who will win, rather than how they will vote.  Both factors suggest that betting markets will be more predictive in forecasting votes.  The issue, though, is that punters in these markets are unlikely to be decentralised or independent.  Although there are lots of them, they are drawn from a relative narrow pool of those interested in both politics and betting.

Furthermore, they will be working from similar information – some of it supplied by polling companies – meaning that betting markets around elections can become an aggregated and refined version of perceived wisdom, rather than the diverse, independent, and decentralised crowd that Surowiecki claims is required to be truly predictive.

In a 2011 Swiss study by Schweitzer and others, researchers observed the failings that can occur within a crowd when their information sources lack diversity.  Asking participants to predict the length of the Italian-Swiss border, researchers found that the accuracy of the median prediction decreased as participants were given more information about other participants’ estimates.

Similar things happen in financial and betting markets.  Information cascades, where participants, all acting on the same information, artificially skew the pricing within a market, occur all the time, seen most catastrophically in financial market bubbles which, invariably, burst.

In racing betting, a horse’s odds shortening can often – counter-rationally – cause other punters to want to back the horse, in the belief that the market somehow “knows” something that they don’t, and that the horse must be a “sure thing”.

This kind of short-term skewing of betting markets can provide opportunities for the well-informed punter – who is able to adopt a more objective view of value.

The approach of questioning crowd wisdom is explored in the now little-read, but aptly-named, 1995 betting manual, Against the Crowd, by Alan Potts.  In it, Potts argues that a degree of contrariness is desirable.

Potts’ view is backed up by data.  A 2006 study looked at 65,000 horses who had run in the UK, categorising over 5,000 of them as either “steamers” (their odds shortened by 5% or more in the last two hours before the race) or “drifters” (their odds lengthened by 5% or more in the last two hours before the race).

It found that, although both groups of horses had similar strike-rates, winning around 20% of their races, backing the steamers would have resulted in a 349-point loss, whereas the drifters returned a 231-point profit.

Next time we’ll look at other ways that you can identify when the crowd is likely to get things wrong, and how you can use all the principles of crowd wisdom.

So far, we have looked at the concept of crowd wisdom and how it applies to Betfair markets, and their long-term accuracy.

We’ve also seen, though, that crowds can sometimes fail – especially where they lack diversity, independence, and decentralisation – and that it can be advisable to take a contrary opinion.

The challenge for punters, then, is to be able to identify, in advance, where the conditions are ripe for such a market failure.

One option is to think about markets whose participants lack diversity and are not decentralised.

Take tennis.  If you bet on the men’s final at the Australian Open this year, between Roger Federer and Rafa Nadal, you were likely competing with thousands of other punters, from all around the world.  If Surowiecki is right, it is likely that the odds were an accurate reflection of chance.  You might, therefore, have bet in that market and won, but advocates of crowd wisdom would argue that, eventually, betting in similar markets would make it hard to return a long-term profit.

However, betting in the early stages of the women’s draw at the Sydney International, a few weeks before the Australian Open, is likely to see you competing against far fewer punters, from a much closer-knit group, suggesting that the market might not be so accurate.

This perhaps explains why some punters specialise: if they develop quantifiable methods of analysis that allow them to correctly assess probability in a niche area, then they can outperform the market, because that market is not big enough, nor diverse enough, to be perfectly predictive.

Another option is to think about markets that lack independence, especially of information.

In the run-up to the Rugby World Cup final in 2015, the Australian TAB reported a huge weight of money for Australia, at $3.00, to beat New Zealand, especially in the last 24 hours before the final.  This is despite the odds on an Australian win being $3.60 on Betfair.  This pricing rick can be explained by what is sometimes called patriotic punting: bettors being driven by the emotion of wanting to see their nation prevail and betting accordingly, often striking their bet close to the event.

Examples like this are seen the world over.

The odds on most national soccer teams shorten in local betting markets in the minutes before kick-off and they demonstrate a lack of independence in participating punters, as they are all effectively acting on the same piece of “information” – I want my country to win and therefore they will win.

Aside from the influence of patriotism, a narrow range of information can artificially skew many other markets, especially where the media has influenced public opinion.

An unusual example of this can be found in markets that ask punters to bet on the highest temperature likely to be reached in a summer.  In recent years, the odds available on the highest temperatures in those markets have been shortening.  To many punters, this would seem to make sense.  After all, there is a scientific consensus that the world is heating up, year-on-year, and the media is full of stories of freak weather events caused by global warming.

However, global warming is based on measures of average temperatures, and whilst these have been increasing, this has not always correlated with a continued increase in the record high-temperature reached in any single country in a single year.  The record high temperature in Australia, for example, was recorded in 1960.

Punters then may be well advised to seek markets that are likely to be inefficient because of a lack of diversity, independence, and decentralisation.

The usefulness of understanding the wisdom of crowds doesn’t stop with helping punters choose the markets to bet in, though.  Its principles can also be used to help any punter improve their own decision-making processes.

Many researchers have investigated whether trying to replicate the conditions of the crowd as an individual can provide similar levels of predictive accuracy.  The idea is that you adopt mental processes that force your decision-making to consider more than one answer, creating a “crowd within”.

In one study, participants were asked to make two estimates in answer to a question, over a timespan of three weeks, giving average answers consistently more accurate than either of their two specific ones.  In other studies, participants who were asked to consider contrary-but-plausible answers to questions were similarly able to improve their accuracy.  Very recent research has successfully experimented with a fascinating approach that sees participants averaging their own guess with what they think the popular view will be. Read More

The research behind methods that try to harness the “crowd within” is not compelling: there are varying levels of accuracy reported in different studies, with some reporting limited benefits to the methods employed.  Part of the difficulty for researchers is in pinpointing the type of answers that such techniques are best at predicting, and it is likely that we are someway off being able to quantitatively apply these models-of-thinking to punting.

Nonetheless, the principle of adopting a humble mindset when punting – where are the areas that my analysis might be wrong? – must be valuable.

Arrogance is not at all a desirable trait.

and if the wisdom of crowds tells us anything, it’s that we should at least be mindful of this when placing any bet.

Related Articles

Framing Effect Bias

Published author, Jack Houghton, is our resident expert on Betting Psychology and is on hand to discuss the framing ...

Illusion of Knowledge

Jack Houghton discusses the Illusion of Knowledge.

God-Shaped Hole