Chess Puzzles: Why Your Puzzle Rating Doesn't Match Your Game Rating

Chess puzzles train a narrower skill than most players assume. A puzzle tells you a tactic is there and asks you to find it. A real game never makes that announcement, you have to notice on your own that a tactic exists at all before you can go looking for it. That gap, between being told and noticing, is the main reason puzzle rating and game rating so often disagree, and it is the honest starting point for anything else this page says about how to train.

Lichess gives unlimited chess puzzles for free: no account tier, no daily cap. Several other Lichess database powered tools do the same, offering millions of puzzles at no cost. That fact matters for how to read the rest of this page. Nobody, including us, is going to out supply Lichess on raw puzzle count. The useful question is not where to find more puzzles. It's how to use the ones you already have access to.

What's actually free right now

Lichess's own account confirmed it plainly: unlimited free puzzles, plus a Puzzle Dashboard that breaks your results down by tactical theme. No paywall anywhere in that flow.

Chess.com runs a different model. Its free tier limits how many puzzles you can solve in a day, and unlimited puzzle access sits behind a paid membership, with the fastest timed modes such as Puzzle Rush and Puzzle Battle reserved for its higher tiers still. Chess.com's pricing page renders its plans and limits dynamically, so rather than reprint a number that might already be stale, check chess.com's own pricing page for the current figures.

ChessTempo is the other name that comes up often in this space. Its official membership page returned an error when we tried to load it directly, so current pricing couldn't be confirmed first hand. One secondary source, checked in August 2026, puts a ChessTempo premium plan at roughly 35 USD a year. ChessTempo's own pricing page did not load when we tried it, so treat that as someone else's unconfirmed number rather than a price we have verified.

A related question sits underneath all of this: when an engine flags a move you missed, was it a move someone at your rating was ever going to find? We put real numbers on that in what chess analysis can and cannot tell you.

Why your puzzle rating and game rating don't match

Search chess.com's own forums for this and you find the disagreement playing out in public. In a thread titled "Reached 2000 puzzle rating, however it's a waste of time," user POOLHALL_HUSTLER reported a puzzle rating near 2000 against a rapid game rating of only 700 to 800, and argued puzzle rating "has 0% relevancy to your chess rating or ability."

The same thread carries the opposite complaint too. Commenter TheFiancheetoGambit wrote, "I have been able to reach 1500 rapid rating with only 2300 puzzle rating," a high puzzle number attached to a comparatively ordinary game rating, the mirror image of the original post. A third poster, ppandachess, self reporting an online rating over 2400, put it simply: "Ratings in puzzles and ratings on a given time control don't correlate."

Not everyone in that thread agreed the gap means puzzles are wasted effort. Commenter Duckfest pushed back, arguing there is a strong correlation between puzzle rating and chess rating, and that a gap of 900 to 1000 points between the two is normal rather than alarming. Duckfest cited no source for that figure, so treat it as an experienced player's impression, not a measured result. Another commenter, magipi, looked at the original poster's actual games and found tactical blunders that puzzle training would directly target.

Read the whole thread and the honest conclusion is that a gap between puzzle rating and game rating is common in both directions, and its size alone doesn't settle whether puzzles are helping you. What it does tell you is that puzzle solving and game play measure related but distinct skills, exactly as the opening of this page argued: one is retrieval on request, the other is retrieval on your own initiative.

Does puzzle training actually transfer to games

The clearest evidence against a naive "more puzzles, more improvement" story comes from a 14 day head to head test run by the Say Chess newsletter. Across that window, the participants who solved the most puzzles did not improve as much as the rest of the group. Volume alone was not the variable that predicted improvement.

That result doesn't say puzzles don't work. It says how you solve them, and what you do after solving them, carries more weight than how many you get through. It's consistent with advice repeated across chess.com's own coaching forum: go slowly enough to actually calculate rather than pattern match on autopilot, and review the full solution, including the alternatives you didn't play, after every attempt rather than clicking straight to the next one.

The Woodpecker Method, explained correctly

The best known structured answer to the "how you repeat" problem is a book, not an app feature. The Woodpecker Method was written by Axel Smith, with the training method itself developed by Grandmaster Hans Tikkanen, and published by Quality Chess. The book collects a little over 1,100 tactical exercises drawn from real games, graded across roughly five difficulty tiers.

The protocol is simple to state and hard to do. You pick a set sized to your level, which doesn't have to be the whole book, and solve every exercise in it once. That first pass is cycle one, and in the authors' own example schedule it takes about four weeks. Every cycle after that, you resolve the exact same set again, aiming to finish it in roughly half the time the previous cycle took. The example progression the authors describe runs about four weeks, then two weeks, then one week, then three to four days, then one to two days, continuing to compress across up to about seven cycles with a single set.

Times above are the example progression the authors describe, not a fixed schedule. Set size and your own pace will change the real number of days.

Tikkanen developed the method himself in spring 2010 and is credited with gaining three grandmaster norms within roughly seven weeks of using it, the method's own headline result.

Does the Woodpecker Method actually work?

The case for it is mechanistic and reasonably intuitive: forced repetition of the same positions, compressed into less and less time, targets pattern recognition speed directly, which is plausibly the actual bottleneck the puzzle rating versus game rating gap is describing.

The case against it is worth stating plainly rather than as a footnote. Repeating identical positions risks teaching you the specific solution to that specific position rather than the underlying pattern, the same overfitting problem a machine learning model runs into when it memorizes training examples instead of learning to generalize. Seeing a pattern across many varied positions is stronger evidence of real learning than solving the same fixed set faster each time.

There is also a possible confound in the method's own headline result. We have not found a named critic making this argument publicly, so take it as our own inference rather than an established criticism: Tikkanen did not just solve the set, he and Axel Smith built it, and finding, vetting and annotating more than a thousand tactical positions is itself hundreds of hours of deliberate chess work that a reader of the finished book skips entirely. His three grandmaster norms in seven weeks may therefore bundle that curation work together with the cycling. The following point is on firmer ground: no controlled study isolating the cycling from the curation, or comparing Woodpecker style repetition against simply playing more games or studying master games, appears to exist. Treat it as a credible, well reasoned technique with real anecdotal support, not as a proven one.

How many puzzles should you actually do a day

Given all of that, the practical guidance is less exciting than a big number and more useful than one. Short daily sessions, on the order of five to twenty minutes, beat long infrequent binges. Go slowly enough on each puzzle to actually calculate the line rather than guessing from the shape of the position. After every attempt, whether you got it right or wrong, look at the full solution and the alternatives, not just whether you were correct. One piece of forum advice worth keeping: when you miss a puzzle, solve that same one again at least a couple more times so the pattern actually sticks, rather than moving straight past it.

How Chessdrive approaches this

Chessdrive, the app we build, is the interested party in this section, so read it as that. Chessdrive's free tier gives three puzzles a day. Against a market where Lichess hands out an unlimited number for nothing, that isn't a volume offer and there's no point pretending otherwise. Nobody is going to out supply Lichess.

What Chessdrive built instead is the structured side of this problem. The Woodpecker Method is implemented as a feature inside the app: you work through a set and the app tracks your cycles and your time per cycle for you, rather than you managing it by hand in a notebook or spreadsheet, which is closer to how most people who try the method on their own actually give up on it. Separately, Chessdrive uses FSRS, a spaced repetition scheduler, for drilling your opening repertoire, so the same underlying idea, review a fixed set at the right interval to force retention, shows up in two different parts of the app for two different kinds of chess knowledge.

Paid access is $5.99 a month or $39.99 a year, with a 7 day trial on the annual plan and a 3 day trial on the monthly one. Chessdrive is live on iOS and Android.

The honest summary

Puzzles are necessary and not sufficient. They build the pattern library you need, and done at a reasonable pace, they lower your blunder rate. What closes the gap between your puzzle rating and your game rating is not more puzzles solved faster, it's applying that same pattern library under the actual conditions of a game, where nobody tells you a tactic is there and you have to notice it yourself. A structured protocol like Woodpecker cycling, used with its real limits in mind, is a reasonable way to build that recognition speed. It is not a replacement for playing and reviewing your own games.

Last verified: 28 August 2026, against the sources cited above.