(Updated) game analysis for improvement in play

Game analysis setup with CB18

At the start of this blog, which coincided with a new phase of more serious (and continuing) chess improvement efforts, analyzing your own games was identified as the necessary heart of an improvement strategy. This continues to be the central driver of the process for me, which is essentially an ongoing cycle of self-diagnosing issues with my game, then investigating and applying the chess knowledge gained as a result.

There are multiple guides on doing your own game analysis out there that may be helpful - although I've found some of them lacking sufficient detail on the "how" part of it, or sometimes recommending particular procedures that I don't find practical for my own purposes. Since the methods (and some of the tools) I've used have shifted a bit since "Game Analysis For Improvement in Play" was originally published, I thought it would be worth providing an updated analysis methodology outline. Caveat: this is not necessarily a template that will work best for everyone's practice, or that should be followed 100%. Rather, the intent is to provide a example of practical study methods in action. (I am a big proponent of the "whatever works for you" school of training, covered in more detail in "Do study techniques matter in chess?")

Some core tools are required for any setup: databases (at minimum one with your games, paired with a large reference database for comparison); an analysis program/GUI; and a top-class engine. Widely available options, both commercial and free, are covered in detail in the Chess Computing Resources (2021) post. Despite the age of it, the relevant links still work, and you just need to make sure to get the latest versions of the products; some of the main computing resources are also permanently linked in the sidebar. (For reference, my current analysis setup is displayed up top: ChessBase 18 running with the Dragon 3.2 engine and Annotated Game #323 displayed.)

Here's a description of my current step-by-step process, which typically takes around two hours for a fully annotated game:

• The game is entered into the database; I have one for all of my tournament games and a separate one for training games. If you are playing online at Chess.com, lichess.org or other sites, normally you can simply download a pgn file of your game and copy that into your database. For OTB games, naturally you'll have to enter it manually via your analysis program GUI - unless you're part of the elite where you play on an electronic board that records all your moves in the cloud. 
• During a tournament, I will make sure to enter all the games into the database before it is over. However, this will not necessarily happen on the same day they were played, if I'm too tired. Because it would be too much of a time and energy suck, I won't do a full analysis of my games while the tournament is ongoing, but I will normally take a quick pass through them with an engine running in order to validate (or refute) my understanding of what just happened in them. (For example, I would feel stupid to get caught out by the same error twice in a tournament.)
• During the initial manual entry of moves, or as a first pass during the analysis process if the full game notation is already copied, I capture my thinking process on significant moves in text comments. This could start as early as move 1, if the opening selection choice is worth mentioning, or it could start at the end of a typical main line sequence. In any case, I don't worry about polishing the language for publication, since it is more important to first record the thought process (whether accurate or flawed) and any other significant considerations at the time (physical or emotional stress, time trouble, etc.) that may have affected decision-making. This is also a good time to record other candidate moves that you considered, along with any related commentary and/or sample variations.
• During a full analysis session, I review the game from the beginning move-by-move with a reference database and engine running. Normally I will also have my opening repertoire database up in a separate window, so I can refresh my memory of the relevant opening line and note any deviations made by myself or my opponent. Here is where I will also investigate any new moves in the opening - new to me, at least - and update my repertoire database as needed. Considerations at this stage include the move's score percentage in the reference database, combined with the engine's evaluation and my own understanding and preferences in the opening line we are following. It is often useful to review some of the master-level games in the database that continue in the line being examined, to get a better idea of resulting middlegame positions and plans.
• After the "book" opening phase is past, the analysis process is devoted to examining and validating (or refuting) the actual moves played in the game, comparing them with the engine's and database's top other moves (for as long as the database has identical games). I normally have the top 4 engine moves displayed for variety, to get an idea of what good candidate moves are available in the position. Key variations are identified, understood (important!) and entered into the commentary, including alternative viable strategic plans and any tactical opportunities/threats overlooked (sometimes by both players). During this process, while the engine serves as an excellent reference, it is very important to avoid simply taking the top engine line for granted as "best" in a position. This is something which especially should not be done if you do not understand the moves it recommends. ("Pitfalls of Computer Analysis" goes into more detail regarding the practical considerations of using an engine for analysis.)

It was surprising to me how much rich content could be derived from just one reasonably thorough analysis of a tournament game. Especially when done regularly and in a timely fashion - something I have not always been the best at - it also inevitably highlights meta-trends in your performance. For me, that has included the recognition of a lack of a consistent thinking process, recurring flawed moves in a particular opening, or a repeated failure to maximize piece activity. Perceiving and rectifying these types of bigger problems with your game, not just identifying individual blunders that are made, can represent a major investment and payoff in long-term improvement potential. Sometimes, of course, positive trends in our game are also highlighted; it's especially important to recognize and celebrate when your mental toughness helps you draw (or even win) a lost position.

Some other typical training benefits I've observed, after employing the above methods:

• Openings: playing, then reviewing and analyzing what occurred in the chosen opening, has been the most effective way for me to understand more deeply and then remember opening lines for future use. This process also naturally results in the refinement of your repertoire based on new information and the practical challenges being encountered in your chosen lines at the board; there really is no better feedback loop.
• Tactics (and not just for the middlegame): what tactics were completely missed? Seen but miscalculated? Visualized improperly? Recurring patterns such as missing the idea of potential backwards piece moves, not visualizing discovered attacks that appear after several moves of calculation, etc. are in fact common phenomena. Although we cannot completely eliminate them, being more consciously aware of these issues in your games and mentally calling them out can help us overcome these blind spots.
• Candidate moves: in both the middlegame and endgame, the number of different types of candidate moves considered (both positional and tactical) has expanded significantly for me. Other than highlighting specific forced wins or clear blunders, this I believe is the most important function of engine assistance in analysis: showing the player possibilities that they would not otherwise have considered (either at the time, or during unassisted analysis).

As mentioned above, I believe the exact methods used are probably not as important as the fact game analysis is consistently occurring at the effortful study level for a player. In fact, I think it would actually be rather difficult to avoid improving your game as part of this process, if it is done regularly and you deliberately follow up on what you discover with both curiosity and a positive intent to apply the lessons learned.