Confirmation Bias - Rules Of Thumb

Hopefully this is entirely unnecessary but let me start out by saying I am not advocating murder, the darlings referred to in what I say next are a metaphor for pet theories. 

In the betting world there are a myriad of theories, "don't bet odds on" "don't trade in running" "don't back a horse to do something it has not done before", whatever sport you bet on theories will abound. 

That said here goes, I think you should "murder your darlings" various great writers are quoted as having used the same phrase when talking about the secret of good writing. 

The idea is that if a writer themselves particularly likes a certain part of what they have written, they should probably take it out as it is likely to be self-indulgent and to subtract from others' enjoyment of the work. 

In terms of bookmakers, betting and in running betting I am not saying that all pet theories are wrong or that those who bet on that basis never win, but we need to do a check, and if we prove ourselves wrong, to move on to something else. 

Confirmation Bias

My last post warned of the dangers of confirmation bias, the phenomenon in psychology that we are more likely to pay attention to information that backs up what we already thought than to information that contradicts us. 

My point here is that we have to expect that at least some of this bias has helped form our ideas, so if we want to be successful in betting we need to be able to eliminate at least most of it.

The expression “rule of thumb”, is said to originate from a time when it was allowed under the law for a man to beat his wife as long as he used a stick that is no thicker than his thumb. 

A quick Google search reveals that this is thankfully not true, but I think a kind of rule of thumb is something that could be useful for us in testing whether our selections are suitable. 

To eliminate this bias, there needs to be an element to our process for deciding our bets that is quantitative rather than qualitative. In other words, we need to have numbers that describe what we bet on and we need to make rules about what these numbers should be. 

If our proposed bets don't meet these rules then we mustn't make the bets, even if we think at the time that we should. If the rules we set are sensible then I believe this should help to remove the bets we were going to make as a result of the bias.

Of course, sometimes we will stop ourselves from making a successful bet like this, but in my mind, it is certain that this will happen less often that it will happen that we leave an outcome that our pet theory is telling us should win and that turns out not to win.

The reason for this is if we make rules without a particular outcome in mind then we are unable to build cognitive biases into them. 

We might say for example that we won't back a horse unless it has at least a place in more than 50% of the races it has run this season and last and unless the number of races is greater than three.

Or we may say that we will never back a team that has conceded more than 8 goals in its last six games. I am not saying that these rules are the right ones to use, readers must decide on their own. 

What I am saying is just that it is helpful to have some sort of quantitative rule of this sort. If such rules are stuck to then I am confident that this will help betting returns over a period of time, as long as the rules used are sensible.

The St Leger Example

Perhaps the example of Windshear in the St Leger from the last post is a useful one to describe what I mean.  As I wrote then, I had seen a lot of good in the form and I thought he had a good chance, but with the benefit of hindsight I realise that I had only been seeing the positives and not the negatives in the form; I had fallen victim to confirmation bias. 

The St Leger is a race for three-year-olds and is run over a distance of 1m 6f so 14f. The vast majority of the horses in the race haven't run over 14f yet, so it is not an option to compare their performances over the distance of the race, but I do think it is important to pick a horse that will stay the distance.

I looked back at the horses that finished in the top three in the race for the ten years previously to last year's running and one thing I noticed was that most of the horses (22 out of 30) had won over a distance greater than 10f. 

Of the ten horses that had won the race in that time, only three, Encke in 2012, Mastery in 2009, and Scorpion I 2005 had failed to win over a distance greater than 10f before winning the St Leger. 

With the benefit of hindsight then, perhaps I could have used this as my rule for making my St Leger selection; if I had done this I would not have been on Windshear whose longest distance win at the time was over 10f at the time. I am not saying that the way to pick a horse to win the St Leger is to just find a horse with an entry that has won over a distance greater than 10f; there will be many examples of such horses that do not go on to win. What I am saying, is that this is a good rule to apply after a main selection process, to weed out any selections that have been produced by confirmation bias. 

My advice would be for each race, before looking at the cards, look at previous runnings of the race and make a rule that the majority of winners fall into. Then conduct your usual selection process and if the horse that you select doesn't follow your rule, don't have a bet on that race. Obviously the rule will be different for different types of race and completely different when we are looking at different sports. 

You may want a single rule or you may want a couple, but the best ones will be strict enough to rule some of the entrants out of the race out, but loose enough that it doesn't rule out many past winners.

The racing fans amongst you may be racking your brains to remember if Kingston Hill, the winner of the St Leger had won over more than 10f prior to that victory or not. I can tell you that he hadn't, although the second and third horses, Romsdal and Snow Sky both had. 

As I have said, these rules of thumb will sometimes rule out a selection that goes on to win, the idea is that they do this far less often than they rule out a bad selection. I think I will be happy to continue to use this rule for this year's St Leger, although I think I might need to add some others to it as well. 

For his part, Windshear was still there or thereabouts when turning for home in the St Leger but was outclassed by the first two and beaten to third by a short head. It seems that his connections agree that he is not a 14f horse now, he has had five runs since the St Leger and four of those were over the distance of 12f (his most disappointing run was fifth of eight over 13f). 

He is not a bad horse and he showed this around Doncaster back in March when he worked hard to return to winning ways under Sean Levey, the jockey that rode him in his first-ever race. Although I think it was a mistake to back him for the St Leger I am still following his career with interest.