Chess.com vs Lichess: the one difference that matters

The short answer

Neither platform will make you better. The only axis on which they differ for improvement is friction — how many clicks and how much money sit between finishing a game and seeing the move where the position turned. On that axis Lichess is ahead, because unlimited engine analysis is free and always has been. On everything else that actually matters, the two are close enough that switching will not move your rating.

If you want the practical version: play wherever your opponents are, analyse wherever it is free, and stop treating the choice as a decision worth a week of thought. The interesting question is not which site. It is what you do with the twenty games you played last month.

What is actually costing you games

Across 193,204 engine-classified moves in 6,167 games analysed on Mated, 3.8% of moves played are outright blunders and another 4.2% are mistakes. That is roughly one real error every twelve moves, before you count inaccuracies, which run at 18.5%. In a 40-move game you are making three or four moves that a stronger player would not make, and around seven more that are merely loose.

Those errors are not spread evenly. In the same set, 9.1% of middlegame moves are a mistake or worse, against 2.3% in the endgame. The middlegame is where the damage happens, by a factor of four.

One more number that changes how you should read your own analysis. The average move in that set gives away 94 centipawns against the engine's best, but the median is far lower, because a small number of very bad moves carry most of the damage. Your games are not a slow bleed. They are three or four cliff edges with a lot of flat ground in between.

That last point is the one that decides the platform question. If most of your loss comes from a handful of moves per game, you do not need a subtle, feature-rich analysis suite. You need to find three moves quickly, for free, every day. That is a low bar, and both sites clear it — one of them just clears it without a subscription.

Where the two sites genuinely differ on friction

Lichess gives you server-side computer analysis on every game, unlimited, at no cost, plus an opening explorer over its own database and a studies tool for keeping notes. Chess.com's Game Review is metered on a free account — at the time of writing it is one a day, but both sites change their tiers, so check rather than trust this sentence.

Chess.com's advantage is the pool. More players, more games at any hour, more events, and a rating that most of your friends also have. That matters more than people admit, because the input to any improvement process is games you actually played, and you play more when opponents are instant.

Both expose a public API that lets you pull your own game archive as PGN. So neither site locks your data in. If you ever want to run your own engine over your last 200 games, you can, from either account.

• Lichess: unlimited free analysis, opening explorer, studies, open source, smaller pool.
• Chess.com: bigger pool, faster pairings, metered free review, more polished lesson content.
• Both: public API, PGN export, puzzles drawn from real games, no lock-in.

A worked example of what you are looking for

Take a real pattern rather than a list. You play a Four Knights, everything is fine for twenty moves, the engine has you between +0.3 and -0.2 the whole time. On move 23 you play a natural-looking rook lift, your opponent has a knight fork you did not see, and the eval goes to -3.4. You lose forty moves later after a long grind.

The temptation is to analyse the grind. That is where it felt like you lost. But the centipawn loss from moves 24 to 63 might total less than the single move 23. That is exactly what the median-versus-average gap above describes: one move carried the game.

So the review you want takes ninety seconds. Sort by centipawn loss. Look at the worst move. Cover the engine line and ask what your opponent's last move threatened. Nine times in ten at 800-1800, the answer is a capture, a check or a fork you did not look for. Write down what the threat was, not what the engine's improvement was. The improvement is a fact about that position. The missed threat is a fact about you.

Both platforms let you do this. Lichess lets you do it on every game without thinking about a quota, which is why the practical answer to "chess com vs lichess for improvement" is usually "play there, review here" — and the two accounts cost you nothing.

The part no platform does for you

Finding the move is fast. Not repeating it is the slow part, and neither site is built around that. You look at the fork, you nod, you close the tab, and three days later you hang a piece to the same geometry. Recognition on a board you are told to look at is not the same skill as noticing in a game where nobody has told you anything is wrong.

This is the gap Mated was built for. It reads your recent games through the Chess.com and Lichess APIs, runs Stockfish over every position in your browser, and builds a fifteen-minute session out of the positions you actually got wrong, so the fork you missed on Tuesday comes back on Thursday when you are not expecting it. The point is repetition of your own errors rather than someone else's puzzle set. You can build the same thing by hand with a PGN export and a spaced-repetition app, and some people do.

A caveat on the endgame number

The 2.3% endgame error rate is tempting to read as "endgames do not matter at your level." It does not say that. Fewer positions reach an endgame at all, the branching is narrower, and a mistake in a rook endgame is often instantly fatal in a way that a slightly loose middlegame move is not. A low error rate per move can still mean a high rate of games decided.

What the number does support is an allocation argument. If 9.1% of your middlegame moves are a mistake or worse, and you spend your study hour on Lucena positions, you are working on the smaller pile. That is a claim about where the volume is, not a claim that endgame study is wasted.

Nobody has measured the platform question properly, as far as I know. There is no study showing Lichess players improve faster than Chess.com players, and there could not easily be one — the populations self-select and the confounds are enormous. Anyone telling you one site's users gain rating faster is guessing.

How to settle it in two weeks, empirically

If you still care, run it on yourself. Play twenty games on each site at the same time control. Export both sets and check two things: how many of the games you actually reviewed, and your blunder rate per game across each set.

The review count is the number that will differ, and it will differ because of friction, not because of chess. If a quota, a paywall or an extra click is enough to stop you reviewing, that is real information about your habits and worth more than any feature comparison.

The blunder rate probably will not differ much, and if it does, twenty games is far too small a sample to tell you why. But you will have measured something about yourself instead of reading another opinion, and the measurement outlives the platform argument.