Forget About Accuracy Scores, Do This Instead
===

[00:00:00] Welcome back to Next Level Chess Podcast. I'm Grandmaster Noel Studer and I help chess players train deliberately with what I call the Simplified Chess Improvement System. Deliberate chess players live by three rules. Do what matters. Do it well. and do it consistently. If you're tired of training randomly and want to follow a simple proven system, this podcast is for you.

Nowadays, everything gets gamified and that sometimes is good, but it can also lead to a lot of features that have this fancy syndrome, right? They are fancy, they seem cool, but actually when we [00:01:00] boil it down to chess improvement, they don't help at all. And today I want to talk about one such feature that I believe is super unuseful for your chess improvement, but it's useful for chess companies, it's useful for making chess more fun, it's maybe bringing you into chess in the first place, but then once you say, okay, now I want to improve my chess, it really doesn't help anymore.

And this feature is called accuracy scores. I guess everyone knows what an accuracy score is, because if you play on chess.com or you play on Lichess, once you finish a game, basically immediately when you analyze them, they give you like this, Oh, here's your accuracy score. You played with this rating, with this accuracy in percent.

And then you see threads online of people posting, Hey, I had a 95 percent accuracy score or 80 percent or whatever. And I, as a chess coach, I'm no fan of this accuracy score because I have a lot of interactions with my students where they get super [00:02:00] confused. They get the wrong takeaways because of this accuracy score.

So let's get into this episode. 

I have this interaction a lot. A student tells me, Hey, I thought I played well, but I only got a 60 percent accuracy score. Now I'm sad about it. What should I improve next time? Then I go click through their game. And I have to tell them, well, you actually played a great game.

You took a lot of decisions I'm very happy with. You should be proud of yourself. But the accuracy score is low because of other factors. So accuracy scores can be fun, but they can give us a misleading sense of how well we've played. Often students play a good practical chess. I'm a lot emphasizing this.

Practical chess, we don't want to play computer chess, we play against humans. So practical chess is super important, but practical chess sometimes means not playing the top engines move, and that means a lower score. [00:03:00] Or, students can blunder early on, so you give away queen, a rook, a knight or whatever, you then fight on for 60, 70 moves, and these 60, 70 moves you play well.

You get a high score, so you get the wrong takeaways. When you get a high score, you actually should be upset that you blundered something and you only played well once it didn't matter anymore. Yeah. Congratulations. You played well, when you were a queen down. I don't think that really matters from a practical point of view. And on the other end, maybe your opponent gave you a big advantage, but then you didn't convert precisely according to the computer,

but you converted very well in a practical sense. So I would be very happy with it, but you see a low accuracy score, you're confused. So I'll try to get a real world example and an analogy that hopefully will help. We take investor A and investor B. So we go into the investing world.

Investor A [00:04:00] is making 99 percent perfect decisions out of a hundred. So very good, right? 99 percent accuracy, full accuracy. Amazing. They build up a lot of wealth. They just have one single problem. Every 100th decision is a massive mistake. They fall into a crypto scam or gamble all their money away in a casino on a single day.

So they build up all of this wealth, then one problem, but all the money's gone. And then you have investor B that gets a lot of stuff wrong and knows that they will get a lot of stuff wrong. And that's why they focus on what really matters. So from a hundred investments, 30 turn out to be failures. But knowing that, they never invest all of their money, but they split it accordingly and they can afford to lose the 30 percent of the time, because 70 percent of the time they win, and this will outperform their 30 percent losses.

So now the question is, do you want investor A with a 99 percent [00:05:00] accuracy or do you want investor B with a 70 percent accuracy? Because investor A is bankrupt every time after a hundred investments or decisions, you should definitely pick investor B. And this is probably the most quoted stuff by Warren Buffett.

So you might have heard this already, but he has two rules that he gives away for investment a lot, which is rule number one: never lose money. Rule number two, never forget rule number one. And the interpretation or how he seems to explain it is like, obviously you will lose money in a way when you invest, but you should never be reckless.

You should never overestimate. You should never put everything in one basket. And so my interpretation of it is you should avoid that huge mistakes. So the same happens in a chess game. Not every decision is equal. Getting most moves right doesn't matter if you royally screw up on one, two moves per game.

Mostly for us that's lack of focus, tactics that lead to these big mistakes. And thus the [00:06:00] accuracy score doesn't necessarily indicate good play. So that's very important. The higher accuracy you get, it doesn't mean that you should be happier with your game. So this is a little bit unintuitive and that's why I prefer to just not look at it.

But let's get deeper into what even a good accuracy score is. If you're googling online, you will see a lot of threads. You will see threads on chess.com, Lichess, Reddit, with people saying, hey, what is a good accuracy score for my level? I get this, I get that, I'm happy with this, I'm happy with that. But there is no real answer to this question.

That's the huge problem. You can get a 90 percent accuracy score. But I look at your game and I say, you shouldn't be happy with the game, or you can get 60%, but I, as a coach, can be very proud of you and happy with how you played. There is no strict correlation. Relying too much on this number to judge how well you played can be very misleading.

It often also leads to the [00:07:00] phenomena that people with a 90 percent accuracy score will just not even look at the game in their mind. It's like, I play perfectly. Let's go on to the next one. And maybe they are missing a huge, huge, huge leak in these games. And that's so important. It's just this score is driving us the wrong direction.

Remember, our goal or your goal should be to improve and not to chase a perfect score. You should understand what you want to do better next time and focus on what matters. Big mistakes in general. But don't only take it from me. Lichess has a page where they explain their accuracy score. I will link it in the podcast description and I will quote from it now twice because it's super important to realize that even Lichess is basically saying, Hey these aren't as great as people think they are, or they aren't as precise to, if you played well, you had a high accuracy.

So here is Lichess quote. "A very high accuracy [00:08:00] percentage isn't necessarily indicative of superhuman GM level play", end of quote. This is very important because there are a lot of cheating accusations against people that play with high accuracy. Also there, we have a problem because it doesn't mean that you cheated if you have high accuracy.

And then, this is even the much more important quote, it's a little bit longer. Again, quote. "While there is some correlation between the player's ratings and their accuracy, it is not straightforward. A more skilled player tends to play more principled theory, heavy openings, and put more practical pressure on the opponent.

This can create more complicated positions and provoke inaccurate play on both sides. Moreover, lower rated players are often more reluctant to resign as we discussed above. Protracted loop sided endgames can increase the accuracy score", end of quote. In other words, playing for a [00:09:00] hundred moves a queen down, that's the example I already used, you can play nearly perfect chess for a long period and thus get a high accuracy.

But the practical value of that is basically zero. You can also get better at chess and your accuracy drops, as we heard. Maybe play tougher opponents that put you in more difficult spots, you play more direct complicated openings that put you in more difficult positions and your accuracy score is dropping.

The accuracy score also depends on the opening, just mentioned that, you get on the board, how well your opponent plays, how long the game went, most of the things are out of your control. As there is no definite real correlation between your accuracy score and whether you can be happy with your game, ....

I recommend simply ignoring this score altogether. The risk is too high. Most people get confused by it more than it helps. So see it as a playful number, but not [00:10:00] something you take into account if you're saying, Oh, I'm improving or I'm not improving. It's not a score you should track. So what should you focus on instead?

Let's talk about what really matters. Three main things matter in every single game you play. Of course, there are more once you go more in depth. So the stronger you get, the more you can take away from a single game, the more things you can focus on. But on a lower level or on a adult improver level, I would put it this way.

Think about these three. Point number one is what matters is your process. How well did you focus during the game? Did you apply, for example, the blunder check on every move? How was your process when playing the game? An accuracy score cannot give you that at all. You need to decide that for yourself.

Did I focus well? Did I play when I scheduled to play? All of these things. So your process matters. That's point number one. Point number two is blunders and that's usually [00:11:00] double question mark if you talk about analysis done by chess.com or Lichess. Did you miss any big obvious things? Anything that shifted the engine evaluation drastically, when it's still mattered, this is super important.

That's why the double question mark can, maybe make an episode about this as well, is sometimes also misleading because who the heck cares if you're already minus five, and then you're going to minus 10. Again, that's not that important. Sure, you can say process, maybe you lost focus, that's something important to note.

But from a practical point, that doesn't matter as much. Also, if you have a totally winning position, right? So the computer is giving you plus 15. And then it's saying double question mark, you're only plus five. Now, maybe you on purpose decided to trade down instead of going to hunt for another piece or whatever, into what you know is a totally winning endgame.

This should be applauded as a [00:12:00] nice, smart, practical play and not scream that with double question marks for bad moves. So when looking for blunders, not only check for the double question mark, but also ask yourself, did it have a practical value that I missed this? Or did I maybe even on purpose, not even go for the line that is shown as the best line because it seemed complex and I didn't need complexity.

And then the third thing would be missed chances. Were there any big tactical oversights by your opponent that you didn't punish? So sometimes there is a difference between double question mark and missed chances, at least on chess.com, I believe that. But if you're looking for yourself, if you're going through the game for yourself, what I teach my students is, hey, just switch on the engine.

Look, is there somewhere a drop of, let's say, roughly two score, like you go from plus one to minus one, for example. So that's a two drop somewhere around there. It's [00:13:00] very hard. Every position is unique, but one and a half to two points is to start where it's really a big mistake. And usually it has something to do with some concrete things that you missed.

That means that's a tactical oversight probably. And then you want to understand why did I miss this? How can I do that better next time. That's the three things that matter, right? Your process, blunders, and missed chances. So you might wonder, well, but then why do accuracy scores exist? And I led this podcast a little bit with why I think they exist, because gamification is a great tool for making chess engaging and accessible, but it's important to remember that not all of it serves the purpose of improving your game. So whenever there is a tool that is coming out from, especially from a big company, it doesn't mean that this tool is going to help you achieve what you want. But [00:14:00] most likely it helps them achieve what they want. So more players to the game is better for chess.com, for example. So when commenting on a game, it's great for a commentator to say: Oh, wow, the players are basically playing perfect chess, 99. 8 percent accuracy. That's a great way to hype the game up. So that's why this score exists. You see that in other sports as well, for example, in football, soccer, they have the expected goal score.

I think that didn't exist until a few years ago. And I highly doubt that a coach will say, Hey guys, we only had 0.7 expected goals, or we need to do better. No, he's like, okay. You should do this better. You should do that better. What was the process? How can we improve our game, but not based on this score.

And then when you see that commentary and you play on your own, you get a game with 90 percent accuracy, 95 percent accuracy. You feel on top of the world. You're like, Hey. So now I just saw the world [00:15:00] championship match. They had 93 percent and I just had 95%. Hey, I just played better than the world championship, right?

So this is kind of the comparison that it leads to you. But as you know, sadly, no, that doesn't mean that it's true. It's just based on your game. And if somebody else would have played your game, probably they would have had even better accuracy. So I hope this makes sense. This accuracy score is mostly meant to bring more people into chess, to make it easier to follow chess games and to make it more hype and to give you a good feeling.

What does it mean? A good feeling means that you're going to play more and the more gamified chess is, the more you will feel like, oh yeah, I want to do more of chess, which is good for all the players within the chess ecosystem, if we're talking about economics. So I'm not saying websites shouldn't have this, but I'm saying if you really care about [00:16:00] improving, if you want to be a deliberate chess player, deliberate chess improver,

you should ignore this score, because, okay, this is gamification, but you need something else. As a small reminder, you need to watch your process when you play games, you need to watch for blunders, and you need to watch for missed chances. And that's basically it, especially if you're only starting out to analyze your games on your own, these three things will be enough, you'll learn something from your game, and then, that will, down the line, improve your chess.

Hey guys, just two quick things before you take off. If you enjoyed this episode and want more structured chess improvement tips from myself, check out my newsletter at nextlevelchess.com/newsletter. It's totally free. It will always remain free and it goes out every single Friday with the best [00:17:00] latest chess improvement tips that I have.

Most of the podcast episodes that I record are based on a previous newsletter. So, getting the newsletter, you'll get the advice earlier and you'll get it directly into your inbox every single Friday. It's totally free, as I mentioned, and you can unsubscribe any time. So, go to nextlevelchess.com/newsletter to sign up. 

And one last thing, if you enjoyed this episode and if it helped you, then please take a few seconds and review this podcast. This helps a ton. It helps other people see, oh yeah, many, many people profit from the advice given in this podcast. Let's give this podcast a try. And if you can, if you know anyone in the chess world that would profit from this episode or any other episode, make sure to share it with your friends, with your people online.

That's [00:18:00] super helpful. Podcast growth is really just working through mouth by mouth recommendations. So thank you. Thank you so much for listening and thank you for spreading the word about the Next Level Chess Podcast. Now that's all from me. Thank you for listening and see you next time.