Man v Model Pt. 2

Yesterday, I wrote about my Man v Model. The most interesting thing was running the results through Claude to see if it could identify the reasons why.

Here are some of the highlights:

1. Betting odds almost never recommends draws accurately. Of the 84 games that we played (I missed a few days early on), they only predicted a draw ONCE, whereas there were 24 draws in the tournament. The Model and I picked up a ton of those points. The lesson here is that there is some alpha in picking draws which seem to be underpriced.
2. The model failed to pick up the high scoring nature of the tournament. Whether it was heat causing games to open up, or other factors, but their average goals was 0.94 goal a game, whereas in reality the average game had a touch under three goals, including the 10 goal madness that was the 3rd place final. I should read Nate’s opinion on why he was consistently so conservative, with almost all picks being 1-0 or 0-0.
3. The Market and Model were WAY better at predicting the winner of the tournament than me. This is consistent with a lot of predictions; it is a lot easier to predict which stocks will go up over time, but extremely difficult to predict how the price will change tomorrow.