Know Thyself, Part I: Specific Preparation

 "If you know your enemy and your know yourself, you need not fear the result of a hundred battles."  That sounds absolutely great, and it applies to chess.  Indeed, preparing for specific opponents goes back centuries, but I am more interested in the first part: know thy self.

This week, I have written a two-part process on the idea of preparation.  Normally we prepare for a specific opponent, trying to learn strengths and weaknesses.  I have never done this, but I have prepared against myself.  In other words, if I had to play against me, what would I do?  What would I find?  If I know this, then I know more about myself, and I can use that knowledge to improve my game.

Today, I will cover the idea of specific preparation.  This is what most people think about: what openings do I play, how do they do, what vulnerabilities may appear, that sort of thing.  The second part will cover general preparation, which basically encompasses everything that isn't opening related.  Without further ado, let's prepare against Smithy.

Opening Analysis: White

Preparation has become near effortless with modern sites.  If someone plays online games, you can literally click on the profile, load their games and see the play rates, the win rates and an engine evaluation all on the same page.

The temptation is to exactly that: load the page, look at my most played move and then compare that with a) the move that scores best for my opponent in practice, and b) the engine move.  I have over 1,300 games, so there's a nice sample to work with.  If we try this, we get the following:

1.e4 (most played) c5 (best score with more than 50 games played), 2.Nf3 Nc6 (best score) 3.d4 cxd4 4.Nxd4 and suddenly we run into a problem:

Against the normal mainline, 4...Nf6, I score really well, and my overall score has increased as well.  I have a poor score against 4...g6 and 4...e5, but Accelerated Dragon and the Lowenthal, but the engine is skeptical of these moves and gives White a slight advantage.  Conversely, the best engine move is 4...Qc7!?, a very rare move with very few games.

Perhaps 4...e6!? is the right approach: the engine still gives White a small edge, but my score has been atrocious:

27% is very poor ... but look closer.  More important than the score, look at the recent games: they all show spring of 2019.  That's nearly five years ago.  We risk preparing against historical Smithy, not the current me.  Indeed, I was below 2000 rating in all but one of the games to reach this position.

This is the first lesson: we need to look at recent games.  If I played the Open Sicilian without any theory knowledge half a decade ago, that's one thing ... and very different from the current me.  Indeed, if we limit our search to 2021, so just the last three years, it turns out I haven't played 1.e4 at all.  We nearly risked preparing for a position that had zero chance of coming up.

Unfortunately, because I have played so few game in this period, the sample is less convincing.  For example, I score 45% against 1.d4 e6 ... but that's across 10 games.  Instead of looking at individual move orders, let's look at general set-ups or "families" of openings:

• Queen's Gambit: In the mainline QGD, I have a positive score in the absolute 4...Be7 Mainline but a negative score everywhere else; in the Slav systems, I have great scores against everything but the 4...g6 Schlecter Slav.  This suggests I am much more comfortable in Slav positions than QGD ones.  Interesting.
  
• The ...g6 Indians: In the mainline KID I have 7 wins in 8 games; in the mainline Grunfeld I have 4 wins in 7 games. Not a large sample size, but very good results.
• The ...e6 Indians: Here we go!  I played 3.Nf3 exclusively, scoring 35% across 17 games.  This includes negative win rates in the QID, Bogo and the QGD transposition.  These are definitely positions I do not play well.
• Others: Not enough data to get a sense of other options. I will point out that I am 6/6 against the Englund gambit, giving me a 2900 performance rating, so don't play that against me.
  

Anyone preparing against me (or more relevantly, for me trying to improve my own performance) should look at the QGD and the ...e6 Indians.  These are less forcing and don't rely heavily on theory or memorization, so it mostly comes down to understanding the position.  Perhaps that's precisely my problem: thousands of games in 1.e4, so lots of experience; only 145 games in 1.d4, so missing that practical experience component.

Opening Analysis: Black

It might help to start with a snapshot: here is a comparison of my all-time stats with my last three years.

My Black performance has increased across the board ... except against 1.d4, where it has gotten worse.  That must be the weak link.  What has changed?

Historically, I have played either the QGA (with a 58% win rate!), the Nimzo (63% win rate!!) or the KID (49% ... still working on that one!).  Since 2021, I have played the Semi-Slav exclusively, scoring 42% after 1.d4 d5 2.c4 c6 and a measly 25% in the mainline Semi-Slav tabiya.  Indeed, in 2023 I have lost every (!) game I've played in the mainline.  I am 0/7.

That last point really intrigued me, so I looked into those seven games in particular.  The results:

1. Game 1: equal middlegame, one move blunder.
2. Game 2: steadily outplayed in the Cambridge Springs (opponent found nice tactic with Bxe6)
3. Game 3: slightly better out of the opening, one move blundered into a fork.
4. Game 4: played the Botvinnik without knowing theory, lost terribly against +200 rated opponent.
5. Game 5: completely winning, had a mating attack, miscalculated, lost to counterattack
6. Game 6: White had slight pull, I broke the tension and self-destructed
7. Game 7: played Botvinnik again, King never found safety, really nice attack from White 

The Bg5 line definitely seems the point of attack: I don't know how to play the Botvinnik (clearly), and I got rolled in the Cambridge Springs.  In the other four games I was completely fine out of the opening in three of them, and only divine intervention stopped me from winning the fifth game.

Results of Preparation

If we want to attack me at my weakest links, here they are: play either the QGD or the ..e6 Indians as Black, aiming for a strategic game where I don't have the experience to play the resulting positions to the full.  I sometimes play 1.c4, but the QGD transposes into that fairly easily.  As White, the mainline Semi-Slav is ripe for the picking, so play 1.d4 and aim for that.

For my purposes, this presents an interesting conundrum: as White I score well against the most aggressive and dynamic openings (KID and Grunfeld primarily), but as Black I score worst against White's most aggressive and dynamic option against the Semi-Slav.

I may need to do some soul searching on this: why am I playing the Semi-Slav so poorly, especially when I really enjoyed GM Sam Shankland's course on it?  Indeed, it's probably my favourite course of his.  It might be a question of experience, it might be a style thing ... or maybe it's just a blip and will self-correct over more games.

Note that I did not do the part where I walk through my repertoire with an engine and select the absolute line I would play against myself: feel free to do that if you want.  For my purposes, I have found where I am struggling in general, and I will aim to correct this before diving into any more specifics.

Finally, I do think it's important to keep in mind the historical results. I have played the Semi-Slav exclusively over the last three years, for example, but I have played 3x as many games with the QGA and have a much better score.  "Complete" preparation against me should include looking into these lines; the last thing you want is to be an only 1.e4 player, prepare the Semi-Slav against me, and then I use the QGA, something you have literally zero experience with.  For the purposes of this exercise, though, I have zero interest in analyzing my results with openings I don't currently play, so we will draw the line here.

Conclusion

I've demonstrated how to prepare against me, and in the process, hopefully you can either use this process to prepare for your next opponent or, more relevantly, use it to pinpont your own strengths and weaknesses. Hard data is often illuminating, especially compared to our subjective thoughts.  I've really enjoyed learning the Semi-Slav, hence why I'm surprised the results have been so poor relative to everything else.  Definitely something I should look into.

Preparation and playing with statistics can be a lot of fun, but I would caution against reading too much into this.  The specific opening is only one part of the process, and general considerations (attacker vs defender, dynamism vs solidity, open positions vs closed positions, etc) play an equally large and maybe even larger part.  I will examine how to do that in the second part.