Elo Maxing works through the motifs one at a time, from forks and pins to deflections, back-rank mates and rook endgames, and keeps a skill graph of how reliably you solve each. It serves the next puzzle from whichever motif you are weakest at, so practice goes where it is worth most instead of wherever the random puzzle lands.
There are 40 motifs. Each one carries its own rating, seeded at your puzzle rating the first time you meet it, and every solve moves that motif's number rather than only your overall one.
While a motif is still new the server keeps returning to it, finishing one before opening another, so the picture fills in a track at a time rather than everywhere at once. That takes about eight sessions, and by the end of it all 26 have been measured.
After that the queue splits. Three puzzles in four come from your three weakest, least recently served first. Three rather than one, because drilling only the single worst motif turns a session into the same pattern forty times over. The fourth comes from everything else, oldest first.
That last quarter is what keeps the whole graph live. Every motif goes on being tested, so each rating stays a current reading rather than a snapshot from weeks ago, and every one of them keeps tightening as you solve.
A motif is scored on clean solves: first move right, no hint. That is a harder bar than simply finishing a puzzle, and it is the one that reflects recognising a pattern rather than working it out. A track stays unscored until you have attempted it three times, so an early lucky run does not read as mastery.
A rating with a question mark beside it is provisional. The number is there, but it is still moving. It settles after roughly eight attempts at that motif, and hovering the mark tells you how many of them you have done and about how many are left.
Alongside the motifs there are four calculation tracks, for one move, short, long and deep combinations, rated on exactly the same scale. They answer a different question. A motif asks whether you know the pattern; these ask how many moves ahead you can still hold it.
Every puzzle has a depth, so unlike a motif these are scored on every single solve rather than only when they are the thing being served. They fill in about ten times faster and are the first part of the graph worth trusting, usually settled inside a couple of sessions.
They are measured, never aimed at. Half of all puzzles are short combinations, so serving more of them is not aiming anywhere; the value is in the reading. Being 1800 at forks and 1450 at deep combinations says something no single rating can: you know the pattern, and you stop seeing it at six ply.
A wrong move ends the puzzle. The answer is played out on the board and the next position loads. There is no second attempt and no escalating hint, because the first mistake has already settled the result: it is what the motif is scored on, and dwelling on a puzzle you have lost only costs you the session.
A session is ten puzzles. Ten because coaching guidance lands between six and twelve carefully solved positions in one sitting, and because grinding a hundred quickly is worse than twenty with attention, because pattern recognition degrades as you tire.
At the end you get a medal, scored on how many you solved clean rather than on rating gained: nine of ten takes Gold, seven takes Silver. Rating is a poor measure here because a correctly rated player nets close to zero by definition, so it would make the medal either trivial or unreachable. Clean count is the part you control, and the difficulty is already matched to you.
Every finished session moves you one level. Levels are grouped into eight tiers, Ember through Spark, Flare, Lantern, Beacon, Halo, Aurora and Zenith, and the tiers lengthen as you climb: Ember is four sessions, Zenith is fourteen. Sixty-four levels, six hundred and forty puzzles, from the first spark to the summit.
Past the summit the names begin again and the cycle number rises. Ember II sits above Zenith I, because the level underneath it is higher and the level is the part that never resets. The road goes on being drawn as far as you climb it.
Elo Maxing improves by accumulation. Every puzzle is recorded in full: the motif it belonged to, whether you solved it clean, how long you took, and what it moved, so each solve is another reading rather than a one-off score.
Early on the picture is broad: every motif available at your level measured a few times each, enough to rank but not to trust in detail. As the count builds, the ratings tighten, the question marks come off, and the ranking stops shuffling on a single miss. The aim gets better with it: the weakest three become the genuinely weakest three rather than the three that happened to go badly on Tuesday.
And it keeps moving. Bring a motif up and it leaves the focus, the next weakest takes its place, and the practice follows. The longer you use it the sharper the reading and the better aimed the work, which is the whole point of keeping a graph rather than a single number.
Motif availability depends on rating. Pins are rare in beginner puzzles and common in master ones; back-rank mates are the reverse. Tracks only appear once there are enough puzzles at your level to train them properly, and more open up as your rating climbs.
Elo Maxing reads and writes the same puzzle rating as Puzzle Training and the Daily Puzzle. One number across all three, not a separate score for this mode. Solving here moves the rating those pages show, and a rating earned there arrives here already settled. The 26 motif ratings sit underneath it and are Elo Maxing's own.
Puzzle Training serves random puzzles at your rating. Nothing it learns about you changes what you see next. Elo Maxing measures each motif separately and picks the next puzzle from the result, so the two differ in what decides the puzzle rather than in the puzzles themselves.
A motif is a pattern: a fork, a pin, a back-rank mate. A calculation track is how far ahead the solution runs, from one move to deep combinations. Motifs are what the next puzzle is chosen from; calculation is measured on every solve, because every puzzle has a depth whether or not it was the thing being trained.
No. It uses your Lumichess puzzle rating, which starts at 1500 and moves as you solve. Your Chess.com and Lichess accounts are not involved.
A new rating is uncertain, so it is allowed to move further per puzzle until it finds your level. As it settles the steps get smaller on their own. The same solve that moved you fifty points in your first session moves you under ten once the rating is established.
That the rating is still provisional. It appears next to your puzzle rating and next to any motif that has not been attempted enough times, and it comes off once the number stops moving much.
No. You can leave at any point and the solves already logged are kept; the session picks up where you left it. The medal is only awarded once ten are done.
Yes. A motif is scored on clean solves, and a hint means the solve was not clean. It still counts as a puzzle in the session.
Yes. Every solve moves a rating that belongs to you and the whole graph is kept against your account, so there is nowhere to put a session otherwise.