Moving Beyond: Meta-Analysis
Hi!
When I started blogging on here, I wanted to use this online space mainly for the purpose of writing analyses of my own games. I felt that I was stagnating in my play because I hadn't been working on my own improvement hard enough, and if I ever wanted to become a serious chess-player I should start behaving like one. Over the past several months I allowed more different types of writing to figure in my Dailies and my other blogs, because many thoughts and topics have potential to be converted into a somewhat coherent story, and because the writer in me refuses to go dormant. Still, game analysis is the raison d'être of my blog.
My lengthy post on checkmating with knight and bishop led @VOB96 to ask me what I value and focus on when conducting my work on analysis and research. For the occasion, I abbreviated my response to only a snippet, which can be read here. However, I have quite a lot more to say on the subject.
In this blog I aim to provide a few more thoughts on what I consider to be good analysis. So I'm analysing the analysis. Since the purpose of analysis is to identify mistakes and prevent them in the future, it only makes sense to use some examples of very bad analysis as well. There's a reason why bad analyses fail, and you can learn a lot from what to avoid.
Who analyses the analysis?
Introduction: My personal history with analysis
3. How Not to Analyse Chess
B) Genuine
C) Fundamentally correct
Introduction: My personal history with analysis
When people ask me how long I've played chess, my go-to answer, which is essentially correct, is "I can't remember ever having been taught the rules." In other words, I've played chess all my life. I learned it within the family, and I've been taught the fundaments of the game by a related generational peer and his father. Most of those memories have been obscured by the mists of time by now, but I still remember one lucid moment. At some point when I was pursuing my opponent's king with my rook, I became very aware of what I was doing. That moment that you're standing in front of the St Peter Basilica in the Vatican and suddenly realising "I'm here." It was something like that, but with moving a rook over an otherwise almost empty chess-board.
Learning the rules and playing quite many games with my peer had me hooked on this game. I wanted to read and learn more about it. My peer had a small booklet that contained themed exercises: four diagrams on the left page, their solutions to the right. These exercises were above my level back then, and so I didn't really try to solve them. Instead I looked at the diagram on the left and read on the other page how the position would go.
After finishing this book (which also included the Rellstab-Schönmann game that I used in "Chess Inferno"), I set out to look for more material. One of the oldest works I acquainted myself with was Max Euwe's Schach von A bis Z, which remains one of my favourite chess-books ever. One of the things that makes this work stand out is the no-nonsense approach in the language with which Euwe speaks to the reader. An excerpt from Chapter 4 (which I translated into English for the purposes of this blog):
The sham sacrifice
With a sham sacrifice, one usually aims for an improvement of sorts. Below follow a few examples of such sham sacrifices.
The reader now finds himself in the midst of the theme of combinations and countercombinations. Naturally he will have enough trouble with all of this. He should not think, however, that the combining of a seasoned player will all go smoothly; no, it remains difficult. One who is able to calculate the complications above in his head on his own has already made a lot of progress. However, the reader doesn't need to become discouraged when he doesn't succeed right away. Everyone makes mistakes, including the greatest masters!
Imagine 200 pages of such concise, accurate and beautifully written chess material, and you can understand why this manual is one of my favourites. Euwe's analytical approach has been a great help in my development both as a chess-player and as an analyst. At first I tried to emulate Euwe's style a bit, which lead to some remarkably primitive attempts that could very well fit in my Dunning Kruger blog. Though it would take another so-many years before I was even remotely near my current level.
When I returned from a tournament and was asked at the club to do commentary on one of my games, I found myself a bit at a loss. I had a good game in that tournament, but I'd never done anything like it before. It was a spontaneous bunch of blabbing that I don't think made much sense whatsoever. I felt uncomfortable, someone in the "audience" started laughing at how convoluted it undoubtedly came across, and one of the trainers said something to the effect of "Some people can do this, and others can't," implying that I'd belong to the latter category.
I later analysed the same game for our club bulletin and decided to give it a fitting theme. The idea to play the King's Indian, by the way, came from Euwe's work:
And so my journey through the world of chess analysis began. Since those early days I've accumulated a lot more experience, read a lot of good and bad material, and learned vastly more than what the sneering trainer from a few paragraphs above could have ever taught me. At this point in my journey I can look back and safely conclude that the implication of that trainer was quite wrong.
Much has changed since the days of Euwe's world championship. Thanks to the efforts of titans like Botvinnik, Polugaevsky and Kasparov, the level and depth of opening preparation had already increased greatly before digital databases were even invented. For their times, they were armed to the teeth with preparation, but those levels of prep pale in comparison with what was yet to come. The development of newer and stronger engines meant that more and more opening lines had to be reassessed. Euwe deemed the Sicilian insufficient at the highest levels of chess, and we know how well that assessment stood the test of time: the Chess.com database now contains more than twice as many games with the Sicilian than games that started with 1.e4 e5.
One of the games that did start with 1.e4 e5 that was decided by engine preparation work was the following World Championship game between Kramnik and Leko:
Kramnik's engine-backed home preparation failed in this game because of Leko's human grandmaster intuition. This debate was thought to have been settled after the Kasparov-Deep Blue match of 1997. This may well have been the human's final say in the matter, and given the 2025 controversies surrounding the white player's social media rampage crusade, it's convenient that he found himself drawing the short end of the stick in this specific setting.
Technologies of the present day have radically changed the landscape of chess work, and will continue to do so. This includes analysis. These days, almost no serious aspiring chess-player can hope to operate without including external sources like game databases, chess engines, and personal repertoire playbooks. The amount of available sources with which to learn, enrich, practice and perfect your game (literature, YouTube, Chessable courses, and what-not) has exploded. We're living in an age of information abundance, and surviving and thriving in the world of competitive chess has come to depend to a good extent on the player's abilities to put the technologies to their best use.
Since chess remains a game between humans, this begs a few questions. Does the abundance of information manufacture an improved understanding by default? Are your sources reliable, and why are they (not)? Will every bit of homework be useful? Or can you overload your brain with your homework so that it won't function in your next OTB game?
To provide my quick view on these: I believe in hard work, but I regard people who only stop their training to eat and sleep with high scepticism. Overstretching yourself in the gym could cause serious injury, meaning you'll have to recover for months, and all the progress you thought you made turns out to be wasted energy. The same goes for doing too much chess training or playing.
However, these questions lay beside the scope of this text. What I instead prefer to focus on here is the role of technology in the quality of analysis.
A game of chess is a journey chronicled by two authors with diametrically opposed agendas. Both teams want to checkmate the opponent's team's king. If this option isn't available, the best alternative is to prevent the opponent from checkmating us. In any position, the best moves resonate with the demands of the position and progresses that side's case towards the goal of the game.
Philosophically, the task that the game of chess poses to its practitioners is to find the best move in every position. This poses an immediate practical problem: it's impossible for a human to do. Within a handful of moves, the number of possible games is larger than the amount of seconds that humans live. And so we have to be more economical with our time and resources. We can't look at everything.
In most cases, analysis is the process of finding the best move and/or discovering what is happening in the position. There are several methods that you can employ, such as Silman's Imbalances method, Dorfman's four questions, or more recently: the KIMPLODES method. Most of them start with observing what is happening in the position. How you go from there depends on your prior knowledge and experience, the method that you're accustomed to, the circumstances in which you perform the analysis, the external material that you're using (if allowed), and the goal that you have in mind for your work.
The way that you use analysis depends on the goals you pursue, the stories that you want to tell. Writing an opening guide for beginners would require a different approach than an analysis of central pawn thrusts in the early middlegame. When you analyse your own games, you hope to discover whether your thoughts and calculations were spot-on or off the mark. If you're preparing your opening repertoire for a tournament, you hope to get a combative position with good chances for success, as well as a good understanding which pitfalls to avoid and which traps to lay for your opponents. When you write a fictional chess-story, analysis can help push the narrative forward or even take the place of the storytelling.
When I wrote the comparative story between chess and the plundering of the pyramid from above, or the retelling of the Trojan War, I had the analysis take the place of the storytelling and used the chess-game as its background decorum. And these are certainly not the only ones: in Choose-Your-Own-Adventure blogs such as "The Lord Of The Pawns" (which hit on the same day as I publish this blog), analysis is an essential skill to get you through the story.
As I said in the @VOB96 blog mentioned above, analysis has to meet three criteria: it has to be functional, genuine, and at least fundamentally correct (provided that it's in line with the other two). I'll discuss each of these in turn.
Functionality means that the analysis should serve the supporting role that is laid out by the goal of your undertakings. This includes taking your target audience into account. Your analysis doesn't have to be 10 variations wide, 47 moves deep on every turn, and engine-level accurate. That would most likely make it factually correct, but hard to read or follow for your intended reader.
If you write an opening guide for beginners on, say, the Caro-Kann, your goal is to make this opening accessible for these aspiring players. You start with what they know (the basic principles of the opening) and expand on this knowledge by explaining how these principles apply to the opening in question.
I’d expect such an approach in a book called First Steps: the Caro-Kann, or The Caro-Kann: Move By Move, but alas: Andrew Martin and Cyrus Lakdawala (otherwise brilliant authors and coaches) leave a lot to be desired here. However, much of their starting point is describing things that the beginner doesn't really need to know. Jeremy Silman approved of the idea to feed people advanced concepts, but when I started to learn other openings it left me very confused.
Let's look at the introduction of these two books. The following is taken from Andrew Martin's First Steps:
You can certainly pick up a few advanced ideas, but I'm left to wonder why it's worth mentioning the open diagonal of Bc8, what that white initiative consists of, let alone how it is supposed to be quelled, why I should know how popular the line is, what piece placement I should strive for and what pitfalls I should avoid, and how this opening translates to the basic opening principles.
In MBM, Lakdawala's book deals with a different variation:
In the ensuing Q&A discussion, Lakdawala talks at length about his preference for this line over the Capablanca variation, but never mentions 4...Nf6. And the most important Q that I'd have (How in the world does this variation constitute a correct implementation of the basic opening principles?) is never asked nor A'd.
I would’ve been okay with this if in either series a standard volume had existed that would shed light on these basic opening philosophies and how these opening lines relate to them. But that work doesn't exist.
Tactics in the Chess Opening 1: Sicilian Defence
When I was growing beyond the basic opening principles, the first real opening about which I learned something was the King's Indian with black (thanks to old works by Max Euwe). The second opening that I wanted to learn was the Sicilian. This opening was popular at the highest levels, and I had seen my first ever trainer play this opening line in some games. I never understood this opening, nor did I have the guts to ask him what this opening was about. My resources at the time were very limited, and I went for the very first book that I could find: Tactics in the Chess Opening 1: Sicilian Defence. This book was to take my game to the next level.
Right?
This book has no foreword, no preface, and no introduction. Except for the back cover of this book that promised to make me a stronger player, what you see in the viewer above is essentially page 1. The very first communication between the authors of this book and the reader is that 7...Nc6!? is "[a] provocative move for which there is no known refutation." No fundamental opening idea is explained in any of this excerpt: all of that is presupposed as prior knowledge.
It almost goes without saying that when I worked my way through the first two chapters of this book (the Najdorf and the Dragon), I didn't learn anything whatsoever and basically just wasted my time. I don't think that this book is bad, but it would've saved me a lot of headaches if the authors and publishers had indicated a target audience somewhere.
These days I'm able to make this book much more accessible for my younger self. I'd start with a chapter on the basic philosophies behind the Sicilian opening for both sides, dissecting the strategical core of the opening in a true move-by-move fashion. The reason why I believe this is essential is that it provides the reader with a solid understanding of what happens, so that they'll be better able to find their way if they're taken out of book by an offbeat move early on. It's lamentable how little this is done.
If I'd redo this Sicilian book, I'd proceed to write a short piece at the beginning of each chapter to explain the basic ideas in the line at hand and the ways in which it differs from any of the other variations. This is a bit like John Emms did when he wrote Starting Out: The Sicilian, which I didn't have access to at the time.
But the book on the Sicilian wasn't the most confusing and opening book I've ever bought. That questionable honour goes to the following book.
Sabotage the Grünfeld
When I started becoming interested in adding 1.d4 to my repertoire, I thought to look for a way to counter the King's Indian and the Grünfeld. I especially knew that one of my clubmates at the time really loved playing the Grünfeld, so I thought to prepare something against him.
The back cover of this book reads the following line:
Sabotage the Grünfeld offers a highly practical repertoire based on 3.f3, covering all of Black's answers. Larry Kaufman, who has first-hand experience with the variation, has analysed the lines deeply with the high-end computer programs Komodo and Houdini 3.
Highly practical, engine-approved, and experience. Well, let's have a look, then.
Chapter 2 begins the theoretical discussion:
If I'm not mistaken, all of the royalties have been spent on covering the massive energy bill that resulted from running the engine for months on end. An analysis like this is completely useless from a mortal human being's perspective: I don't want to have the silicon substitute for confidence. I want the real deal.
And that brings us to the second aspect of the analysis:
Genuine analysis means that you use your own words and insights. This is very different from reinventing the wheel at every turn. Every present-day author uses external sources for their work: other literature, databases, programs, courses, engines, and verbal advice by their peers. Most books include a bibliography that lists the sources that the author consulted during their writing.
Blogs, on the other hand, are not academic works, so in most cases I don't expect to see a bibliography at the end. What I do expect to see is that whenever an external source is used, the author, narrator, commentator or other content creator will be transparent about them and mention their sources in one form of another. Even just "the engine" would make sense here.
I'm quite allergic to something like this:
This is what you'd get if you just let the review batch do its work and don't bother to break this down for your audience. None of the annotations are explained at all, no alternatives are suggested. It's just a chess-game smudged with a few different colours. This would've been a more genuine analysis if it was left completely blank.
It has to be acknowledged that chess (at least competitive chess) is primarily a game between human beings. During chess-games, you're not allowed to receive assistance from your silicon friends, or from your flesh-and-blood ones for that matter. You're on your own, and you have to find your own moves. In order to best approximate the best moves, it becomes essential to know what is happening on the chess-board.
But how can you know as a human being what is happening on the board?
And how would you explain this to another human being?
Precisely this is one of the things that your engine is insufficiently able to really help you with. Engines calculate, compute, and assign numeric evaluations to the position from move to move. You can conclude from a sudden shift in evaluation which moves are suboptimal, ranging from "good enough" to "blunder." Engines don't tell you why the move is a blunder. Nor do engines tell you what caused you to make this or that mistake. Behind most human mistakes is a psychological cause that your engine will not be able to understand.
Let's look at this position, which occurred in one of my recent OTB games:
The human thought (my thought during the game): "This is a good move. The rook enters the 7th rank with tempo against b7 and can then team up with the queen to exert pressure against f7, which would paralyse black to a very large extent. Provided that I'll be able to restrain black's queenside pawns, black will be unable to develop any activity, and I'll be able to look for a win on the kingside at my own leisure."
This is what the chess.com review batch returns: "29.Rd7 is an inaccuracy (+1.51). 29.Qe3 is best (+1.84)."
A very similar thing occurred in the game between WIM Anna-Maja Kazarian and GM Eldar Gasanov, London Chess Classic 2025, Round 4, played at the time of writing. After 19 moves, the following position had emerged:
White has the option to push 20.d5 here. It's a heavy decision to play this move, because it drastically changes the situation on the board, as it gives up the tension and closes the centre down completely. The one proposing this move in the Twitch-chat of the Dutch WIM justified the move as "Because it's the best. It's +1.5." There had to be a rationale behind it, but the one proclaiming that it was the best move wasn't able to back up their statement. It was then aptly explained in chat, and the explanation goes a bit as follows:
Black has more space on the queenside and could be looking to open up the c-file at an opportune moment. The queen on c7 and the knight on a5 are positioned to welcome this course of events. By closing the centre with 20.d5, the queenside will also remain closed (b5-b4 will be answered with c3-c4, and c5-c4 will be met with b3-b4). This means that there are no targets that black can assail, and in turn the knight and the queen are no longer productive. The pawn on d5 also takes away the squares c6 and e6 from black's pieces, and especially Na5 will find it difficult to join the game after that:
At the same time, white stabilises the centre and secures a space advantage, which can work well in conjunction with an attack on the kingside. Since the centre and the queenside are now closed, black can't initiate a counteraction on the queenside, and will thus be forced to defend. This is extra difficult with one piece as badly offside as the knight on a5.
None of this becomes evident just from reading off the engine that 20.d5 is +1.5.
The following excerpt is from a YouTuber that I'll leave nameless here because of ongoing allegations of cheating. The general consensus among anti-fans is that the content creator's alleged playing strength and the level of analysis that they provide are incongruent. Inspired by the discussion on the related forum that I won't link here for the same reasons, I found the following excerpt. I could've placed this "brilliancy" in each of the three sections of this blog, but chose to place it here because of its similarity to the previous example: a central pawn that crosses the middle.
The breakdown of this excerpt by the commentator is simply garbage. They seem to have no memory for what happened just one move ago (scrolling back), but this is the least of the issues here. The thinking is very simplistic and lacks the nuances that I'd expect a 2300+ rated player to have developed along the way. They seem completely oblivious to the far-reaching consequences of the pawn thrust to e5, justifying it by wishful thinking. This is a very peculiar positional concession to white's central structure as it cedes control over the d5-square to black and opens the diagonal for the enemy bishop.
I've fed the same position to the engine, but rather than the atrocious 8...b5 and the cryptic 10...Bf4, I made black play 8...h6 and 10...Be7. In this adjusted position, the engine evaluation dropped significantly upon the pawn push:
This proves that just "attacking and winning the knight if black fails to spot the threat" is a bad justification for playing this move. The only reason why 11.e5 worked in the game is that it wins a pawn with tempo by force. A positional error like this will need a tactical justification to be played. If such is unavailable, such pawn moves should be played with the highest caution, if at all.
The argument from authority ("I'm 2300+") doesn't hold for this content creator, because they're unable to back this up with analytical efforts that meet the standards that one may expect from an expert. It certainly doesn't make a convincing sales pitch for content that as insincere and practically abysmal as this is.
There are several problems with conducting insincere analysis. From a writer's perspective, it's likely to make your content miss its purpose. If you're trying to learn something from your own writing (which is the reason why I started my blogging journey), it's much more fruitful to immerse yourself in the material. If you don't know something, that's also fine. Like Stephen Fry once said in a QI episode: "There's the Socratic acceptance of the limits of one's own knowledge, and there is ignorance."
Of the three elements, factual correctness is ironically the one that's most subjective. The point is that the level of required correctness has to be proportional to the purposes of the analysis. In opening preparation, many of the world elite deliberately steer away from the objectively best lines to avoid falling into the territory of known draws. When you're writing a practical opening guide for beginners, you can get away with recommending variations that are less challenging but easier to learn, as long as they don't lose by force. If you write fiction, it's very forgivable to sacrifice accuracy for the purpose of the narrative.
In cases where you try to get to the bottom of your material, the correctness of your analysis becomes more important, and in such cases it's the easiest requirement to provide. This is where the machine comes in as a trusted helpful assistant. Its evaluations still need to be translated into human language to make them functional and genuine, as the excerpts in those sections demonstrate.
Given the philosophical goal of chess mentioned in #1, you'd be inclined to think that the engine can take over the analysis in full. Ever since the Deep Blue-Kasparov match (or, rather, since Leko clapped his opponent in the 8th round of the 2004 World Championship match), engines have become rather much stronger than humans, and any review tool can indicate the best move much quicker and much more accurately than your trainer can. You're able to generate a full review of your game within sometimes seconds, and you don't have to wait for your trainer's schedule to clear up.
To cut Kaufman some slack: these days, every book is flayed by the use of the engine. However, as I said to @SPK1729 when asked for a quote to be used in the collective effort "Game Review," engine evaluations are completely meaningless if you have no idea how they relate to the situation on the board. I just don't swallow "White may be marginally better" accompanied by a numeric evaluation. In this sense I'm happy that Kaufman's approach has improved in Kaufman's New Repertoire for Black and White. There's still the numeric engine evaluations, but Kaufman now makes an effort put these numerics into human language. Even if this is often abbreviated to "Bishop-pair," it's much more functional than just printing an engine evaluation.
The "fundamentally correct" bit could almost be skipped because it's the most trivial one, and because I've already included several examples of its importance. Still, there are a few notable examples that I'd like to discuss here.
The Alien Gambit
The Alien Gambit was proposed by Bulgarian CM Volen Dyulgerov (@Witty_Alien) and popularised through his Twitch and YouTube content. This response to the Caro-Kann is fundamentally unsound, but the game could turn complex, which might make it work in blitz and bullet encounters. As a Caro-Kann player I'm always very happy to see my wishful-thinking opponents employ it against me: my current score against this is 17/22. I'd like to mention that @NMChessToImpress has attacked this El Dorado system in his Hallowe'en post. To his credit, Witty_Alien himself admits in his course Vaporize Your Rating: Alien Gambit that the opening line is questionable.
Levy Rozman (@GothamChess) has proposed the Alien Gambit in this video. This I can't take seriously. The statistics of a 60% win rate (based on the lichess database) are fallacious: 6.Nxf7 doesn't appear in the Masters database, and this percentage drastically shrinks to 38% as soon as black has prepared the antidote based on ...c5, ...e6 and ...Kg8.
The only way to reconcile this is by seeing the video as mostly entertainment, but it's too much a stretch to do this since the material is presented like a promo. Rozman eventually does get to black's most critical moves 7...c5 8.c3 Nc6, and only mentions the truly relevant information (sub-2000 and blitz) very off-handedly, almost as an afterthought. That's 10 minutes off my life that I'll never get back.
As can be derived from all of the above, good analysis is functional, genuine, and preferably factually correct. It fits to look at what I think to be good analysis. I'll choose three different types of media: a YouTube video, a blog, and a book excerpt.
GothamChess
While he has its occasional slip-ups (see above), Levy Rozman is capable of providing expert analysis in a very entertaining and accessible way. In this video he broke down a Carlsen game with the purpose to show the difference between a "regular" GM and someone who's been the highest rated player for the last 12 years at the time of the video. I decided to embed the video rather than transcribe it, because the delivery is an integral part of the analysis:
As an informative video, it delivers magnificently. Rozman mentions the alternatives, but doesn't linger on them because they fall outside the scope of his video. He pauses to investigate and demonstrate the ideas behind 13.Rae1 very elaborately, spending almost 5 minutes (10:22-15:15) for a move that I suspect most others would just gloss over. Since this way of deep investigative analysis goes hand-in-hand with his showmanship very well, it fits his purposes to entertain and educate, it's highly genuine (open and transparent about the use of outside sources, and using his own voice to string it together) and is factually correct. I'd classify this analysis as superb.
Max Illingworth
Although he is no longer active as a competitive player, Australian GM Max @Illingworth has a lot of expertise as a coach and annotator. He provided the New In Chess Yearbook with numerous opening surveys, and he's also the co-author of Dismantling the Sicilian (a collaborative effort with Jesús de la Villa, who has written the endgame classic 100 Endgames You Must Know).
As an annotator, he has written numerous blogs with very elaborate analysis. Considering that his Chess.com profile indicates that he hopes to help 2000+ rated players improve to master level, we can assume that the target audience for his analyses is advanced players. And for his intents and purposes, Illingworth delivers very well. The analysis below is taken from this blog:
Understanding that Illingworth is a fulltime chess coach, I can overlook his frequent sales pitches and see the analysis for what it is, which is a high-level endeavour. With his own style, Illingworth is able to turn an opening blog into something that hardly feels like your typical opening blog. A good example is his engaging "Why Australian GMs Love the Scandinavian" blog entry.
Illingworth has also written and recorded a tribute to Daniel Naroditsky upon the latter's untimely demise. And he is the last person I'll talk about in this blog.
Danya
Daniel Naroditsky is undoubtedly one of the most tragic figures of this year. Many people have created something in his honour already, and so have I. Within my entry, I showed an example of what the 16-year-old American was capable of in his books. Here I'll reprint the analysis he presented in Mastering Complex Endgames. As I mentioned, this analysis is perfect in every sense: it takes into account what happened during the game. It's no-nonsense, and makes a very good case for his own book.
In my opinion, analysis is one of the most beautiful and interesting parts of writing about chess. How you use analysis, and what methods you choose to put to practice, is up to you. It's important that you know what you want to do with your analysis, and that you make an effort to make your analysis functional (it should support the goal of your work), genuine (it's your work, and be open when you consult the work of other authors and/or the engine), and fundamentally accurate (no glaring errors). If you choose to use annotations, don't just copy them off the engine. Make an effort to explain what you mean.
I have a very long history with analysis, and with this text I hope to be able to provide you with some of the insights that I've amassed over the years. What analysis is, can be, and should (and shouldn't) do is very diverse, and I hope that it will be able to help you to enrich your work.
Good luck finding your own voice and using the beautiful tool that is analysis for your own works in the future!
