Puzzle rating vs game rating
Almost everybody's puzzle rating is higher than their game rating. Nothing is broken: the two numbers are answers to two different questions.
They are not the same measurement
Almost everybody's puzzle rating is higher than their game rating, and it is not a mistake in either number. They are measuring different things, and the four differences are all in the puzzle's favour.
- You are told there is something there. This is the big one. In a puzzle you already know the position contains a winning idea, so the only question is which one. In a game nothing tells you. Most tactics that get missed are missed because nobody was looking.
- The clock is different or absent. Given long enough, most players solve most puzzles at their level. A game gives you the time you can spare and not a second more.
- A puzzle has a right answer. Somebody has already checked that a best move exists. A middlegame with no tactic in it asks a different and harder question: what is the plan.
- Nothing else is being tested. Openings, endgames, structure, technique, nerve and the ability to not blunder for forty moves in a row are all part of the game rating and none of them are part of the puzzle rating.
So which number is the real one
Both, for what each measures. The game rating is what you would put on a scoresheet. The puzzle rating is the ceiling your pattern recognition allows: it says what you can find when you know to look. The distance between them is the interesting part, because it is the part that training moves.
What this page will not tell you. A number for the average gap. You will see figures quoted in forum threads and they are somebody's impression, not a measurement. We have no player population to measure and will not invent one. The gap is personal anyway: it is made of how much you rush, and a site cannot know that.
Why this site keeps two ratings
Reflex Chess carries two ratings and not one, because a fast pattern-recogniser and a good blitz player are not the same person and one number would make each ladder noisy with the other's results. Collapsing them would make each ladder noisy with the other's results: a run of good duels would move a number that is supposed to describe how you play a slow game, and a bad night at blitz would drag down a number about how fast you see a fork.
| Duel rating | Game rating | |
|---|---|---|
| Moved by | puzzle duels | rated games |
| A contest is | 15 puzzles, both sides racing the same set | one game on a clock |
| Decided by | solved count, then total time | the result on the board |
| Time limit | 3 minutes for the whole run | 1+0 to 30+20 per player |
| Rating system | Glicko-2 | Glicko-2 |
A duel is the missing middle
A duel sits between the two. It is still puzzles, so you are still told a tactic exists — but the clock is back, somebody else is solving the same 15 in the same order, and a wrong move costs you that puzzle and moves you on. That is much closer to the pressure of a game than a solo puzzle is, and it is why its rating is worth keeping apart from a bare puzzle rating.
The duel opponent is a computer opponent whose solve chances are derived per puzzle from the gap between their duel rating and that puzzle's own difficulty, so they stall on hard ones and fly through easy ones the way a person does. A constant rate would read as a progress bar.
How to close the gap
If your puzzle rating is far above your game rating, the missing skill is almost never tactics. It is noticing. Two things move it:
- Solve under a clock. Untimed solving trains the wrong thing: in a game you get the time you have, not the time you need.
- Ask the question when nothing is prompting you. Every move of a real game, on both sides: what is undefended, what is forced, what changed. A puzzle asks that question for you, which is exactly the help you do not get when it counts.
- How to stop blunderingThe other half of the same problem, counted from the corpus
- What is my chess rating?Twelve adaptive positions, free, no account
- How chess ratings workGlicko-2, deviation and why a result is worth what it is
Why is my puzzle rating higher than my game rating?
Because a puzzle tells you a tactic exists and a game does not. Add an unlimited or generous clock, a guaranteed best move, and the absence of openings, endgames and forty moves of not blundering, and the puzzle number has every advantage.
Which rating is my real chess rating?
The game rating is the one that describes your results. The puzzle rating describes what you can find when you already know to look, which is a real thing to measure but a different one.
Should a puzzle rating and a game rating be the same?
No, and a site that made them the same would be hiding information. They measure different skills, which is why this one carries two Glicko-2 ratings that never see each other's results.
How do I make my game rating catch up?
Solve under a clock rather than untimed, and practise asking what is loose and what is forced when nothing has prompted you. The gap is made of noticing, not of tactics.