I tried analyzing the past ten years of FIDE rankings by using Monte Carlo sampling to simulate future ELO trajectories and got the following results:

Magnus highest rated in July 2025: 54.6%

Magnus stays #1 throughout 2025: 23.4%

Magnus drops below 2800 before 2026: 35.2%

Someone breaks Magnus's peak by 2026: 19.4%

But then you actually look at the Elo trajectories of the current top-ranked players and it looks like this. Or you can compare him to the other top players in, say, 2015 and note that while other players may "fall off", Magnus Carlsen does not "fall off", but consistently cruises at an altitude that stratifies chess rankings into into "Magnus Carlsen" and "everyone else". So maybe something's wrong with my code, or maybe puny human statistics don't work on entities of this caliber. I think it's worth at least, like, whispering in the direction of doubt, though. If only to reassure myself that the man is mortal and bound by the laws of physics and etc.

@cherrvak I'm not sure how your code works, but I think it's <1% for anyone to break Magnus's peak rating that soon.

@TimothyJohnson5c16 there's absolutely no way it's less than 1%, but 19.4% seems a tad high. I'd probably estimate like 5-8% or something. There are 2 years to go.

For reference, Magnus went from 2772 to 2826 (a 54 point jump) in < 1 year, I think it's quite possible (and in the 5% range )that one of Gukesh/Erigaisi/Abdusattorov/Firouzja gains 100 elo in 2 years. Obviously very unlikely but still possible!

@benshindel Caruana went from ~2750 to ~2850 back in like 2014, in about 14 months btw!!! So it's quite possible for this to happen.

@cherrvak also looking at this, Magnus has been above 2800 for a decade so I don't see why the odds of him dropping below 2800 in the next year could possibly be 35%?

There was in fact an issue: my simulation updated the Elo used to calculate win probability with every game, which is only realistic if you predict that Magnus Carlsen’s reaction to losing is to spiral into a depression that makes him more likely to lose future games. If we instead run the simulation with a static value that we take to represent “true” chess ability, we get the following probabilities:

Magnus highest rated in July 2025: 66.5%

Magnus stays #1 throughout 2025: 31.3%

Magnus drops below 2800 before 2026: 20.5%

Someone breaks Magnus's peak by 2026: 13.6%

But the above also assumes that Magnus Carlsen plays 5 games per month, which is a generous estimate. At 3 games per month:

Magnus highest rated in July 2025: 75.4%

Magnus stays #1 throughout 2025: 43.9%

Magnus drops below 2800 before 2026: 12.8%

Under this paradigm Carlsen's mean projected Elo is about as good as it is right now, maybe a little better, so if you expect it to pop up again then don't listen to me lol. Magnus probably wins more often than not but there's still a chance, Gukesh 2025 lfg!!!!!!

@creator @mods This is still ambiguous.
From https://qc.fide.com/faqs/ "FIDE publishes its new rating list on the last day of every month, to come into use from the first day of the new month."

So at the close time, the July rating list will have been published (assuming no delays), but the June rating list will still be the effective list. So will this resolve based on the June list, or the July list?

FAQs – Qualification Commission

@DanielTilkin interesting. The goal fo the market is to discuss Magnus's rating through the close date. If the "come into use" on July 1st "in effect" number refers to the period prior to that time (which it seems that it must do) then it seems to me that that number is the last one validly related to this claim.

i.e. the claim is about magnus's rating through end june 2024. On june 30 a number is publicly released relating to his rating during the period. The number as released is noted to come "into effect" on july 1st. Yet it clearly refers to the period we care about. So, that number counts for this market.

Market close time seems wrong? When precisely is "mid 2025"

@MingweiSamuel thanks fixed.

@Ernie Still not clear to me what exactly counts at 'mid 2025' here: Is this 30 June 2025 (which would mean this would resolve based on the June rating list) or 1 July 2025 (which would mean this resolves based on the July rating list)?

or july 2, which is the 183rd day of the year (365/2=182.5)

I use the quarter system. Sorry for any confusion caused by this