TRUTH About Chess Accuracy

[@JohnDoe-yg5sm]

Such a great video! The way you hold yourself is undoubtedly the best for chess videos, I hope to see more of you on this platform, take care.

[@kesetokaiba]

Thank you for the kind words :)

[@crazyape968]

1:25 Here KeSetoKaiba says "'E', 'L', 'O' band". Just a reminder to beginners that "Elo" is not an acronym. It does not stand for anything but is instead the last name of the person who invented the system: Hungarian-Amerian physics professor, Arpad Elo. en.wikipedia.org/wiki/Elo_rating_system If his last name was "Jones", you wouldn't call it a "'J', 'O', 'N', 'E', 'S' band"!

[@kesetokaiba]

Good point lol

[@Geelse_zot]

Good video to adress something that really needed to be discussed. I have seen lower rated players score 90% accuracy and think they played at GM level. It's all about how much pressure your opponent has against you. Magnus vs MVL at 80% accuracy is far better chess than Geelse_zot vs John Doe at 80% Titled players just put so much pressure on the position. They force me to play bad moves.

[@kesetokaiba]

Thanks you. It was a needed video. I just had a 600 rated friend of mine show me a game where they had close to 100 percent accuracy and wanted to know if this was normal. It was. The opponent blundered on move 3.

[@jaybingham3711]

The analysis includes an estimated rating level which takes into consideration the player's current rating. Not sure how anyone presumes they're at GM level when the estimated strength is staring them right in the face.

[@vasilyborodin]

Б и .ю. ю

[@johnwalker1058]

@@jaybingham3711 I'm guessing it's some sort of cognitive bias where people tend to want to overestimate their skill at something.

[@cydonyn]

In a bullet game my opponent hung their bishop and I missed it for like 3 moves, after I took it they resigned, but our caps score was in the 20s and 30s lol. It was kind of funny.

[@rottenfifa5391]

Game difficulty, that's the point, yes. I once Fried Livered someone and got 100% accuracy. One day, my friend was going through my game archive and suddenly shouted "You Stockfish" seeing the accuracy of that game. I mean, would you care to look into the game instead of going for the accuracy? It's a freaking Fried Liver!

[@kesetokaiba]

lol I like seeing all of chess :)

[@dragonore2009]

Yeah, I mean even low rated players, such as myself, get 90+% accuracy from time to time. So yes, it doesn't determine cheating.

[@kesetokaiba]

Congrats, still :) 90%+ accuracy is great at any rating.

[@FloydMaxwell]

In fairness to Vladimir Kramnik's work, he is looking at large numbers of games for a given player, also the average rating of opponents and above all simply comparing players to Magnus/the best players. When unknown players are getting multiple CAPS scores like (or even above) the top player's score, it is safe to assume they are at least suspect.

[@rawr333r]

I always assumed the accuracy was per opponent. I play 90% accuracy against 500 rated players and 70% accuracy against 1100 players.

[@kesetokaiba]

Yeah, it isn't exaactly opponent-based, but higher rated opponents are more likely to find good moves and this makes it tougher for you to find good moves, so indirectly it is sort of true.

[@SenatorBluto]

Accuracy has nothing to do with ratings.

[@rawr333r]

@@SenatorBluto lower rated opponents play worse and therefore I can find the best moves against them and have a higher accuracy more often than a higher rated opponent.

[@SenatorBluto]

@@rawr333r Of course. But the computer evaluation doesn't care who you're playing. To play a GM a long hard fought 80 move game with 95% accuracy, now THAT'S impressive.

[@kesetokaiba]

@@SenatorBluto Individual games, then probably not much. Averaging out accuracy over a large sample size, then rating is a big factor. Of course it just makes sense that higher rated chess players have a better chance of finding the top engine moves in their games.

[@tripolarmisfit]

I trust online chess one million times over online poker.

[@kesetokaiba]

Fair point. Much of online poker is trustworthy too, but there's also a ton of poker sites which are unreliable. WAY more bots in online poker than chess percentage-wise. I'm virtually certain of that.

[@justinkauffman731]

Poker is over seen by the states of MI, NJ and PA for me. Comfortably grinding at either Poker Stars or MGM daily

[@kesetokaiba]

@@justinkauffman731 I seriously considered doing that as well, but instead I have been putting more time into my KZbin channel for several reasons.

[@jaylenlenear3944]

I thought it was more or less understood that blunders/opening prep often leads to high accuracy and that doesn't make it wrong but going off one game is wrong. accuracy based off a bunch of games is way more just.

[@kesetokaiba]

I'm sometimes surprised by common sense not being that common, but at least this one I can understand how CAPS could be misleading to people unfamiliar with it.

[@shadeburst]

I place little reliance on the accuracy and game rating stats. In order to improve I'm more interested in the numbers of brilliants (open the champagne), greats, bests, mistakes, misses and blunders, especially the last three! An opponent's game rating far above their ELO, and a consistent pattern of long pauses between moves; those are red flags to me. But there's nothing that can be done and I remind myself that chess is like golf. Every shot pleases somebody.

[@kesetokaiba]

I actually enjoy playing golf a lot :)

[@nickkestler4848]

Very informative! I(1200) played a 1600 and beat him with a 92% accuracy and I thought I was going to get reported and banned haha.

[@kesetokaiba]

Congrats on your win! If you play fairly, then you never have anything to worry about; glad this video helped you feel better about it.

[@AyrtonTwigg]

I find it hard to believe that their cheat detection is the very best, when there was an account that was making every move in 0.2 seconds, and it didn't get banned for months!

[@kesetokaiba]

Sounds like someone experienced in bullet chess. If every move was that fast, it would be difficult to cheat since looking at an engine would take longer than that.

[@AyrtonTwigg]

@@kesetokaiba There are chess bots that can play moves instantly. No human can play a whole game in 0.2 seconds per move.

[@kesetokaiba]

@@AyrtonTwigg No chess.com moves happen instantly. Even premoves take 0.1 seconds. There are bullet players who move really quickly even without premoves. My reaction time for most moves is about 0.3 or 0.4 seconds, but it could be faster if they "hover" the pieces before moving, or especially if they premove, but if literally ever move was 0.2 seconds, then that would be suspicious.

[@Ebobster]

I wonder if anyone has correlated average accuracy of a large data set of games with likely ELO rating. Or existing ELO ratings with anticipated average accuracy.

[@kesetokaiba]

I think I recall chess.com doing this in the Insights feature in the same way that they display your rating to the global percentile. I'm sure with some effort, people could compile a list of various accuracies and ratings.

[@PHeMoX]

Many of these statistics are a bit useless actually. For example, is a move really a brilliant move when the follow-up of the game does _not_ match the line that takes complete advantage of the brilliant move? In other words, someone played the brilliant move accidentally? This stuff happens all the freaking time and it shouldn't count as a brilliant move. More so as 'it had the potential to be a brilliant move', with only those moves following it up in the correct way truly being brilliant moves. The chess statistics are filled with problematic statistics like this one, unfortunately.

[@kesetokaiba]

True. I've also seen really Brilliant ideas which engine only calls "best."

[@futureanimator522]

Hey kaiba How are u doing? Haven't talked to u and ur brother in a while.. glad ur channel doing good Its me jay to Jay to jay

[@kesetokaiba]

Hey, friend :) Yeah, we should chat sometime. I have been having some internet connection issues for the last several weeks (which also made recording videos tougher), but I'll try to message if I see you online :)

[@futureanimator522]

@@kesetokaiba tell your brother I said wassup

[@kesetokaiba]

@@futureanimator522 I will :)

[@GlobalWarmingSkeptic]

Accuracy also needs to be compared to the opening and openness of the game. In a game with a lot of sharp forcing lines, accuracy will be very high. In a game where there are more complications, even strong players will have accuracy dip to the 80s and for some openings and variations, even the 70s.

[@kesetokaiba]

chess.com Insights in Stats does also reveal accuracy sorted by opening :)

[@Jezreel_Barriga]

I'll summon the Exodia... its over now kaiba -Yugi

[@kesetokaiba]

"That's impossible! No, wait. What am I saying? It's totally possible." - Kaiba Yugioh Abridged series

[@chadbishop-sr]

How do I tag Kramnik on youtube?

[@kesetokaiba]

lol

[@theclimbto1]

LMAO! Okay, When I Win 67.8%. When I lose 64.3%. Not much difference in my Accuracy when I Win or Lose. But when I DRAW? 72.2%. I seem, somehow, to be rather accurate when I Draw.

[@kesetokaiba]

Could also be that many draws are played out to the end in positions where the game is pretty much decided like opposite-color bishops shuffling. Playing this to 50-move draw rule likely gives 50 moves to boost the accuracy as you both won't blunder. This inflates the accuracy.

[@IvanBaAl961]

Accuracy depends on playstyle. So even average value doesn't tell that much per se. Lichess shows you quite a lot of various stats, although you don't have simple access to other players. I doubt their cheat detection is worse. I barely meet any suspicious players there.

[@kesetokaiba]

Part of the reason could also be how small lichess is. It only has about 150,000 members compared to chess.com's over 100 million. Less people means less cheaters, but not necessarily due to cheat detection; it is size number of people.

[@IvanBaAl961]

@@kesetokaiba Well, Lichess is a little more populated. It's over 100000 online players right now.

[@noobnoob98noob]

Saw this point in comments and so i am going to summarize the idea. Accuracy is kind of a representation on how close you are to your opponent in skill in a specific game. It is also a how close you play to a computer, which is the closest thing we have to know what perfect chess is. Have a nice day!

[@PHeMoX]

It is NOT a metric that shows 'how well you actually did that game' though. Like, not at all. It is moreso a percentage score that shows how closely you followed theory / book moves and made the best moves _as a response_ . Because in reality you can have a 80-85% score in a deadlost game. It also overvalues the opportunity for a draw. The percentage score does not show let's say your probability towards a win. The percentage is _obviously_ heavily influenced by a bunch of best moves, but you only need one inaccuracy or one blunder that gets punished hard to end up in a totally losing game. As such, these percentages actually say nothing about how well you did. The actual moves, ie. did you blunder or not blunder are way more important here. The elo system itself judging only wins, draws and losses are waaaaaaaaaay better to show 'how well' someone does. It is also really a problem that many best moves are kind of 'easy' to find, versus some other opportunities are basically impossible to find for average to higher range players that are not GMs or super GMs. By the way, I very much doubt the accuracy rating takes into account a player's or an opponent's elo rating. It should 100% ignore it.

[@kesetokaiba]

System isn't perfect, but I don't even know if this really a "percentage." Yup, it is out of 100, but I don't know if best move 10% of the time would actually give 10.0 CAPS. I think they skew the results to be higher than a certain number (30 perhaps?, which exception for zero Danny emoji).

[@asexualnonbinaryoterkinfem6236]

The Elo Gate Keepers oppresed ya accuracy

[@theclimbto1]

My FAVORITE thing when I do my Post-Game is when the Computer tells me "That was a bad move". Okay, then my next move... set up by that move... "BRILLIANT!". Yeah, that's why I didn't take that pawn last move guy, was to set up this move.

[@johnwalker1058]

Same. I've had a couple instances where the computer looked at a particular move and was like "meh," but then the next move was deemed "great" or "brilliant" but it was only possible because of the previous move that set it up.

[@aaronb2181]

​@@johnwalker1058That happens because there was a refutation to your move that would put you in a worse position, but you're opponent couldn't find it. Do not think you're a genius and the computer is wrong.

[@johnwalker1058]

@@aaronb2181 If there was a refutation to a move that would result in a worse position, then the computer would say it's a bad move, not just merely a suboptimal move (what I meant by "meh"). And the eval bar would drop down significantly instead of staying at roughly the same level. Meanwhile, if the opponent's response to the setup move is identified as a mistake or blunder by the computer, then it might be true that the setup move was refutable, but the opponent failed to find the refutation (hence the mistake or blunder label on their move for not finding it). However, in instances where this happened, the opponent's very next move does not refute my setup move, but was not labeled by the computer as a mistake or blunder, and then the following move was labeled as "great" or "brilliant." *What I was getting at is that it's rather strange for the result of a good play to be praised more than the move that made the whole thing possible in the first place.* It would be like finding a move that sets up an unstoppable mating sequence, but that move is not credited as that good of a move. But then the final move in the sequence that delivers checkmate is considered very clever, when it's just the finishing result of that earlier setup move that made it possible in the first place.

[@Jlang-es9lc]

This screams 700 elo

[@theclimbto1]

@@aaronb2181 I'm pretty sure the Computer gets things wrong. The reason I say that it from time to time it will grade one of my games as a 2000+, and I'm nowhere near that. And I don't mean like "I'm 1800, and that's a huge gap to 2000", I bounce back and forth in the 700 to 950 Range. I suck. All I'm saying is, sometimes the Post Game Engine does some wonky things.

[@ChimeMemeh]

tbh im 2200 rapid and most my games are 85+ accuracy cause i sweat in those games

[@kesetokaiba]

Hey, that counts lol. I'm still trying to get to 2100 rapid myself.

[@josefkaras7519]

really good explanaition. i have a feeling that even kramnik might be able to understand it if he saw this video...

[@ТимофейЧерников-щ2х]

He obviously knows that already

[@kesetokaiba]

lol I hope so.

[@kesetokaiba]

@@ТимофейЧерников-щ2х Joking aside, Kramnik tried an aggregate of CAPS when accusing Hikaru of cheating in his famous blitz streak, but as staticians have shown: Hikaru's streak was impressive, but also essentially inevitable to happen eventually for someone of his level.

[@Cer_Pro]

You can become GrandMaster, are you trying for it?

[@kesetokaiba]

I don't have dreams of becoming a titled player, but even NM. FM, or CM someday would be cool. It isn't a current goal of mine though.

[@jaybingham3711]

6:38 Wth? This is decidedly sub-1000 play. Did this opponent suffer a stroke during the game?

[@kesetokaiba]

lol they didn't really have good options. That was the only legal move which doesn't walk into Ng3 with check.

[@firesickle]

I feel like as long as I win I got 100% accuracy, I dont care if Im 300 elo

[@kesetokaiba]

xD Reminds me of a coffee mug that says "It doesn't matter if you win...it matters if I win."

[@firesickle]

@@kesetokaiba hahahaha.... exactly!

[@VivPlayzX777]

​@@kesetokaibaI'm 826 is it bad that I get 90 plus accuracy often

[@kesetokaiba]

@@VivPlayzX777 Nope. That's great. It probably means your opponent's are blundering early and you are correctly exploiting that.

[@chill-warlords]

What if you it's a complex middle game

[@Nostalgicallyy]

This illegally feels llike a GothamChess video

[@kesetokaiba]

lol that's a compliment because of how much GothamChess does for the chess community, but I wasn't imitating him at all with this video. I think it is partly because he covered the cheating scandals that it feels similar.