Quote:
Originally Posted by NajdorfDefense
BJJ, I agree with Tim that match results are not 'independently' occuring games. And Nepo is a bad stylistic matchup for him. Let's not forget Magnus is only +1 in classical time in his last two defenses. And, certainly, 'luck' can always play a role in these things. All it takes is one of your prep lines to have a mistake that leads to a loss and you're -1 and having to rebuild on the fly under the greatest pressure.
Thinking more about this point, I guess I should explicitly clarify that I also agree the games are not independently occurring in a true statistical sense. The thing is, I don't know in what way the games being related to each other actually affects the odds. I can easily run probabilities as if the games were independent. Once I've done so, I don't have an adjustment I can make that I think makes those probabilities more accurate. I don't have a better way to model things that feels like an improvement.
If Nepo wins the first game that puts pressure on Magnus, but it also gives him more incentive to play for a win when he's already the stronger player. He could crack under the pressure or he could be motivated by it. And meanwhile if Nepo is suddenly leading the match that puts a kind of pressure on him too. Does he become risk averse, to try to make sure he doesn't give up the lead? Does that work or does it backfire?
I will have my baseline odds for game 2 (once colors are assigned) from treating as if it's independent. I agree circumstances will change after game 1 is complete, and those odds may shift in some way, but I'm not sure how. And I don't have any data that tells me how. So I'm just not convinced the lack of independence between games really creates a problem for my model.