The endgames you actually get, and the ones you never will

The short answer

The endgames you will actually reach are rook endings, king-and-pawn endings, and single-minor-piece endings with pawns on both wings. Everything else is a rounding error in your game history.

The famous ones you can ignore for now: bishop and knight mate, two knights against a pawn, queen against rook, rook and bishop against rook. You will play thousands of games without needing any of them. If you want to learn them because they're beautiful, do — just don't file it under training.

Why nobody can hand you a frequency table

I am not going to give you percentages for how often each endgame type appears, because I do not have them and neither does anyone else who publishes them. Lichess and Chess.com don't release endgame-type breakdowns of their databases. The numbers you see quoted in books usually come from small samples of master games, which is not your games.

What you can reason from is how pieces leave the board. Rooks are the last pieces developed and the hardest to trade, because trading them usually requires an open file that both sides have contested. Queens come off in a huge number of lines — early trades, perpetual-check avoidance, forced sequences after a middlegame tactic. Minor pieces get traded constantly from move eight onward. So the material that survives to move forty is disproportionately rooks and pawns.

Pure pawn endings then arrive as a second-order effect: they are what rook and minor-piece endings turn into when one side simplifies to convert or to draw. That is why king-and-pawn technique matters more than its raw frequency suggests. You don't just need to play those positions — you need to evaluate them from three moves away, at the point where you decide whether to allow the last trade.

What our own move data says about the endgame

Across 204,371 engine-classified moves from 6,481 games analysed on Mated, 3.8% of moves played are outright blunders and another 4.2% are mistakes — roughly one move in 13 is a real error, before you count the 18.4% classified as inaccuracies.

Those errors are not spread evenly across the game. In that same sample, 9.1% of middlegame moves are a mistake or worse, against 2.3% in the endgame. On the face of it that says the endgame is where you play best.

Read it carefully before you act on it. Endgame positions contain more forced recaptures and only-moves, which inflates the denominator with moves nobody could get wrong. And there is a selection effect: games decided by a middlegame blunder never produce endgame moves at all, so the endgames in the sample are the ones where both players were still functioning.

The third number is the one that changes how you should train. In that sample the average move 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. Chess losses are not a slow leak. They are two or three moves. In a rook ending with four pawns each, one of those moves is the whole result — there is no material left to cover for you. We have not measured result-swing per error by phase, so treat that last step as reasoning rather than a finding.

The five things worth learning, in order

Start with the opposition, because it decides king-and-pawn endings and because you need it as an evaluation tool long before you need it as a technique.

White king e6, white pawn e5, black king e8. Black to move loses: 1...Kf8 2.Kd7 Kf7 3.e6+ Kf8 4.e7+ Kf7 5.e8=Q+. White to move draws: 1.Kd6 Kd8 2.e6 Ke8 3.e7 Kf7 4.Kd7 stalemate. Same pieces, same squares, opposite results depending on whose move it is. That pattern, or its mirror on some other file, is behind a large share of the pawn endings you will ever play.

After that, in this order:

• The Lucena position — how to promote with rook and pawn on the seventh when the defending king is cut off. You will get this one.
• The Philidor defence — defending king in front of the pawn, rook on the third rank until the pawn advances, then checks from behind. This saves half points you are currently losing.
• Rook behind the passed pawn, yours or theirs. One rule, applies constantly.
• Rook and four pawns against rook and three on the same wing — the most common practical rook ending there is, and usually a draw with active defence.
• Opposite-coloured bishops: know that a pawn or two up is often nothing, and that the drawing method is to blockade on your bishop's colour.

The famous ones, and why they never come up

Bishop and knight mate is the standard example. To reach it you have to have traded into exactly that material with no pawns left, which means a specific and unusual sequence of exchanges. My honest estimate is that a club player meets it once in several thousand games, and I have no measured number to offer. Fifty moves is also enough time only if you know it cold, which means the study cost is high and the payoff is near zero.

Two knights against a pawn and the Troitsky line: essentially never, and the technique is not learnable in an evening. Queen against rook: rare, and a tablebase win that strong players still botch. Rook and bishop against rook: rare, and mostly a draw with correct defence you will not remember.

Réti's study and Saavedra's are worth seeing once. They teach you that kings move diagonally with two purposes at once, which is a genuinely useful idea. They are not positions you will get.

How to actually drill this

For the named theoretical endings, you do not need a product. Set the position up against a tablebase or the engine on any site, play it from both sides until you stop needing to think, then come back to it in a week. Ten minutes for Lucena, twenty for Philidor including the short-side and long-side defences. That is most of the theoretical work you need before 1800.

The harder question is which endgames your own games are producing, and where you are throwing them away. That is not something a general endgame course can tell you, because it depends on your openings and your trading habits. Reading your last few months of games and sorting the endgame errors by type is the useful version of this — Mated does that from your Chess.com or Lichess history and builds the drills out of the positions you actually reached, but you can do a slower version by hand with an engine and a spreadsheet.

Either way, the sorting matters more than the tool. Twelve losses in rook endings where you defended passively is a different problem from twelve losses where you miscounted a pawn race.

The skill nobody calls an endgame skill

Most of your endgame results are decided before the endgame starts, at the moment you choose whether to take on d4 or leave the tension. The evaluation you need is not "can I win rook and three against rook and two" — it is "is the position I get after this trade one I know how to play".

That reframes the study list. Your openings produce specific endgames. An Exchange Slav player gets symmetrical rook endings constantly. A Berlin player gets that queenless middlegame with the doubled pawns. If you play the London, you will meet the same minor-piece structures for years. Learning the three endings your own repertoire keeps generating is worth more than working through a general endgame manual front to back, and almost nobody does it, because it requires you to look at your own games first and pick.

One practical test: next time you're offered a trade that heads towards an endgame, name the resulting material balance out loud before you decide. If you can't name it, you're not evaluating it.