The 6 tips for getting from 400 to 3200 puzzles in 5 months as a 200 blitz player

Some people have asked why I am so low rated in games, relative to my tactics. As of this writing, I'm 3209 puzzles, 813 rapid, and 185 blitz. I'm so bad at live games, I haven't played in several weeks.

Basically, I have fairly poor board vision and intuition, but have lengthy mental checklists, and decent calculation. I've also developed some tips and habits for myself to apply to puzzles specifically, some of which I'm sharing below.

#1: Understand how puzzles score

Time bonus: If you solve a puzzle quickly, you can get a substantial time bonus, as you probably know. But if you take too much time even an hour or more, you'll always get at least +5 anyway for solving the puzzle. Although it depends on the target time for the puzzle, generally, if you've already spent 1.5 to 2 minutes on a puzzle, you've likely lost your time bonus anyway, so take your time on that puzzle.

Personally, I'm extra cautious on puzzles, haven't gotten a time bonus in weeks, and still maintain this rating.

Puzzle difficulty: The puzzle immediately after you get one wrong will usually be an easier one. Conversely, if you're on a winning streak, the puzzles will tend to increase in difficulty. This algorithm is probably to pull you out of excessive tilts and streaks.

You can use this tendency to your advantage by anticipating the difficulty of the current puzzle. If you're on a nice win streak, the next puzzles will warrant caution. Conversely, if you just lost a puzzle, the next one may be a gimme that will help you conserve your energy.

#2: Use the tools

Arrows: Draw arrows, and highlight squares on the board as needed. There are four different colored arrows, and four different colored square highlighting too. You draw these with the right mouse button, with different colors selected by depressing shift, ctrl, or alt.

#3: Determine the purpose

Material count: Count the relative material on the board. If the sides are even, the puzzle may be to win material.

If you're way behind, it may be to checkmate, or queen a pawn.

If you're ahead in material, it's often a defensive puzzle, where you're challenged to just hold on to your advantage despite an enemy onslaught.

#4: Notice the clues

Steering: Because all puzzles have only one right answer, there's often some subtle "steering" in the way the board is arranged, to ensure that you're given only one good path. This can sometimes provide a clue.

For example, if a bunch of pieces happen to be arranged so that your queen has only one possible back rank check, that might be a clue that the puzzle wants to steer you towards the back rank. On the other hand, if your queen has a choice of 2 different back rank checks, and they both have absolutely the same result, then they're both wrong, or you're wrong about them both being equivalent.

First move: A puzzle always starts with your opponent making a move. I wish I could tell you that this first move is always a blunder that provides some clue, but I've unfortunately found that is not always the case. Sometimes it is, but I've also found cases where the same tactic would have worked if the opponent had not made that last move.

#5: Record what you learned

Mistakes: If I feel that a puzzle has taught me something important, I take notes. If I encounter a puzzle where I get something wrong, or even almost get something wrong, that I feel I shouldn't, I record what I learned in a document, and when possible, a brief remediation plan or tactical tip. When I started a few months ago, my tips were quite dumb, like "Remember that queens can give check diagonally, idiot!", which over time became more concrete and refined, like "Draw arrows for all possible checks when starting a problem".

#6: Game the rating by taking time

Obviously, the difficulty of a puzzle is subjective. The given rating of a particular puzzle apparently depends on how past solvers had performed on that puzzle. Therefore, the rating of a puzzle presupposes that solvers in the future will approach that puzzle in the same way as solvers in the past. And if you happen not to fit that mold, a puzzle may turn out to be much harder, or much easier for you than its rating indicates. It's possible to take advantage of this, if you care to.

You see, 3xxx rated puzzles do not necessarily involve lengthier or more difficult calculations than 2xxx rated puzzles. Instead, a puzzle's higher rating is probably often due to players instinctively falling for a tempting looking red herring, which some superficial calculation would refute. A lower rated puzzle, on the other hand, may require difficult calculations to see the underlying idea, but happens to have forcing or tempting moves that everyone is attracted to. Such a puzzle may only require you to play the prelude to a 12 move checkmate that the puzzle did not require solvers to complete or understand.

Therefore, taking time to calculate a puzzle to evaluate the consequences of your moves, tends to negate the difficulty difference between 2xxx and 3xxx puzzles. Therefore, when I got to 2000, 2500, 3000, and so on, the apparent difficulty of the puzzles to me only changed a little.

Although a disadvantage of the way I do puzzles is that it takes a long time. As I've taken more notes, my checklists have grown larger. As I do higher rated puzzles, there are more branches to go through one by one, and more arrows to draw. At this point, I average about 10 - 15 minutes per puzzle, when it used to take me under a minute when I started. This is about as much time as an entire Rapid game, which is why my tactical techniques are not very applicable to my games.

So I encounter 3100 rated puzzles that I think are easy, because although they have a lot of false leads, one of those branches is a perfectly calculatable mate in four. On the other hand, I've come across 1500 rated puzzles that are totally beyond my understanding.

BTW, there was another thread about correlation between puzzle and game ratings. Here's a poll I did of some people who I knew were recently active with puzzles. You can see from the trendline and r-squared of 0.5 that there is a significant correlation between the two. It would be even stronger if my data point - the one on the bottom right, weren't there.

According to the regression line, someone who is 2000 puzzles, for example, would be predicted to have a rapid rating of about (2000 * 0.529) + 216 = 1274.

And someone with a 1000 rapid rating would be expected to have a puzzle rating about 1482.

Play more!! Having a high puzzle rating is good, but if you can't apply those techniques to actual games, then they're practically useless. You shared some interesting points, but to me it's all about application.

I use chesscom's puzzles as a quick warmup before my games, because they're random in terms of the pattern recognition. You never know what pattern you're going to get next. I take a few seconds to see the position, and then I just play what I see.

For more serious tactics training, like what you do on chesscom, I use CT-Art 4.0's mobile app. I don't take notes or anything, but I do take my time. They organize their puzzles by what type of tactic they are, so you learn the individual types of tactics and then get to repeat them over and over again. This is good for burning the patterns into your skull, so it becomes easier over time to spot tactics in game. But to recognize tactics in game, you have to play!!!

Cannot recommend listudy.org blind tactics enough. The position will not move at all until you’ve finished the entire line. It’s great for visualization.

We miss you at the club!! You should come back

These are good points, thanks - I'd like to be able to drill by theme, and chess.com's puzzles have such inaccurate theming. I'd like to be able to drill tactics by type.

To illustrate with an example, here's the puzzle that today happened to put my rating above 3300 for the first time. This puzzle (https://www.chess.com/puzzles/problem/2086634) is rated 3356. You can try it out for yourself, and I think you'll agree that it's very doable, and that the ideas in it are fairly elementary.

Yet, it has a pass rate of 14%. Why? I think it's the third move (below) 3... Qxc7, grabbing white's passer, which when the solver thinks about it for a moment, would agree is necessary. But at the same time, the position dangles some other superficially juicy (but losing) candidate moves for you to consider, like 3... KxR taking a free rook, or 3... Qe4+ forking and winning a bishop. But no, don't take the rook or the bishop, take the pawn.

Below are some notes I wrote on my thinking as I was working through the calculations.

Notice also how I highlight capturable pieces and checks for both sides after each move. When calculating, I also highlight the positions in the line I'm calculating in my head for each major piece, so that I don't fall victim to ghost pawns and other visualization mistakes from just holding the calculations in my head.

I took 33 minutes on this puzzle, which is very long even for me. It's partly because I knew this puzzle would put me over 3300, so I was double-checking all calculations.