Framing Effect Bias – Part 2

Last time we looked at the Framing Effect Bias, and how the language used in data presented to us can significantly affect how we process that information.

Framing is not just about language, though. How we frame information in other ways can also change how we perceive it.

Take This Scenario

Let’s imagine you feared nothing more than earthquakes, to the extent that, when relocating, the chance of a major earthquake affecting your new home would be the determining factor in your decision. Would you move somewhere where, each day, there was a one-in-ten-thousand chance of a major earthquake? What about if it was one-in-seventy-three-thousand?

The issue with such calculations is that, in such a narrow timeframe, the difference between the scenarios is hard for the human brain to quantify.

Framing it differently changes this. According to the United States Geological Survey, where the data above comes from too, Anchorage will experience a major tremor (bigger than 6.75 on the moment magnitude scale) once-every-thirty-years, whereas Salt Lake City will be largely spared, only seeing a major tremor once-every-two-centuries.

Depending on how much you feared the consequences of a major earthquake, you may decide to give both options a miss (after all, it’s also a choice between being surrounded by grizzly bears or Mormons), but what’s clear is that thinking about the data in a different timeframe allows us a very different perspective.

Onto Betting

The effects of different timeframes on how we see information can have a dramatic – and often unprofitable – effect on how we bet.

Take another scenario. You are watching a tennis match. I offer you odds of $1.77 on a player winning the next point. Do you take it? If I said that I would offer you $1.16 on the same player to win the entire match, would you take that?

It’s hard to answer those questions without knowing more information. The players in question are Roger Federer and David Ferrer.

Across their careers, they have played 17 times. In these matches, Federer has won 56.38% of contested points, which translates to the odds of $1.77 above. The $1.16 is the average price, according to my records, that has been available on Federer at the start of his matches against Ferrer (with a range from $1.03 (Miami 2006) to $1.35 (Madrid 2010)). Federer has won all 17 of their encounters.

This example is, of course, illustrative. It’s never possible to have this kind of historical perfect-vision when betting. It shows, though, how difficult it is to assess probability when it is presented in such different timeframes. A player with only a little over a 50:50 chance of winning a point may not seem like a strong bet to win a match, but even a small advantage in every point, when extrapolated over a match, can turn that player into a near certainty.

In the example above, our chances of making the best decisions depends on our ability to put together, or aggregate, lots of smaller data points. Aggregation of data, though, has its own risks.

The Trump Card

Take the 2016 Presidential Election. Data from FiveThirtyEight shows that in 48 out of the 50 most-educated counties, Clinton increased the vote share won by Obama in the 2012 election. Conversely, in 47 of the 50 least-educated counties, her vote share decreased. It would be tempting to extrapolate from this data, therefore, that the less-educated were voting for Trump. However, this would be an example of an ecological fallacy, where an analysis of a group is used to draw conclusions about an individual. In the case of the 2016 election, things are more complicated than they initially seem, with race, income, geography, and media consumption all showing significant correlations with voting patterns.

We need to beware of similarly fallacious extrapolations when betting.

The Darren Weir Effect

Horse racing can provide an example. Across the world, large, successful training and breeding operations tend to dominate the world’s biggest races. Because of this, it can be easy to overvalue runners from those types of stables when pricing up individual races. That’s not to say that horses representing large training operations don’t have some advantages over other horses – the greater experience and wealth of their handlers may afford them better travel arrangements, for example – but it’s likely that such advantages are only marginal, and that individual differences in the ability and form of horses, rather than aggregate differences between their stables, are going to be the determining factors in deciding which horse wins. Therefore, it is past form that needs to be at the centre of pricing up these races, not the colour of the silks on the jockey’s back.

Similarly, how often are New Zealand rugby union players overvalued? Whilst it is true (although this might be difficult for any Australian to accept) that New Zealand is the best rugby playing nation in the world – three Rugby World Cups and a near-80% strike-rate is impressive for a nation of fewer than 5 million people – it is false to assume that individual New Zealand players are therefore necessarily better than their international contemporaries. Yet Northern Hemisphere clubs will pay high premiums for players from the South. How much of this value is because of the cache they bring from their home nations, though? As punters, we need to beware of falling foul of the same aggregation bias. How often have we overvalued a team because they have a new, star player from a glittering background?


Whenever we are assessing information as punters, we need to be aware of how that information is being framed in terms of the level-of-aggregation. Often, seemingly insignificant data is crucial when extrapolated over time. For example, an F1 car that can lap a tenth-of-a-second quicker will be 8 seconds ahead after 80 laps. Conversely, though, data presented to us in its aggregate form can seem significant in predicting events, when it is of little value. For example, draw and track biases can often dominate pre-race discussion, drowning out analysis of relative horse form.

Successful punters are always aware of the type of data they are dealing with, and will constantly question the extent to which it is relevant in their decision-making. Often, being able to identify where others are over-reacting to some of the aggregation biases above, and simply betting against the crowd, is enough to ensure a profit.


As a passionate sports’ fan and punter, Jack has written about sports and betting for over a decade, winning the Martin Wills Award for racing journalism in 2002 and writing Winning on Betfair for Dummies, first published in 2006 and now in its second edition, having sold over 35,000 copies in two languages.

Related Articles

Framing Effect Bias – Part 1

Humans, no matter their intelligence, are cognitively lazy. We have to be.

Framing Effect Bias – Part 3

In my last two articles, we have looked at framing biases. Specifically, how the language that is used, and ...

Illusion of Knowledge

In 2015, Atir and Dunning, researchers at Cornell University, ran a series of experiments to investigate the phenomenon of ...