Usefullness of analysing ones own games with an engine

Question

There are different opinions on analysing your games with an engine. Some say you shouldn't use an engine at all, and others say you can use it, but you should do your own analysis first. Most say that you shouldn't use the engine without a manual analysis first.

But what can you learn from an engine analysis? They can point out tactical mistakes, but if the mistakes imply you must correctly play a long variation, and you don't manage this, the whole variation fails. But of course, often the mistakes are easy to see. But I have started to ask myself, how valuable is this at pointing out tactical mistakes? Often I can find the easy mistakes by myself, and the hard ones are probably above my understanding.

Maybe the engine is more useful for endgame? But even so, it can be hard to understand the engine moves.

For the opening I think an opening-db or opening books are much better then the engine.

The best option is to probably analyse games with a stronger player/coach, but I do not have that option. You can also try to get help from players on sites, asking for help with analysis. I have tried this on chess.com several times, but I have gotten very few answers. I am trying Chess Stack Exchange to see if I get more responses.

I found this article that has lots of good advice.

Answer 1

I guess a lot of it depends on how strong you are. I understand what I am trying for in my games, but that does not always mean that I am efficient in achieving the aim. Sometimes, the computer shows me a good line quickly. In addition, I am human, and short of being Carlsen, and even he misses things, the computers is great for pointing out what you miss, but not really great teachers. If you are rated 1000, and have little positional understanding, then it is really not going to teach you much beyond any forcing lines you missed, so I would not spend a great deal of time with it.

Personally, I do not spend a lot of time analyzing with a computer, and in fact, just a few minutes per game, more as a blunder-check to show me obvious things that I missed.

I do not think that they are great for positional lines as what looks good to a human is often not what the computer suggests. In concrete, forcing lines, they are the best. Again, you can look at Carlsen since he often does not play the optimum move per the computer (although he does more than any other human still), but he clearly has a concrete plan in mind in most cases.

The days of computers playing endgames poorly is long gone, even if you can find exceptions here or there...especially with tablebases, which play endgames with less than 6 or 7 pieces perfectly. That said, sometimes, their hidden resources are very deep and are very specific to the position, so they do not help you gain any long-term knowledge. I cannot find the game again, but just the other day, a GM could have saved a draw in a rook endgame with Rd3 instead of the natural Ra3 (maybe b3). There was a peculiarity way down the road that no human would be expected to see, so he missed it. The odds of that coming up again are minimal, and certainly even less if you are an average player. Sure, it is interesting to look at, and to get the "truth", but how helpful that is in the long run is debatable.

The best way that computers can help in the endgame is if you have previously studied some aspect of the endgame, and you feed the computer a similar position, and play against it.

Answer 2

First of all it kind of depends on your playing strength, but every player can benefit at least a bit from an engine.

Personally I use the engine for the following analysis:

• Opening preperation: I play out different variations of my opening with the engine running. I do this so I dont miss any critical good moves of my opponents. For myself I dont necessarily chose the best engine move, but the move I think fits my style the best. At the end I check if I like the position and what the engine says of it. I dont care about +- 0.2, as long as I feel the position is good for me.

• Open issue control: If I play 10+0 or 15+15 online, I dont analyze every game by hand afterwards. But I will go for a quick engine run and check, what kind of mistakes I made (if there are any). This way, I get a overview about my typical mistakes and know where I can improve. I also check, if I followed my prepared opening ideas (if present) and how the position developed.

• For club games, I will always first analyze them by hand. The time spent depends a lot about the position and how many ideas I find which are worth analyzing (or ideas I had in the game but could not calculate due to time). Of course I will analyze my mistakes I spot myself. Afterwards I will run the engine and check if I missed any mistakes. In addition I will play out some lines of the engine which I find intresting. Afterwards I think about, wheither I could have spotted it during the game and if it was something I could play myself, regarding the resulting position.

In general, I think an engine gives you intresting insights, but you should always remember, that you dont have to be able to play like the engine. So the engine will find some lines, which you should dismiss, because they are too hard for you.

For additional benefit, you can use a book to improve. I will sit down with a chess book go over some content and later on use the engine to double check the ideas I learned in the book.

Answer 3

I've done it both ways (computer first, and human first) and I think you can learn either way. The important step is "human last," meaning the human needs to go over the computer's analysis to be sure they understand it, because that's where a lot of the learning occurs.

If I'm going to start with myself analysing, then I will play through it, trying to get a feel for when the game turned on me (assuming it was a loss, I find most of my problems with losing games comes from not recognizing those potential "turning points" in the game until the chance has passed). If I missed a tactical shot, I'll ask myself why I didn't see it; maybe that answer will point to a flaw in my approach to analysing a position.

Then, when that's done (or when I start with the computer) I'll let the computer loose on the position, probably overnight. At every point where the computer and player disagreed, I'll try to understand what the point of dispute is. Possibly it's something, possibly it's just a matter of taste. (IOW, maybe the player's choice differs from the computer's only in the path, not the result; the simplest example is a difference between a forced 8-move checkmate and the forced 5-move checkmate the computer saw. Same result, different path to it.) The important part of this phase is I arrive at some understanding of the computer's point.

Maybe it is that the computer sees this long variation; if it's the kind of variation only a computer can play well, I take that as a sign I don't want to go down that path, and I should look for a clearer course.

It's always possible I don't end up understanding every point the computer has. Or, even more probable, I misunderstand it, drawing the wrong conclusion. I know that happens, and I'm sure I carry in my head many examples of that of which I'm currently not aware. That's called "being human" -- all I can expect is those misunderstandings will be cleared up at some future point in my study.

But the point is in trying to understand them. From that effort comes the fertilizer that enables your chess understanding to grow and prosper.

I prefer to start doing the analysis on my own, these days, and only start with the computer when I don't have time. But you can make your own choice which way to start, I think either works well.