Which endgames should you practice? A statistical analysis

RB+6p vs. RN+5p are classical endgames. Lasker was famous for these. It is no middle game any more. Players must coordinate their rook with their minor piece. Players must always calculate trading rooks or trading minor pieces.

Yes RR+3p vs RR+2p is also and endgame, what else would it be? There are some famous examples by Botvinnik. Players must coordinate their rooks and must always calculate trading a pair of rooks.

Distinction between opening, middle game and endgame is artificial, as said above. The opening is usually until the pieces are developed and a pair of pawns is exchanged, typically until move 20. The middle game typically involves some threats to a king, which tries to shelter and is typically over before move 40. The endgame centers around promotion of a pawn and activity of kings and starts usually around move 40.

This was nicely presented but could be better if it was a study :)

Based on the statistics provided by the Chessgames database (685,801) games, the average number of moves is 40.04
These stats are from high quality games where the endgame is of note for study. Where the middle game ends and the endgame begins is open to interpretation. But it is safe to say it is well before move 40 in most all games, as early as move 20 for many.

When practicing endgames it makes sense to build up from 2 kings, and 1 pawn to gradually more material.

Although I read all this stuff incl. the mighty Endgame Manual by Dvoretzky I rarely get into endgames. Don't know why but sometimes none out 7 or 9 rounds, sometimes only a few. But the lions share end somewhere in the middlegame or in no-brainer endgames, say a piece ahead or down. Overall it is by far a small part of all games.

So I prefer studying the openings in combination with the ensuying middle-game manoeuvring. :D

@Sarg0n... very true for fast online games. However, between evenly matched GM's, they are likely to reach an endgame if a draw was not agreed upon earlier. It is their games that are considered worthy of study as the positions reached generally had few mistakes. The endgame can be a clash of ideas, either side thinking they have an edge, or how to draw the game when worse.
The title is which endgames to practice. Positions can be reached by amateurs that are interesting/challenging and make for a good puzzle. The Books only consider endgames by Masters as their play was of high quality. A database of endgames at lichess between players can be useful, but in determining which games to study requires more research. The OP's stats only included games over 40 moves, thereby excluding a great majority of games.

"So what should amateurs like us study? To answer this question, it seems reasonable to ask which types of endgames show up in amateur games most often." OP

Imo, this assumption doesn't make a whole lot of sense. Endgames that appear between amateurs in fast time controls, are most often "illogical" in that they appeared after many mistakes. Why study them? Endgame study, as has been suggested, is more about "theory", knowing the fundementals and not so much the study of positions reached.

Actually I was talking of my otb games. I won some tournaments this year against decent opposition but there were hardly any endgames I remember.

@mdinnerspace

I agree with you. The 40 half-moves rules is limiting, but I'm too lazy to revise that.

However, in some sense, even if the endgames reached on Lichess are often "illogical", they are still quite relevant. For example, I only play online, and mostly 15+15. So it's of great practical interest to me which kinds of endgames are reached in those types of games, even if illogical.

Obviously, if you're a GM who only plays long time controls OTB, then the sample I analyzed will be of no interest (but there already existed statistical surveys more relevant to these folks).

OP writes: ..."In amateur games, Rook endgames are absolutely dominant! If my statistical analysis is correct, amateurs should spend most of their endgame study "budget" looking at rooks, and it's not even close."

Yes. It is not even close. It applies for Masters, GM's alike. Rooks are the last pieces developed, initially begin as support behind the lines and after trades they emerge as the dominant piece along with remaining pawns. There being a single Queen, it is more likely at least a single Rook will remain. Rook endgames +p's have always been known to be the most common and the most difficult to play properly.

Ashley today in his commentary said.. "and they are going directly into an endgame" at move 15. There remained 2 Rooks and 2 Bishops for each side plus pawns. An example of the wide disparity in opinions about what makes for an endgame. Exchanges may be anticipated, but not always forced. Very possible a great many Rook + pawn or King + pawn endgames are equal in material, but clearly won or drawn. No need for "study" except to know basic principles as opposition or a Rook restricting a file or rank.