Why Engine Accuracy isn't as useful as it seems (at least for sub-2000 players)

1. The Appeal

Everyone knows how appealing it is to click that "request computer analysis" right after each game finished, for the same reason many videos on YouTube, or chess books, or streamers, or any other thing that promises your Elo to skyrocket quickly. All of these things promise that if you just watch their video, or buy their book, it'll be the final piece of the puzzle and you'll gain 100, 200, 300, 500, however much rating they say you'll gain and it'll be fast and easy.
The same appeal happens with the engine analysis. Everybody probably hears "Analyze first, then check with engine" but what percentage of people actually do this? Over 95% of the people at my current level either don't analyze at all, or if they do, when I go to analyze I can immediately see that loading symbol and that engine analysis popping up. Why do we always do this? For the exact same reason as we are automatically driven to those videos or books with the bright, eye-catching titles and thumbnails. It makes our "analysis" quicker. Why would we want to spend the time to analyze the game ourselves when the computer can go through it for us and tell us what we need to do better? It is another promise saying that if we just click on this button, our rating will go straight up to the right.

2. Why It Isn't Actually Useful (especially for sub-2000s)

There are multiple reasons why Engine Analysis is not useful for people under the 2000 level. The first one being how terrible it is for your confidence. I've gained over 200 rating in rapid in the past month and a half, and it isn't because I've done any more studying than I was doing before, or playing more often. My confidence has been so much better recently. Some of the worst mistakes I've ever made in OTB tournaments were because I was too afraid of messing up to actually play the move I was considering. My dad, who recently started playing chess and is around 500 on chess.com, has told me that his biggest problem with playing games is that he's too scared of blundering a piece to play a logical move and instead will play too passively. This has nothing to do with chess knowledge or skill, but all about mindset. One of the biggest walls I hit was in June this year around 1380 rating and then I dropped 120 points in a couple of weeks. I tilted so miserably because once I lost a few games at the start I was convinced that I was a bad player and that every move I was going to play was going to be a mistake. I started playing less frequently and sure enough I had a better mindset going into each game and I gained that rating back fairly quickly and more.

What does this have to do with engines? Engines (especially the lichess stockfish) are the harshest critics of your games there are. Why is this? They are just simply better at chess than humans, which has been a true fact for years now, so every slight error completely ruins a game in their opinion because they know how to capitalize even though even a top GM might not be able to find why the move works.

A great example of this lack of confidence from me personally was a OTB game I played in October (I wish I could find the imported notation but unfortunately I was unable to) where I was outplaying my opponent tremendously, they hadn't even made a true mistake but I was in a position that the computer analyzed as +2.5 after 15 moves. I went in to attack, and my opponent defended well and got me into a difficult position where my only way to keep my advantage was to sacrifice my queen. I saw the correct move and knew that I needed to play it, but I was too worried that I had miscalculated that I didn't play it and lost the game soon afterwards. This had nothing to do with my chess knowledge (as I saw the tactic), but the constant visual of those double question marks had made me to afraid of playing a mistake to the point to where the move I actually played got the ?? when I analyzed later and I was really disappointed with myself for not playing the move I knew worked.

This is why engines (especially the lichess engine again) are terrible for confidence because the main thing that they do is highlight the mistakes made in any given game and give no attention to the good moves because being a 3700 elo engine makes those moves expected to it. Granted, the chess.com game review does have the great move and brilliant move possibilities, but those are still much less common (at least sub-2000) than mistakes or blunders. Seeing that everything you do is wrong in a game is terrible for confidence. This is why I believe that having a coach (if you can afford it) is a much better way to analyze your games because they are humans that make occasional mistakes also and are able to highlight your bright spots in games too.

Another reason it is not useful is because the total accuracy percentage doesn't really apply to humans. For example, I have two games I want to show, the first is a rapid game that I was extremely proud of playing, and a classical game that I didn't feel like anything special happened. (For whatever reason no matter what I do the board is in the wrong direction for the first game. I was playing as black, if you would like to see it from my point of view, just click the three dots under the game board and click flip board)

I'm sure everyone over 2000 would think that was an incredibly sloppy inaccurate game, but I was proud of myself for beating an opponent that didn't make any clear ("clear" is relative) easy to spot mistakes. I just simply outmaneuvered my opponent and got clear positional advantages. While the computer doesn't say this game was terrible, even though I had 89% accuracy which is pretty good, I still had 3 mistakes and an inaccuracy which when I was reviewing made me doubt whether I had actually played well at all. Fortunately for myself, I know now that whenever I am reviewing a game and feel bad about it not to play another game until the next day. (The next day I hit 1700) However when my coach reviewed it, he said it was very nice. The reason it was harder for me to be accurate was because the position was much more complicated.

Meanwhile, in this second game, which was a classical game I played just 2 days afterwards, there were much different results.

This game I didn't feel nearly as good about as the previous game I showed because there were no real complications I had to work through. My opponent blundered on move 10 and I had no complications the rest of the game. When I checked the game analysis later, I saw the shocking 99% ACCURACY. The only reason my game was listed this way is because the computer has so many moves that win in the position because it is winning by so much and so simply. If you look at the game, the computer has around 7 moves that are accurate for each move. Anyone over the rating of 1600 would likely be easily able to have the same or if not extremely close accuracy to mine because it is so simple.

This is why the accuracy number doesn't apply to humans because that 99% didn't mean that I played like a super GM (as much as I would like to play like that) it just meant that I played solidly in an extremely simple position. Meanwhile in the first game my accuracy went down tremendously from 3 moves because I missed the same tactic each time. I played just as solid as in the second game throughout the whole other part of the first game in my opinion, but since it was in a more complex and less clearly winning position it felt much better to play.

The funniest thing about this second game, is that when I got my coach to review it later, he said "not bad, good job" which made more sense to me because there was nothing crazy or briliant I did, I just didn't make blunders in that game. Meanwhile, for the first game which the engine analyzed me as not doing as well, my coach said that it was a "very nice game, good job" I agree with his analysis of the games more than the computer because the way in which a computer analyzes games is if the person is a 3700 rated bot like itself and expects it to do everything perfect, meanwhile a coach gets in tune with the way you play, what your strengths are, and what you need to do to improve. This is why I believe having a coach if possible is much more effective than engine analysis.

Now I can hear everyone saying, "I don't wanna pay for a coach" or "I can't afford a coach" The fortunate thing for you is that you can be your own coach. When you go through each game, go through your thought process and see what went well and what didn't, and figure out where your weaknesses are so you can figure out what you need to improve. For example, for me currently my biggest problem is rushing winning positions and throwing them away. This is something the engine cannot tell you. I have found it out myself from analyzing my games myself.

I'm not saying the engine has no place in your analysis, I would just recommend its use after you have analyzed the game yourself to make sure you didn't miss anything or didn't see a critical line or tactic. The engine should be used as a tool and a part of your analysis not your whole analysis.

Thanks for reading my blog everyone, if you want me to make more blogs like this just either DM me or get this blog to 50 likes.