Good at Puzzles, Bad at Games? Train on Your Own Mistakes

You have a 2000 puzzle rating. You can spot a knight fork in three seconds when the app hands you one. Then you sit down for a real game, play fifteen normal-looking moves, and hang your queen to a tactic you would have solved instantly if someone had labelled the position "white to play and win." If that sounds like you, you are not broken and you are not imagining it. There is a specific, fixable reason good puzzle solvers still lose games, and once you see it you cannot unsee it.

Puzzles tell you a tactic is there. Games never do.

Every puzzle comes with a hidden instruction stapled to it: "there is a win in this position, go find it." That one piece of information does most of the work. You already know to switch into calculation mode, so you hunt until you find the forcing line. A real game hands you no such label. The hard part of a real game is not calculating the fork once you are looking for it. It is noticing that a fork is even available, on a move that looked quiet, in a position where nothing told you to look.

Random puzzle trainers drill the calculation and quietly skip the noticing. So you get very good at the half of the skill that was never really your problem. This is exactly why puzzle rating and game rating drift apart. A high puzzle rating means "when told a tactic exists, I find it." Winning games requires "I notice the tactic, mine and my opponent's, when nobody says a word." Those are different muscles, and only one of them shows up on the scoreboard.

Random puzzles are calibrated to your rating, not your mistakes

Here is the second problem, and it is quieter. A generic tactics trainer serves you puzzles tuned to hold you near a 50 percent success rate. That makes a pleasant difficulty curve, but it has nothing to do with the specific things you personally get wrong. Maybe you have a real blind spot for back-rank weaknesses, or for your opponent's checks right after you castle, or for a piece that gets overloaded defending two things at once. The random feed does not know any of that. It drops a back-rank puzzle on you once every few hundred, the same rate everyone at your rating gets.

So your actual leak, the pattern that quietly loses you real games again and again, gets trained at random, if it gets trained at all. You could solve puzzles for months and barely touch the one habit that is costing you the most rating.

What random tactics ARE good for

None of this means tactics trainers are useless, and you should be suspicious of anyone who tells you they are. They do one thing genuinely well: they build your raw pattern vocabulary. If you have never seen a smothered mate or a windmill, you have to meet them somewhere, and grinding themed puzzles is a perfectly good way to load those shapes into your head. For a true beginner with no pattern library yet, a healthy volume of varied puzzles is exactly the right tool.

The trouble starts after that. Once you know what the patterns look like, generic puzzles hit a ceiling fast, because the thing holding you back is no longer "I do not know what a fork is." It is "I did not see the fork in my own game." No amount of puzzles that announce themselves will train that.

The fix: train on the mistakes you actually make

The highest-leverage tactics training is almost embarrassing once you name it: your own losses. Every game you lose contains at least one position where you went wrong, and unlike a random puzzle, that position is real. It came out of your openings, your style, and your blind spots, and it came from a moment where you genuinely did not see the tactic. Redoing that exact position trains the noticing, not just the calculating, because you are re-entering the situation you actually failed in.

Two things make it work:

• Relevance. You are drilling the patterns costing you rating this week, not a generic curriculum aimed at the average player.
• Repetition under real conditions. You saw the position, you missed it, and now it comes back a few days later and you catch it. That is how a blind spot actually closes.

You can do this entirely by hand. Go through your last ten losses, find the move where the evaluation swung, set the position up on a board, and replay it until the right idea is automatic. Then do it again next week. It genuinely works. It is also tedious enough that almost nobody keeps it up, which is the whole reason we built KingChess.

How KingChess does it for you

KingChess connects to your Chess.com or Lichess account, finds the exact moves where you blundered or missed a tactic in your recent games, and turns each one into a puzzle. Then it brings those puzzles back on a spaced repetition schedule, so the position you missed on Monday returns on Thursday, and again next week, until you stop missing it. You are not solving random puzzles that quietly announce a tactic is there. You are re-solving the ones from your own games, the ones where nobody told you, until the pattern sticks.

If you want the bigger picture on what actually costs you games below 1400, we broke that down in why do I keep losing at chess. But the short version is this: being good at puzzles and being good at chess are not the same skill, and the fastest way to close the gap is to stop practising on strangers' positions and start practising on your own.