Apr 7, 2023·edited Apr 7, 2023Liked by Mike Saint-Antoine
This is great Mike! Thank you for putting in the data collection work to produce something interesting! For what it's worth, I usually bias pretty heavily toward prediction markets over 538. In the Good Judgment Open question below, tracking the implied probabilities of the betting markets fairly closely over the season gave me a brier score of -0.043 while a fellow forecaster (who is also a very good) who tracked the 538 numbers came back with a brier score of almost 0.25. As you can gather from our relative brier scores, the crowd was also closer to the betting markets than to 538 for this question. A lot of times, in other sports questions, it seems like 538 has been fairly close to prediction markets, but when they've disagreed, as per your intuition regarding EMH (which I'm convinced is at least weakly true), I've deferred to the betting markets. It will be great to see what you come up with and whether your initial hypothesis ultimately prevails. I suspect it will, but it will be great to see the empirical validation (or disconfirmation).
Thanks Jonathan! Yeah it seems to me that the betting markets should be at least as good and probably better than 538 in theory (EMH -- if it was the reverse and 538 was consistently better, we could get rich by just placing bets on the markets according to 538 predictions until the prices adjusted). But like you said, it'll be interesting to see if this is supported by the data.
This is great Mike! Thank you for putting in the data collection work to produce something interesting! For what it's worth, I usually bias pretty heavily toward prediction markets over 538. In the Good Judgment Open question below, tracking the implied probabilities of the betting markets fairly closely over the season gave me a brier score of -0.043 while a fellow forecaster (who is also a very good) who tracked the 538 numbers came back with a brier score of almost 0.25. As you can gather from our relative brier scores, the crowd was also closer to the betting markets than to 538 for this question. A lot of times, in other sports questions, it seems like 538 has been fairly close to prediction markets, but when they've disagreed, as per your intuition regarding EMH (which I'm convinced is at least weakly true), I've deferred to the betting markets. It will be great to see what you come up with and whether your initial hypothesis ultimately prevails. I suspect it will, but it will be great to see the empirical validation (or disconfirmation).
Here's the link: https://www.gjopen.com/leaderboards/questions/2179-which-team-will-win-the-2022-nba-finals?focused_on_id=87897
Thanks Jonathan! Yeah it seems to me that the betting markets should be at least as good and probably better than 538 in theory (EMH -- if it was the reverse and 538 was consistently better, we could get rich by just placing bets on the markets according to 538 predictions until the prices adjusted). But like you said, it'll be interesting to see if this is supported by the data.
And I'll check out the link, sounds interesting!
Very interesting!
Thanks! 🙂