Analyzing Our Games

Maybe we're asking the wrong questions

Welcome back to Chess in Small Doses, an Substack for adult improvers at chess. Today I’m tackling the issue of game analysis. It’s almost universally recommended that players who want to improve need to analyze their games. How we should do that is an evidence free zone that is full of opinion. Do we use the engine or not? Do we analyze every move or just some? In my opinion, these are the wrong questions to ask. I believe the right question is how am I going to retain what I learn? It’s just my opinion, but game analysis without retention seems like wasted time. The answer to that question lies in The Rule of 3’s. Let’s get into it.

The Wrong Questions

First off, why should we even analyze at all? It may be heresy, but time is limited and we adults have stuff to do. Often it’s more immediately rewarding to just play another game. This can set up the bad habit of playing without analyzing. But there’s a quote that fits here: “Those who cannot remember the past are condemned to repeat it.” Without analysis, we simply repeat the same mistakes again and again. Therefore the value proposition for game analysis is improvement. What analysis gives you is the chance to learn from your mistakes. You can begin to collect “chunks” of knowledge that you didn’t have before. You can see what patterns you missed and add them to your mental library. You get to study your decision making, and look for patterns in your thinking. Just looking again at your games will naturally help you improve as a player.

But what does analysis even mean, and what exactly are we doing when we analyze? This area is sadly an evidence free zone, full of opinion. The general consensus is that we should analyze. Game analysis is almost universally recommended by coaches and stronger players for anyone seeking to improve. After that, it’s anyone’s guess how we should go about doing it. Use the engine or don’t use the engine? Analyze deeply on our own, or superficially for “just the big stuff”? Do we need a coach, a community, or a computer? There’s a lot of options here, and little if any evidence on how to choose. Perhaps that is because “How to analyze?” is the wrong question to ask.

It’s just my opinion but I believe the right question is “How will I retain what I learned?” My perspective is that game analysis without retention is wasted effort, and beyond pointless. Players who automatically run the engine, see what the blunders were, and then move on are very likely to forget what they just learned. There is a whole science behind learning that is still growing, but I believe we can answer this question with the Rule of 3’s (aka Hendiatris). To greatly simplify, we need exposure to something at least 3 times before we can be expected to apply it. Practically, it means we likely need to review whatever we learned in analysis more than just once. Another way of saying this is: One is luck, two is a trend, and three is a pattern.

First let’s talk about how to analyze our games, then we’ll get int to how to retain the learning.

3 Styles of Analysis

I present to you 3 options or styles to analyze your games. I leave it up to you to decide which one works best for you and how to apply it. For each of these options, I’m leaving out learning about the opening. I think both Noel Studer and Nate Solon have covered that pretty well.

Option A: Analyze Like Chernev

In his classic Logical Chess, Chernev wrote a sentence or more about every single move. (Ben Johnson reviewed the book here.) It’s almost comical what lengths Chernev goes to in order to describe 1.e4, but it gets the point across. Every move had a purpose.

In this style, we go move by move through our games without the engine and explicitly write out what that move does (and undoes). One immediate benefit is it forces you to think about the move. What negative and positive things does that move do? Is there a plan? Every move, every time.

Can you say there was a purpose for every one of your moves? Do you remember what you were thinking? Did you miss something critical in your opponent’s moves that cost you a game? Then this style might be right for you. One of the hardest things for improving players is learning to see. Analyzing like Chernev can really help develop your awareness over the board.

The downside is time. This style of analysis (which I have tried) takes a long time. My analyses would take at least 1 hour or more. This kind of analysis leads to steady, but slow, improvement. It requires patience, discipline, and focus. As a parent and an adult who is employed elsewhere, the time (and its opportunity cost) is a challenge.

Option B: You First

Go back through your game without the engine, one move at a time until you get to what you thought were the “big moments” that decided the game only. In each of those critical moments you identify, see if you can find a better move (for either side, it doesn’t matter which side made the mistake). Just do your best and see what you can come up with.

After you’ve reviewed the game, then turn on the engine. We’re not asking for a full analysis, just the engine’s opinion. For each critical moment, ask the engine for an evaluation and a line (or two). Try your better moves out, and see if the engine agrees. Record in your own words what you learned about what you missed or didn't see.

The upside is that you will improve your vision if you do this consistently. You will identify patterns that you missed. The downside is the computer will be showing you. The idea of “Everything is Obvious: *Once You Know The Answer” applies here. Our brain is less likely to retain this knowledge, but there are tricks we can use (see below).

Option C: Start Your Engines

There’s actually two ways to do this, but I recommend only one. The first is to request the full server side analysis and just tap through the moves, noting the parts you missed. I don’t recommend that. It’s hard to manufacture relevancy from that approach in my experience. You get a lot of “What the heck?” moments which somehow pass into the void about 10 minutes later.

Instead I recommend reviewing your game with the engine on, but the lines turned off. Just the eval bar will work. Go move by move and look for the huge changes. Then try to identify what the engine is pointing to. See if you can see it. Noel Studer recommends to only look at the +/- 2 moves in his course. I think that’s because as Jacob Aagaard said, “Modern engines are crap!” They’re too good, and are able to outplay everyone on the planet in almost every position. So just attend to the big stuff. Having the engine lines turned off will force you to engage with the position.

The upside to this option is it’s quick, like 10-15 minutes quick. You can drop the game into a study and just try to find what was missed. The downside is you don’t really understand the why sometimes. The engine will often recommend moves that seem odd to humans. Engines don’t speak English (yet), just algebraic notation and evaluation numbers. That’s why keeping it just to “the big stuff” will likely show you tactical patterns or important strategies you missed.

After Analysis - The Rule of 3’s

Whichever of the three options you select (or do your own version) is very likely irrelevant. That you analyze is likely far more important than how you analyze. It’s what you do after the analysis that will make all the difference. This step is not well spelled out in many resources. In Pump Up Your Rating, there is an entire section on “The List of Mistakes”. (I really recommend that book BTW.) This is a fantastic habit that others have mentioned. But what do you do with that list? Just remember to remember? Try hard to learn the lesson? Review it before big games? To quote David Allen’s Getting Things Done, “Your mind is for having ideas not for holding them.” Learning science has a different answer, the Rule of 3’s.

While this isn’t a rule per se, it is something that resonates deeply with Cognitive Load Theory. That theory states that humans have a limited number of things they can retain in short term memory at any one time. The use of three items in comedy, literature, writing, and public speaking goes back to the Greeks and likely works because it’s so easy for people to keep track of. Decision science seems to indicate that people are best at decision making when considering 3 options. Further more, educational science dictates that a student cannot be expected to retain something until they have been exposed to it at least 3 time.

Our minds are built to recognize and retain patterns. One exposure the brain may treat as random and unimportant. Two exposures and the brain may notice it, but if it stops after that then the forgetting curve takes over. Three exposures and now you have a pattern. You’re likely familiar with spaced repetition, but have you noticed how much longer you get after the 3rd exposure on Chessable? Don’t take my word for it, here’s the data.

So what does that mean for us? My point is that when we do analysis if we do not have a plan for retaining the knowledge, we have likely wasted our time. The plan can be as simple as reviewing the game at least 2 more times over the next week. However, let me recommend some options.

Save the positions

Whenever you found a position where you (or your opponent) made a critical mistake, save either a screenshot or the FEN. Put the position into a study or a slide deck. The key habit is to regularly review these positions. Anki will give you spaced repetition if you make slides (which I do). Lichess will allow to make a study which you’ll have to make a habit of reviewing semi regularly. You’ll get the adequate number of exposures for sure this way

Write it out, Talk it out

Multi-modal learning improves retention. Writing out what you learned from a game will help to reinforce it. Put it into your own words as best as you can. Then read it out loud, like you were teaching it to someone. Seeing it, writing it, and hearing it counts as 3 exposures. Your brain will retain more with this technique than just putting your analysis in a study.

Seek feedback

This is the Chess Dojo way, posting your analysis in a public space and asking for feedback. Each time you get that feedback, you have to go back over it. A coach is great, but so is a friend. Try to get at least two other opinions to qualify for the 3 exposures.

Whatever you choose to do, having a plan on how you’ll retain the learning is critical. Analysis is vital for improvement, but just the first step. Retention is the critical step that’s often missing. Using the Rule of 3, you can increase the likelihood that you’ll retain the learning instead of forgetting. Chess improvement is hard enough, but doing work that gives you no benefit makes it harder. Our time is precious so if you’re going to commit to analyzing your games, commit to retaining as well. Just choose a plan for retention and see if it makes a difference for you.

Thanks for reading and I hope you’ve enjoyed it. Let me know what you think. I’ll be taking a short break as I’m starting back up with the 45/45 league. Until next time, good luck on your chess.

P.S. AI on the Horizon

I came across a lot of stuff about computers and engines during my research for this piece. What was most fascinating to me was this video here. In it Chat GPT-4o is talking to Sal Khan’s kid explaining geometry to him and fielding questions. While Chat GPT currently sucks at chess and is a cheater, it’s likely to get better. AI capacity is doubling every 7-8 months. I can foresee a time within the next few years that AI will be able to act like a coach, giving advice and patiently explaining why one move is a mistake. Maybe… Stay tuned.