Quote:
Originally Posted by coon74
It's bizarre that the rating analysis gives Levon only a 17% chance of winning the event; seeing his current tactical brilliancies, I think it's at least 30%.
I agree it seems a little low at first, but of course the model doesn't adjust for form or his "current tactical brilliancies" in any way beyond using his improved live rating. He's still 2776, while Carlsen is 2854, and they're tied for first with four rounds left (plenty of time for the rating edge to matter). Back of the napkin sanity check might say that if Aronian's odds of winning are X, Carlsen's should be at least 2X. Then there's Topalov just half a point back, also with a big rating edge over Aronian, and with a pretty easy remaining schedule as I've said before. Makes sense that those factors could overcome the rating edge and give him roughly X odds as well. Right there that gives Aronian a cap at 25%, and that's if you thought noone else had any chance at all which is obviously silly because with four rounds left all sorts of crazy stuff can still happen. Give some equity to the long shots, and Aronian drops to maybe 22%.
Then we run the numbers and see that the model gives Carlsen more than a 2:1 edge, and Aronian shakes out at 17% instead. Reasonable enough if you think 2776 is a good estimate for him. If you wanted to give him credit for being in better form than that his odds would increase. I reran the sim with everything the same except Aronian, whose rating I bumped to 2820, and those 44 extra points improved his odds of winning to 27%, right around your estimate of 30. I can get him all the way up to 30% by putting his rating at 2832, just two points above his published peak.
So you're not really questioning the model, you're just saying that it feels like Aronian is genuinely playing at 2830 strength in this event (not just getting lucky so far) and that you expect him to continue doing so over the last four rounds. Not particularly unreasonable