For the past few weeks I’ve been experimenting with database analysis for chess. That is, looking at big picture trends in a player’s game data to identify areas for improvement. This approach was inspired by how poker players study, so in that sense, this week’s player is particularly apt: Ike Haxton is a poker champion and recent chess convert.
In fact, when we started the project, he was delayed from sending me a file of his games because he was too busy winning a tournament FOR 1.6 MILLION DOLLARS!!!!
So it seems like Ike’s poker career is going just fine. Nonetheless, he’s taken an interest in our cozy game of chess.
Unsurprisingly, given that he’s clearly a very smart guy who knows how to study games, he’s been making fast progress. He’s currently rated in the 1800s on Chess.com rapid. But I wanted to see if I could pick out any areas to accelerate his improvement.
Openings
Ike also sent me some of his thoughts on his own game, which I’ll be referring to throughout the post. On his openings, he said:
“Openings are, by far, the part of the game where I most consistently outplay my opponents. I know this is not a particularly important strength to have, especially at my level, but I just kinda like memorizing openings and find it relatively easy.”
Aimchess agrees with this, giving Ike a 95% score for the opening. I would also agree: he’s doing well in the opening. Nonetheless, there is always room for improvement.
Ike’s repertoire consists of the Catalan with White, and the Classical Sicilian and Queen’s Gambit Accepted as Black. Overall, I like these choices.
The Catalan is extremely solid, but also dangerous for Black, which is why it’s one of the top choices at the grandmaster level. It tends to be a more advanced and subtle opening, but it seems to be working well for Ike.
The Queen’s Gambit Accepted is another opening I really like. In fact, I’d even say it’s somewhat underrated. Any time you can change the course of the game early on, that gives you a better chance of playing the game on your “home turf” and the QGA does just that by playing dxc4 on move two. It’s also just a rock-solid opening.
The most interesting choice is probably the Classical Sicilian (1. e4 c5 2. Nf3 d6 3. d4 cxd4 4. Nxd4 Nf6 5. Nc3 Nc6). This is yet another opening I like. It can lead to similar structures as the Najdorf, but it’s somewhat less common so there’s less of a target on your head.
Ike switched from the French to the Classical Sicilian. I would like to once again express my bewilderment that everyone I know seems to play the French. Won’t someone think of the light-squared bishops??
While the Classical Sicilian is undoubtedly a much sexier choice, it doesn’t seem to be working as well in practice. According to the Chess.com opening explorer, Ike was scoring 54% with 1…e6, but only 50% with 1…c5. And if we go to the starting position of the Classical Sicilian after 5…Nc6, he’s scoring only 38%. (I’m a little suspicious of these numbers but I haven’t had a chance to try to replicate them.)
So it seems that the Classical Sicilian isn’t working quite as well as we’d like at the moment. This could be because it’s a strategically complex opening that leads to positions that are tense and difficult to play.
When I mentioned this to Ike, he said that part of the reason he chose the Classical Sicilian was to force himself to play chaotic positions that he was uncomfortable in. I think this is really smart and something more players should think about. Opening choice isn’t only about maximizing your results right now, it’s also about growing as a player. So while the Classical Sicilian isn’t producing great results as measured by wins, in terms of challenging him to play different kinds of positions, it’s working exactly as intended.
Advantage Capitalization
Ike said, “The biggest thing holding me back is that my endgames are horrible. I fail to convert easily winning positions and lose from easily drawn positions way too often.”
Interestingly, Aimchess gives him a 68% score for advantage capitalization. As Ike is no doubt well aware from his poker experience, our perception of what’s happening is often very different from what’s really happening. In fact, this is one of the best reasons to look at stats: by looking at the record of what really happened you can correct biases or mistaken views about what’s happening in your games.
In this case, though, the apparent discrepancy is easily resolved. It turns out that when Aimchess gives you a 68% score for advantage capitalization, that doesn’t mean you’re 68th percentile at that skill, it means you’re converting 68% of winning positions. Aimchess also tells us that players at Ike’s rating level typically convert a higher percentage of their winning positions.
Okay, so it turns out the data backs up Ike’s sense that he is failing to convert many winning positions. What can we do to address this? He said, “I've put a decent amount of effort into studying endgames, particularly the 100 Endgames You Must Know Chessable course, but that knowledge doesn't seem to translate into executing over the board.”
I find that many of the most diligent, motivated players I come in contact with are studying 100 Endgames You Must Know and, well, I have mixed feelings. I think this is a good book/course, but it’s often misunderstood. This book covers theoretical endgames, which are endgames with very reduced material where it is possible to establish a method of correct play that can be memorized or drilled.
The thing is, these endgames rarely happen in real games. As Ike said, “that knowledge doesn't seem to translate into executing over the board.” That doesn’t mean it’s not worth learning. If you’re planning to play chess for a long time it’s probably worth learning these theoretical endgames at some point.
My issue is that many people seem to study this book expecting to get a practical boost in their endgame play, but what it actually gives you is a survey of highly technical positions. It’s probably a good idea to learn these positions at some point, but at any given moment in your chess career, they’re unlikely to be the highest leverage thing you could study to improve your practical results.
With that in mind, the title is somewhat misleading. You definitely don’t have to know all 100 of these endgames, even to reach grandmaster level. A more accurate title would be 100 Endgames You Should Probably Learn At Some Point, but this is admittedly less catchy.
Specifically for Ike’s current issue, 100 Endgames You Must Know isn’t super relevant for converting the type of positions he’s currently slipping up in. So what would I recommend instead?
Less focus on theoretical endgames and more on practical endgames. Play out practical endgame positions with a training partner and analyze the games.
For winning positions you failed to convert, play them out against the engine. There are some positions that are so winning that you want to feel confident converting them against anyone in the world. For these, you can practice them against a strong engine until you’re able to win.
Additionally, there are a few ways you can adjust your strategy to convert winning positions more consistently. I’ve actually given a fair amount of thought to this because at one point I noticed converting winning positions was a weakness in my own game. The mantra I ended up coming up for myself is Safe, Simple, Centralize.
Safe: As much as possible, keep your position safe. Ideally, make sure all your pieces are defended. Pay special attention to king safety because getting your king attacked or checkmated is the easiest way to lose a winning position.
Simple: Keep things as simple as possible when you’re winning. This lets you avoid unnecessary mistakes and makes it easier to stay out of time trouble, which is another common way to fail to convert a win.
Centralize: Everyone knows you’re supposed to control the center, but it’s not usually associated with converting winning positions. But I found that in many of the winning positions that I ended up losing or drawing, my pieces drifted away from the center of the board. It is especially dangerous if your queen gets far away from your king. Keep your pieces centralized and working together.
Here are a few examples of where Ike could have won the game by applying these principles:
Here there was a very simple winning move of d1=Q+. White must give up the knight for the queen, after which Black will be up a clean knight, and has a straightforward plan of taking the pawn on b2 followed by advancing the b-pawn to make another queen. There is nothing White can even try to stop this. In the game, Ike went for the trickier Nc2, trying to prevent White from giving up their knight for the queen. But after Nd1, which was presumably missed, the win was far more complicated and Ike ended up losing on time.
Lesson: when there is a simple win, take it.
One pattern I noticed was that Ike often got into trouble by grabbing irrelevant pawns. Here he had been playing a great game and had built up a clearly winning position. With two minor pieces and two pawns vs. a rook, White has a huge material advantage. Black has only one idea, playing f5 to undermine the bishop or create a passed pawn.
There are many ways to win, but I like the move f5, stopping Black’s counterplay and putting all the kingside pieces on defended squares. From there White could win in many ways, for example by advancing the b-pawn to create a passed pawn. But White’s winning plan almost doesn’t matter because Black can’t do anything.
Instead, Ike played Na5, moving a piece away from the center to go after an irrelevant pawn. After Black played f5, the position was still winning, but much tricker. The game ended in a draw.
One point I want to highlight is that when it comes to converting winning positions, the engine will often lead you astray. For example, it thinks Na5 is okay because after 1. Na5 f5 2. g5! Rxh5 3. Nxc6 White is apparently still winning with various passed pawns and a knight against a rook. But to sacrifice your bishop to get to this position when just a few moves before you were winning easily would be tough to do.
From a human perspective, it’s much more important to focus on the move when the win stopped being simple, not the move when the position stopped being winning according to the computer. As long as there is at least one winning line, even if it’s incredibly complex or difficult, the computer will still consider the position winning, but you don’t want to put yourself in a position where you have to find that line if there was an easier way.
Another way to put this would be that if you need to find difficult, only moves under time pressure to win, it’s hard to see how you could win much more than 68% of the time. But if you navigate the game so that such moves are unnecessary, you can win much more often.
Time Management
“Somewhat relatedly, my clock management is also bad so I'm frequently playing the endgame under time pressure.”
Whereas the previous two players I looked at in database reviews were playing primarily 10+0, Ike is playing most of his rapid games at 15+10. Having an increment changes the math somewhat, but my basic time management philosophy is still the same: spread your time out over the course of the game and try to stay on schedule.
To calculate how many seconds you can afford to spend per move in a time control with increment, you take the starting time in minutes, multiply by 60 to get seconds, divide by 40 (there are about 40 moves in the average chess game), and add the increment.
So for 15 +10 that would be:
(15 * 60 / 40) + 10 = 32.5
Overall, you can afford to spend a little more than 30 seconds/move over the course of the game. Ideally you play a little faster than that in the opening to build up a buffer, and you want to avoid falling behind that schedule as much as possible.
As with the other players I’ve analyzed, I plotted Ike’s average time per move against this idealized schedule. (The orange line is the schedule, the blue is Ike’s average time left on each move over a sample of games.)
We can see that he’s building up a buffer in the opening, even using the increment to go above his starting time; but then spending a lot of time and getting behind schedule in the middlegame around moves 20-30. So far, Ike is the first player I’ve looked at whose blue line drops below the orange line, and I would say he should probably aim to speed up a little bit.
But this advice isn’t very helpful without looking at the positions where he’s spending the most time. If those turn out to be critical positions where he’s struggling to find a move to stay in the game, then forcing himself to play faster in those spots would not necessarily help, and could even hurt his results over all.
GM Gregory Kaidanov has a “15 minute rule” that says you should never spend more than 15 minutes on one move. The idea is that there are generally diminishing returns to spending more time, and once you’ve been thinking about a position for 15 minutes, you’ve seen most of what you’re going to see, and you should make a decision and save your remaining time for another move.
Of course, when the time limit for the whole game is 15 minutes, a 15 minute rule doesn’t make a lot of sense, so I generalize it to a 10% rule: don’t spend more than 10% of your total time on a single move, except in very rare circumstances.
Factoring in increment, 10% of your total time in a 15+10 game is about 2 minutes, so I looked for moves where Ike spent more than 2 minutes on a single move to see if there were any patterns. What I found was that in many cases he was facing a direct threat.
When your opponent threatens something in a chess game, it’s kind of like facing a bet out of position in poker: it’s an emotionally uncomfortable situation, but not necessarily a bad one. In fact, it can even be an opportunity, because if your opponent is straining to create threats, they may stick their neck out too far and give you a chance for a devastating counterattack.
So getting better at responding to threats involves both developing the analytical skill to find the best defensive moves, but also an emotional comfort level in those situations. Of course, these tend to go hand-in-hand, because as you develop your skills you become more confident.
Unfortunately, exercises to work on defensive skills are harder to find than those that work on offensive skills. Puzzles tend to focus on knockout blows with little or no examples of defensive resources, but in a real game, defense is at least as important as offense. There are, however, some resources for working on defensive tactics, such as The Next To Last Mistake by Jeremy Kane.
In terms of finding that emotional comfort level, a good first step is to take a deep breath. Don’t give your opponent too much credit: many players assume that if their opponent plays a threat they hadn’t anticipated that it’s a crushing move, but it could just as easily be a blunder. Then calmly work through the candidate moves.
In Ike’s notes to his OTB games, he mentioned several moments when he spent 20 minutes or more calculating a line. Interestingly, whereas the long tanks in his online games seemed to occur most often in situations where he was under pressure or facing a threat, in the OTB games it was the opposite: he often burned a lot of time on the clock looking for a knockout blow or aggressive sacrifice. I’m not sure what to make of this difference between online and OTB, but in many cases, the effect was the same: running out of time to make critical decisions later in the game.
Perhaps more importantly, these long calculation binges often weren’t in service of winning the game. Here’s one example:
In this position, Ike spent some time calculating the sacrifice Nxd6. (Actually, in his game notes he mentioned that he spent more time on the follow-up after Nxd6, but bear with me…) To me, Nxd6 doesn’t look very appealing. When you sacrifice a piece, you usually want to be able to control the game with forcing moves, but after 1. Nxd6 Qxd6 2. Nb5 Qd8 3. d6 (the game continuation) Black has a lot of options to try to consolidate their extra piece.
More importantly, you don’t have to sacrifice a piece! With the simple 1. Nb5, White can get just as strong of an initiative, but without giving any pieces away. And Nb5 is a move you can play in two minutes.
One poker concept I like to think about with regard to chess is the idea of expected value (EV). Expected value basically means, if I played out this situation many times, how many points would I expect to score on average? In chess, expected value would be calculated as the odds of a win, plus one half times the odds of a draw. Or if you’re not a math person, you can just think of it as the decision that gives you the best chance to win or at least draw.
If you play Nb5 quickly, you have a great position, equal material, and plenty of time on the clock. If you play Nxd6 after a long think, now you have a risky position (there is a very plausible scenario where your opponent consolidates and beats you with the extra piece) and much less time. The sacrifice may end up working, but it’s clearly a lower EV scenario.
Ike’s self-assessment on his calculation was, “I think my calculation is decent for my level, if fairly slow.” I think this is just about right. The problem isn’t really Ike’s ability to calculate, it’s when and how he’s choosing to calculate. Remember that calculation is just one tool in your chess toolbox. It’s like a hammer: you may know how to use a hammer, but if you use it to open your front door, that’s not gonna end well for you. The problem I saw in Ike’s games wasn’t so much miscalculation, as spending a lot of time calculating weird lines in situations where that wouldn’t help him win the game.
This problem can afflict very strong players. As GM Eugene Perelshteyn said in our interview about preparing for tournaments:
“I realized in some complicated position, 15 moves deep – or rather thinking for 15 minutes, doesn't matter if it's five moves deep or 10 moves deep – after I get out from that thought process and I looked at all the candidates, I am not fresher nor do I know more about the position than I began with.”
Here is how he deals with that situation:
“And what I typically do when I get kind of lost in the labyrinth of different variations, I use this concept of comparison. So I get a sense which moves give my opponent less counterplay and which positions sort of narrow down my opponent's counterplay. And I gravitate toward those versus the ones that give my opponent more counterplay.”
In other words, be clear about your alternatives and keep it simple.
Conclusion
As is so often the case, everything is connected. After reviewing Ike’s advantage capitalization, time management, and calculation, it seems to me that his issues in those areas are all related. He’s doing a lot of things right, but by spending a lot of time calculating in spots where it doesn’t increase his EV, he’s making the game harder than it needs to be.
I would suggest instituting a rule: never spend more than 15 minutes on a single move in a classical game, or 2 minutes on a single move in a 15-minute game. Are there some spots where you’d be justified spending this much time? Of course, but I find a hard and fast rule is easier to implement than one that has more wiggle room. The point isn’t to jump directly to an optimal strategy, it’s to make an adjustment that gives you a feel for a different way of approaching the game.
Additionally, whenever you catch yourself going down a rabbit hole of calculation, remember to ask yourself, “Is there a simple solution?”
When you have an option that is both simple and good, the threshold for a complicated option to take precedence is very high indeed. The complicated option has to be much better to justify the additional time and risk involved in going for it.
In one sense, chess is terribly complicated. If you try to play out a balanced middlegame against the engine, you’ll get outplayed inside of five moves, often in a deep way that is hard to understand. But in another sense, winning at chess against most people isn’t that complicated. If you can avoid blunders, manage your time sensibly, and spot big tactical opportunities, you can beat 99% of human players.
Keep it simple!