Trust the numbers and only the numbers

Einstein had a maxim: “as simple as possible, but no simpler”.  (The maxim is falsely attributed, but don’t let that detract from the story).  A punter has had the maxim pinned on a noticeboard above his desk, throughout his betting journey. It makes him focus on producing a betting system that is pure, unadulterated, and numbers-driven: he updates his ratings, he produces tissue prices for a match, where there is a significant discrepancy, he bets.

Gone are the days when he tries to account for all the noise that surrounds men’s tennis. Players getting new coaches; different ball-types and marginally different surfaces at different tournaments favouring different players; erroneous statistics about first-serve percentages and time-on-court; rumours about players adopting new gluten-free diets.

These stories, and much besides, are now ignored. The numbers are uncontaminated; the numbers don’t lie; trust the numbers and only the numbers. Remember what Einstein said.

It’s now late-summer in 2016, and Djokovic is causing a problem. In recent months, he’s become, according to our punter’s ratings, the greatest player of all-time, with an Elo rating that tops out at over 2550.  Until he crashed out early at Wimbledon, Djokovic held all the grand-slam titles. But there are numerous rumours surrounding the world number one that are becoming difficult to ignore. Family “issues”, various injury woes, and disharmony in his coaching team are all being cited as reasons for what appears a sudden drop in form.

Fast forward to the end of the 2017 season. Djokovic is still the best player in the world according to ELO ratings, just. However, since that ignominious departure at the hands of Sam Querrey at Wimbledon, Djokovic has won only three tournaments, the last two being lowly ATP250 events. Our punter continues to make a profit, but has consistently struggled to deal with Djokovic in his pure-numbers-only approach.


The Strength of ELO

One of the strengths of ELO ratings is that they recognise the generally-sound axiom that “form is temporary, but class is permanent”. Unprofitable punters – no matter what sport they are betting on – often rely too heavily on recent information when making their bets. They remember a horse’s unlucky second-place in its last race; or a rugby team’s most-recent scintillating backs-display; or a batsman who hasn’t got out of single-figures in his last ten innings.

Punters should fix their recency bias and begin to consider the entirety of a competitor’s form when assessing likely future performances.

ELO ratings force you to focus on the longer-term: whilst a string of poor performances does cause a decline in a player’s rating, implicit in the model is a recognition that form is, by nature, variable, and that even inexorable declines are usually gradual. A few poor tournaments won’t see a tennis player’s Elo rating plummet.

What if, though, circumstances like those that have plagued Djokovic in recent months, mean a decline in form is more sudden, and potentially terminal? How does the number-purist punter deal with these situations? How do you know when one decline is temporary, whereas another is permanent?


A flexible mind is needed

As a tennis punter – not unlike the character in our story – who relies on numbers and feels uncomfortable whenever I’m faced with information that is difficult to quantify, I recognise the difficult in answering these questions.

One approach is to ignore what you can’t quantify and carry-on-regardless. I know some punters who take this approach. Their rationale is that, if information is not easily quantifiable, then it is open to bias, and so you do as well to ignore it, rather than interpreting it incorrectly. Sure, they argue, there will be times when this approach costs you, but these times will be balanced out by the other times when you inaccurately “assess” the unquantifiable information you are dealing with.

As a numbers guy, this approach is instinctively attractive to me. If it’s attractive to you, too, though, it’s worth conducting a little thought experiment.

Let’s say Roger Federer is playing Andy Murray and on ELO ratings, Federer is around 70 points superior to Murray. That suggests Federer has around a 60% chance of winning, which translates to odds of $1.67. Federer is available at those odds. Do you back him? The numbers say not to: there is no value in the odds offered.

 

What if you now find out that Murray has been struggling with a hip injury? Does that change your decision?  Perhaps you think that Murray’s injury concerns are enough to introduce value into Federer’s odds, or perhaps you are steadfastly sticking to the numbers: after all, how do you quantify the effect of the hip injury? Presumably, if he has chosen to play, Murray thinks the injury is manageable.

Okay, what if you learn that Murray has lost both arms in a freak accident but still intends to play? Are you still steadfast? What if he’s dead?

This thought experiment reveals that, of course, there are certain times when the numbers are not enough, or, more accurately, where the numbers are unable to account for all the information. In these circumstances, a more flexible mind is needed.


Thing's cannot always be quantified

Unfortunately, there is no easy way to deal with information that is difficult to quantify. Clearly, as soon as you start trying to adjust Elo ratings based on unreliable information, you are likely to struggle. If you read on Twitter that a player is experiencing marital strife, what do you do? Reduce their Elo rating by 50? By 100? The rumour’s reliability, and the extent to which it will affect a player’s form if true, make computing it nigh-on impossible.

In the case of our imaginary tennis punter dealing with the Djokovic-problem, he chose to create an Elo rating based only on more recent results. His rationale was that, although this would decrease the size of the data-set driving the rating, and therefore its reliability, it would more accurately reflect how Djokovic’s problems were reflected in his current tennis form.

That seems a sensible solution. After all, the models we create to help us more accurately predict punting futures are just that, models. They are simplifications of reality that allow us, hopefully, to assess the probability of different outcomes consistently and unbiasedly. As Djokovic shows us, though, sometimes the sporting world is more complex than these simplistic models account for, and we need to find ways of accounting for this.

How we deal with this complexity as punters is an inexact science. It’s important we don’t lose the desire to quantify, but we must recognise that such things cannot always be quantified with certainty. A suppler approach – albeit one that is guarded, for fear of introducing bias and unreliability – is likely to bring better results in the long-term.

There’s nothing wrong with Einstein’s falsely-attributed adage, of course. Making your approach to punting “as simple as possible” remains something to strive for. However, we must remember the caveat, “but no simpler”. Sometimes the numbers alone are not enough. Novak teaches us that.


Related Articles

Illusion of Control

Superstition and sport seem inseparable. Whether it is Steve Waugh carrying his grandfather’s red hanky in his pocket, Tony ...

Illusion of Knowledge

Jack Houghton discusses the Illusion of Knowledge.

Overconfidence

In The Wisdom of Crowds by James Surowiecki, he argues that, provided you get a large and diverse enough ...