Sports betting and me
I was seven when I placed my first sports bet, or rather had my Dad place it for me. He told me to pick three horses in the 1973 Grand National to back each way (meaning I got paid if they finished in the top 4). I don’t think my selection algorithm extended much past randomness but clearly I ran well, or my selections did. Spanish Steps came in fourth behind L’Escargot in third, but best of all Red Rum won. Over the year I’ve dabbled to varying degrees. I’ve written in the past about a specific exploit I used to fund in part my college years, and a spell in Hong Kong exploiting some holes in the pari mutuel on horse racing there.
McGregor - Mayweather
In recent years however I’ve restricted my gamble exclusively to the poker tables. A few years ago when McGregor fought Mayweather, I knew the odds on McGregor were way too short and those on Mayweather correspondingly too long. But while most of my friends got in Mayweather for varying amounts from a little flutter to almost allin, I didn’t have a single penny riding on the fight. Afterwards I did experience a tinge of regret at having missed out, and this came back to me recently when considering whether to bet on the outcome of the American Presidential election.
Enter Biden
All of the data seemed to suggest that it was highly unlikely Trump could win re-election. He was running 6 points behind in the polls, more than the margin by which they underestimated his support last time, and he was well behind in several key marginal states. The fact that a lot of people had voted early by mail meant a late swing similar to 2016 was unlikely. The fact that Biden was urging his supporters to vote early and Trump was telling his to wait til polling day also conferred a clear advantage to Biden. Additionally there was no Comey moment, and attempts to scandalize Biden didn’t seem to be as effective as they were against Clinton. Nate Silver’s modelling data put the likelihood of a Trump victory in the 10% region, yet on the day the betting markets were making it more like 40%. All the anecdotal data I had also seemed to point to a likely comfortable Biden win. Two of the three Americans known to me personally who had told me they voted for Trump last time told me they had already voted by mail for Biden.
My biggest ever sports bet
All of this led me to placing my biggest ever sports bet on the day of the election across a number of different sites. I wasn’t feeling all that great about that about 12 hours later when Florida came in for Trump and he moved to a clear favourite on the betting markets. Not that I thought Biden was necessarily going to lose, but had I waited I could have got on at a much better price now.
I would never claim anything approaching expertise these days as a sports bettor. But there were two things I have picked up from friends who are that seemed relevant in this spot:
(1) who is betting on each side is vital information. As a kid I was told by a professional gambler that one should never bet on a horse owned by the Queen, because whatever it’s prospects it always went off at far too short a price because of the weight of patriotic money on it’s back from people betting it purely because it’s owned by the Queen. Some events attract a lot of mug money where people bet with their heart, and elections are often such events. A quick gander at social media was enough to convince me that most of the mug money was being wagered on Trump, and most of what I suspected to be the smart money on Biden. One notable exception to this was Bryan Paris, who not only won bigly on Trump in 2016 but came on the Lock-In this time to explain why he thought Trump would win again.
(2) Markets tend to overreact to early data points. An example of this is when one team scores early in a match their price often drops too much. This certainly seemed to be the case here, and indeed there seemed to be a bit of recency bias built into Trump’s price at every point in this election cycle, with a lot of “look what happened last time”.
Cognitive biases
It has been pointed out that while closer than a lot of people (me included) expected, it wasn’t as close as many seen to think, and this would have been a lot clearer if the results had come in in a different order, with for example the marginals that Biden ended up winning all coming in at the start rather than the end.
At the end of the day (or days), I was rather relieved to not only have won my bet, but to have resisted pressing the panic button after Florida. One thing that did occur to me afterwards though is that I may have placed too much value on the anecdotal data at my disposal. Yes, 2 of the 3 Trump voters I knew switched sides, but
(1) Three people is a rather meaningless sample size
(2) Most of us realize we create our own social network, populated for the most part with people who share our views on most things. Given this bias which I know I am also prone to, I obviously don’t regard as in any way significant that almost nobody I knew was voting for Trump. I may however have assigned too much significance to the fact that two of the three Trump supporters I knew were switching sides. My thinking here was since they were Trump supporters they were exempt from any friend selection biases I might have, but on closer reflection this doesn’t really hold up. It’s highly probable that I’d find it much easier to find enough common ground with a mild Trump supporter than a rabidly committed one to maintain a friendship, and therefore the Trump supporters I knew were far more likely to switch sides than average.
Retirement....not just yet
Despite the favourable outcome, winning my biggest ever sports bet did not encourage me to repeat the experiment often in the future. Despite winning, it felt a lot closer than I imagined it would be. This led me to conclude that I am too prone to all the usual biases to have a very clear edge, at least ahead of time. As I consumed the data as it came in on election night, I did feel I interpreted it very well in real time. Several times I was convinced the market had overreacted to the latest significant data, and in every case I was proven correct with an almost immediate market correction. It was also clear to me that the markets were often internally inconsistent: for example Trump’s overall price tended to be shorter than the cumulative prices of all the states he needed to win, and Biden’s overall price was conversely often longer than it should have been, so it felt like there were lots of opportunities to make plus Ev in running bets. This is something I might experiment with in future, but otherwise it’s very much a case of don’t give up the night job.
0 comments:
Post a Comment