# Why is Magnus Carlsen's Elo rating not higher, given that his win rate is so high?

Magnus Carlsen's Elo rating is about 2870, and according to this quora answer, he hasn't lost a game in 100 games at the time of writing. Assuming that his win rate hasn't changed since then, and assuming this implies conservatively a win rate of 99%, then based on this table, which says that a 99% win rate against a single other player corresponds to about 670 ELO surplus, this seems to suggest that Magnus Carlsen plays players who have on average an Elo rating of 2200. Is this correct? it seems like it is way too low, given that my understanding is that the super grandmasters mostly play each other and other high-scoring grandmasters, at least in long games.

• No - Magnus Carlsen draws most of his games, so his win rate is closer to 50% than 100%. (Most top level games are draws.) Commented Nov 11, 2021 at 19:07
• Re "he hasn't lost a game in 100 games at the time of writing": How do you count that? Magnus lost in the classical time format on 2021-09-12 to Sergey Karjakin (current FIDE rating 2743) (Norway Chess 2021). Doesn't that count? Has he played 100 classical games since then? Commented Nov 12, 2021 at 6:09
• "Assuming that his win rate hasn't changed since then, and assuming this implies conservatively a win rate of 99%" That's not a conservative estimate, that's a wild, overblown estimate Commented Nov 12, 2021 at 11:29
• @David, the others have already clarified that Commented Nov 12, 2021 at 11:55
• Related, 8 years ago: Why has Carlsen's Elo not changed? Commented Nov 12, 2021 at 16:34

A 99% win rate is winning 99% of games played. Carlsen does not win nearly that many games, since at top level most games are drawn.

Carlsen's overall record in life is +1262 =1289 -454, which means 1262 wins, 1289 draws, 454 losses. He has won 42.00% of all the games he's been recorded to have played, and has an all-time average score of 63.44% (wins are 1 point, draws are 0.5 points, losses are 0 points).

Another thing is the amount of rating gained or lost depends on how strong your opponent is. If Carlsen were to play a beginner a hundred times, he'd win effortlessly for a 100% win rate - but he'd also gain almost no rating.

• @Allure - even worse, ELO cutoff! So if the difference is <400 (AFAIK), he even loses a bit with 100%! Commented Nov 12, 2021 at 8:46
• If Carlsen were to play a beginner a hundred times, he'd win effortlessly for a 100% win rate - but he'd also gain almost no rating. 1000% this. Also, if (unthinkably) he were to lose (or even draw) one of those games against that beginner, he would lose a massive amount of points, even with an overall 99-99.5% win rate against that player. Commented Nov 12, 2021 at 19:34
• @HaukeReddmann "So if the difference is <400 (AFAIK), he even loses a bit with 100%!" - No. The 400 point cutoff means he gains a small amount instead of zero (although with the newer rules he only gets the benefit once per tournament.) If you have an established rating you never lose points for winning.
– D M
Commented Jun 11, 2023 at 1:30
• @DM: THX, then I always misunderstood the cutoff rule. Commented Jun 11, 2023 at 12:37

assuming this implies conservatively a win rate of 99%

There are a number of wrong assumptions in your question. First, at elite levels where players don't often lose it is draw rates that are high. Second, according to the FIDE rating rules if the rating difference were 670 it would be adjusted down to 400:

8.54

A difference in rating of more than 400 points shall be counted for rating purposes as though it were a difference of 400 points.

In fact over the last two years according to the data on FIDE's rating website for Carlsen he has won 40.7%, drawn 54.3% and lost 5% of his games.

Here are his results from November 2019 to date.

A "*" next to the average rating shows that some or all of his opponents were rated more than 400 points lower than him and their ratings were adjusted up to give him more rating points.

Date Results Rating Change Avg. opponents' ratings Event
October 2021 +4=5-1 -0.2 2741 Norway Chess 2021
October 2021 +2=1-0 +0.3 2557* 36th European Chess Club Cup 2021
September 2021 +8=6-0 +8.4 2672 World Cup 2021
March 2021 +3=9-1 -15.3 2714 Tata Steel Masters 2021
November 2020 +5=3-2 -0.8 2743 Altibox Norway Chess 2020
April 2020 +1=0-0 +0.8 2462* Ostlandsserien 2. div A 2019-2020 R 9-11
February 2020 +3=10-0 -9.6 2731 Tata Steel Masters 2020
January 2020 +1=3-0 -0.2 2778 London Chess Classic Grand Chess Tour Finals 2019 Standard Games
December 2019 +2=0-0 +1.6 2470* Ostlandsserien 2. div A 2019-2020 Round 1-2
November 2019 +4=7-0 -6.0 2692 2019 FIDE Chess.com Grand Swiss

Totals
+33=44-4

Percentages
+40.7=54.3-5

• Surprisingly (for people who aren't too deep into chess) Carlsen lost rating on some tournaments without losing a single match (e.g. Nov 2019). This is because against lower rated opponents he is expected to win more matches instead of drawing them. So if he draws too much and wins too few he loses rating without losing matches. Commented Nov 12, 2021 at 11:40