How to actually get better at chess

Advice that is true at every rating is useless at all of them. What decides a game at 700 has almost nothing in common with what decides one at 1900, and the positions themselves will say so if you count them.

10,440 rated positions counted
87 skills, 16 sections
rated 404 to 2783
Free, no account needed

The whole ladder in one table

Six of the 87 skills in this trainer were filled by asking the Lichess database for a rating band and nothing else -- no theme, no motif, no filter on what the answer turns out to be. That is 720 positions, and it is the only part of the corpus that can honestly be asked what happens at a level.

Counted from the 720 positions drawn on rating alone. "Slow answer" means the winning move is tagged quiet or defensive rather than forcing.
Rating band	Positions	Ends in mate	One-move answers	Three-move answers	Fork	Slow answer
Under 800	45	58%	12	2%	13	0
800 to 1000	67	27%	7	15%	19	1
1000 to 1200	44	25%	5	20%	11	2
1200 to 1500	151	26%	14	35%	22	7
1500 to 1800	140	22%	9	45%	20	14
1800 to 2100	149	7%	1	58%	13	33
2100 and up	124	5%	0	61%	6	20

Three things fall out of it, and each one is a different answer to "what should I work on".

Checkmate stops being the answer. It ends 58% of positions below 800 and 7% between 1800 and 2100. Below 1000 you are training to see mate. Above 1500 you are mostly not.

Depth multiplies. Answers needing three moves from you go from 2% of positions below 800 to 58% between 1800 and 2100.

The winning move stops being forcing. Quiet and defensive answers go from 0 of 45 positions to 33 of 149. That change is the reason tactics training that worked stops working.

Find your band

The two questions everyone asks

How to read any number on this site. Everything is counted from the 10,440 positions in the trainer at build time. Nothing is quoted from a study, because we cannot check one, and nothing is attributed to a player who did not say it here. Where the honest answer is that these positions cannot tell you something -- how fast you will improve, what time control to play -- the pages say that instead of guessing.

The limits are stated too. No position needs more than 3 moves from the solver, which is a rule we applied when selecting them. Puzzle positions carry fewer pieces than an average middlegame by construction. And a Lichess theme tag is one label per idea, so a move that is two things at once is counted once.

Then go and do the positions

Reading about it is the smaller half10,440 positions across 87 skills, four to a lesson, about two minutes each. Free, no account, works offline.Open the course

The rest of the site

What is the fastest way to improve at chess?

Nobody can give you an honest number for speed and this site will not try. What the positions do say is where the return is at each level: mate and free pieces below 1000, indirect tactics around 1200, sacrifices and endgames at 1500, quiet and defensive moves at 1800. Working on the wrong one of those is the most common way to spend a month and not move.

How do I know what my rating band is?

Use whatever number your usual site gives you. The bands here line up with Lichess puzzle ratings, so if you mostly solve puzzles rated around 1200 that is your band regardless of what your blitz rating says.

Should I study openings to improve?

Every opening pool in this trainer starts above 1000, because that is where opening positions begin producing tactics that players get wrong. Below that there is nothing in them at your level to practise on.

How many puzzles should I do a day?

Small and daily beats large and occasional, which is why a lesson here is four positions and about two minutes. Beyond that, this page has no measurement of its own and will not invent one.