The Hardest Endgame In Chess vs Komodo Computer

IM Daniel Rensch explains how difficult the two knights vs king and pawn checkmate is during his weekly match with Komodo.


  1. It is a win indeed. I won against my computer like this. First you have to blokkade the pawn with a knight on d3 the other one you put on d2 to form a wall. rubenhogenhout – Fritz 91.Nd3 Ke6 2.Ke3 d4+ 3.Ke4 Kf6 4.Nd6 Kg5 5.Nc4 Kg4 6.Nd2 Kg3 7.Kf5 Kh4 8.Kg6 Kg4 9.Kf6 Kh4 10.Kf5 Kh5 11.Nf3 Kh6 12.Kf6 Kh5 13.Nfe5 Kh6 14.Ng4+ Kh5 15.Kf5 Kh4 16.Nf6 Kg3 17.Ke4 Kg2 18.Kf4 Kh3 19.Ne4 Kh4 20.Ng3 Kh3 21.Nf5 Kg2 22.Kg4 Kg1 23.Kg3 Kf1 24.Kf3 Kg1 25.Nh4 Kh2 26.Kg4 Kg1 27.Kg3 Kf1 28.Kf3 Kg1 29.Ng2 Kh2 30.Ngf4 Kg1 31.Ke2 Kh2 32.Kf2 Kh1 33.Ne5 d3 34.Ng4 d2 35.Ne2 d1N+ 36.Kg3 Ne3 37.Nf2# 1–0But it was poor defence of fritz. The table base says mate in 63 moves.

  2. Ok, I tablebased it out of curiosity 🙂 The start position (after Nc5+) is indeed winning for White, but it takes 65(!) moves (assuming perfect defense). These Troitzky endgames can be quite lengthy and difficult to execute.

  3. What a great study. Thanks Dan (and John B).

  4. Lmao mate in 60 when u can't calculate that shit and u just accept a draw. at least that's what I would do

  5. Regardless of Chess talk, there´s a baby crying back there. Congratulations 🙂

  6. Hey Danny do you remember me from nationals

  7. Lomonosov tablebase shows there is mate in 60 moves.

  8. I thought you couldn't force mate with two knights?

    Edit: Nevermind, I assume having the pawn on the board allows white to get into position without stalemating the king.

  9. Mate in 65 but wouldn't game be a draw after 50?

  10. Why don't you think for an hour BEFORE you publish the video if you think you can solve the problem?

  11. You got to beat Magnus Carlsen first, before you attempt this mate in 65 my dear Danny! Well you are a genius, and we all know that .. you show off, you!!

  12. Wow, I played this one out with the help of tablebases and it is winning indeed (mate in 61 from the starting position). However, stockfish doesn't even see the mate when it's already down to mate in 3 (yes, no typo, # in THREE! WTF!!!!!??????), which should be super easy for any engine to calculate in any position. It doesn't even suggest useful moves most of the time, and when it was like mate in 10 it even blundered mate a couple times by wanting to take the pawn (!). So what's the sorcery here? My guess is that it is programmed to just not even look for mate/progression, as it just assumes that 2 knights alone can't win (which would be true if it wasn't for the black pawn).
    Btw, if you were curious, the winning idea here is to use your king and only ONE of your knights to get the black king into the bottom-right corner (ie: the corner square that your d3 knight has fastest access to), which is an incredibly hard task and requires some very inhuman looking knight maneuvers. By the time their king gets to h8, you want to move your second knight to avoid the usual stalemate and put their pawn into zugzwang. Once you are there, the actual checkmate isn't too hard to see.

  13. So it appears that my impulse to check out this video with the intent to gain some actual chess insight was simply another bad move on my part.

  14. I have an engine line of this endgame on my channel if anyone wishes to see it in action.

  15. "I actually don't own a tablebase."

    That's why you use lichess, pal. That's why you use lichess.

  16. You need to swap a knight for the blocking of the d pawn. The pawns that lose against two knights are the a, c, d, e , f and h (as long as they are 3 moves away from queening!!!) Got that? Don't think of the result, first thing first, swap roles…1…kf6 2. nd6 (stopping king f5 to block the kings route) Ke7 3. Ke5! d4 4. nd3…and now begins the absolute amazing roll on how the king and lone knight force the king back. Later, one knight will swap roles. The d3 one will come out to play and the other will block on d2. To avoid the 50th move rule.

  17. do you know who byakuugan86 is? he solved this endgame with a tablebase and made videos about it, but I have no clue who he is

  18. Maybe change the title of this video to “Don’t bother to watch if you want to learn how to actually win with two knights against a pawn as you are wasting your time as I am just rambling on without showing you the actual technique.”

  19. Why is this endgame rarely mentionned as being the hardest to win?

  20. My study is can reach hardest endgame Rook+bishop(300+ move after Black get corner) vs two knights engine spend 1-2 min to find quickest mate

Leave a Reply

Your email address will not be published.