Surprising tactic in a normal looking rook endgame that stockfish struggles to find

en.lichess.org/1Loe501D/white#50

Really happy with this one because it comes like a lightning bolt out of nowhere, and because after leaving stockfish on for a while it still can't find the best idea and just wants to go pawn grabbing. Humans FTW!

To get to the tactical position, click on black's 25th move. Surprised the embedded game doesn't automatically do that... Suggestion time.

just add #<movenumber> to the link, eg

en.lichess.org/1Loe501D#50

ok that doesnt work :)

I've got my rook in this kind of prisons... Not a great feeling :D

*SPOILER ALERT: If you want the opportunity of finding the tactic on move 25. then don't read on.*

In the Analysis board SF moves 26, 27 and 28 are the exactly the same as the moves you played when you found the tactic on 25. The moves you play and the moves SF recommend diverge on move 29. c5?! played by you and fxe5 recommended by SF. Unfortunately your opponent blunders on move 29...exf4 and so it may not be immediately obvious that the line you choose may not have been best. If you take more care you can see that the SF suggested line leads to a win while the move you played allows Black to coordinate his/her Kingside pawns. If you follow along the moves suggested (before Black blunders on 29...) you may find as I did in one of the computer recommended variations suggested instead of 29...exf4 - that white ends up in zugzwang and is forced to attack the Black rook (otherwise its a draw or worse) which frees the Black rook. The position that your *tactic* on move 29 leads to with best play is

en.lichess.org/analysis/8/p2k4/r5p1/1PP2p1p/5P1P/1K2R1P1/8/8_b_-_-

while the position the computer suggested variation leads to after *stupidly pawn grabbing* is here

en.lichess.org/analysis/8/4kp2/r5p1/p1PRP3/P1PK3P/8/7P/8_w_-_-

Now I think its clear that both positions are winning but I would like to see you argue that the *human* variation is a better position than the *SF suggested* one, because it seems the other way to me. Therefore I find it more likely that SF found your *tactic* evaluated it to be inferior for reasons that you did not foresee and decided to play something more accurate.

This is not to say that your tactic was not very good play but I don't think you have shown that it was superior to the SF moves, if you can please do. More worryingly, it seems that you really did not look very carefully into the analysis, saw that your move was *winning* and that SF did not play it and concluded that humans>SF FTW. I am all for Humans>SF, maybe one day, but I don't believe that arguing inaccurately is going to help anyone.

If I have made a mistake or you can show that the move you played was superior objectively, then I would genuinely like to see it.

Cheers!

#5, don't follow the line SF suggests for the "human variation" because it still does not understand the trap. Instead of the suggested b5?? you must play a5 as in the game, thus trapping the rook.
:: 29... e4 30. Kb3 -- 31. a4 -- 32. a5. (SF suggests moving the h-pawn with 30... h4, then realizes it achieves nothing). Try it!

@DVRazor Thanks for your reply, and thanks for pointing out my mistake. I did not play the *human* variation properly I see that now. And I do see SF's evaluation change wildly from a +2~ advantage to a ~+9 advantage in a few moves after that despite *best play* from SF. I was very wrong in how I played out the human moves!

Also @LM OhNoMyPants It appears I was wrong about how deep your analysis was, I did not see that! Apologies!

I have one last question- if the human line is indeed winning- is the computer line not equally winning or good?

Cheers!

Well, it's technically winning — passed pawns, which are close to promotion, more active rook and king, but I'd say it's not the cleanest way. The downside is you have to be alert on your opponents play when the Black rook comes to the kingside and grabs the pawns and Black tries to promote the pawns. Thankfully, White is significantly faster.
Just a typical endgame technique would be required, unlike the neat play by the OP, trapping the rook and not worrying of any kind of counterplay.

@DVRazor Thanks for your reply! I guess I am not familiar with aesthetic play but I do see your point. Especially if a human was playing either endgame, it would be easier to play the one reached by white's tactic.