“If you’re looking for picks you’re in the wrong place, find a tout with a narrative to make a strong case” – “Bet the Process” podcast theme song

Yesterday I had the honour of appearing on the “Happy Hour” webcast with legendary advantage player Captain Jack Andrews and Andrew Mack, author of the excellent “Statistical Sports Models in Excel” series of books. It was my first time doing a live broadcast but I think it went very well, with some interesting discussions and audience Q&A.

I wanted to expand on something that came up at two different points in the session – once when talking about how to validate a prop model with no historical odds data, and once when talking about political betting markets where the lack of standardized data and the uniqueness of each election make analytical approaches very tough.

The word “narrative” gets a bad rap in the analytics community. It’s generally used to mean a conclusion to which amateur bettors jump, with only a superficial level of thought and analysis, to drive their betting decisions. The problem with most narratives is that they are either a) false, b) true but not predictive, or c) predictive but already reflected in the market odds. The BTP theme song, if it can be forgiven for biting a hip-hop classic, provides a textbook example: “blindly assuming a team must be tanking”.

Given this, I’m sure some listeners’ ears perked up when they heard me recommend last night that they look for narratives!

Perhaps this is my fault for taking a provocative term and using it in a slightly different context. In situations where it’s impossible to fully build and back-test a model, you’re left trying to reconcile two numbers: your number and the market’s number. If you have the Giants -3.5 and the market has them -2 then either you are right and the market is wrong, you are wrong and the market is right, or somewhere in between. Over time your results will tell the tale, but you need to decide how to bet right now. In Bayesian terms, you need a prior for “is my number better than the market?”

This is where narratives can be useful. Simply put, can you find a good reason not only why your number is right, but why the market number is wrong? This requires a level of self-examination that is critical to success in this game. If your narrative is “I ran blah blah blah machine learning and I have the Dodgers at 90% to win so they’re a great bet at -150”, you’re going to have a tough time. It’s virtually impossible that the market is wrong by THAT much. The standard comeback is “if I’m wrong and the 90% is really 70%, I’m still good”, but that’s a slippery slope. The 90% is obviously wrong, what makes you think the 70% is right? If a model is wrong by this much, it’s much more likely to be completely wrong than partially wrong.

Some examples of “narratives” I’ve used in the past to convince myself that my numbers were right and the market was wrong:

• “The market is wrong because my numbers are taken from another, more efficient, market than the one I’m betting into.”
• “The market is wrong because I’m betting into a book that does not move lines on action and I know that their line makers are not sharp.”
• “The market is wrong because I can pinpoint a specific error they made in setting their lines.”
• “The market is wrong because analyzing this market involves a specific area in which I have expertise at a world-class level (applied probability distributions, epistemic uncertainty, etc) and so this is one of the rare cases in which I have more confidence in myself than in the market.” (Teaser for an article I’ll write in August about NFL regular season wins markets.)

In summary, “narrative” doesn’t always have to be a dirty word. Instead of using narratives to explain why you are right, use them to explain why the market is wrong. Be honest with yourself, and keep in mind that this type of thinking is quite vulnerable to cognitive biases. If you can’t think of a good reason why the market is wrong, it’s exceedingly likely that the market is not the one that is wrong. It’s a hard truth, but your bankroll will thank you!

Copyright in the contents of this blog are owned by Plus EV Sports Analytics Inc. and all related rights are reserved thereto.