Studies to Troll your engines with

The original one was an error. The latter one that I posted was correct.

ok, is this white to play and win?

Position #3

The engine troll factor on this one is about 9.9/10

White to play and win:

No, it's black to play and win.

Ok, with black to play this is pretty easy for engines:

33/52 02:06 2,298,282k 18,163k +9.28 1. ... f6-f5 2.Qd1-e2 Qd8-h4+ 3.Ke1-d1 Nb8-c6 4.Qe2-g2 Bc8-e6 5.Ng1-e2 O-O-O 6.Ne2-c3 Qd5-c5 7.d2-d3 Rh8-g8 8.Bc1-d2 Nc6-d4 9.Bd2-e1 Qh4-e7 10.Qg2-f2 Bf8-h6 11.Rh1-g1 Kc8-b8 12.Ra1-b1 Bh6-f4 13.b2-b4 Qc5-c6 14.a2-a3 Be6-d5 15.Bf1-g2 Bd5-c4 16.Bg2-f1 Qc6xf3+ 17.Qf2xf3 Nd4xf3 18.Be1-g3 Qe7-g5 19.Bg3xf4 Qg5xf4 20.Rg1-g2 Qf4-d4 21.Kd1-c1 Qd4xc3 22.d3xc4

A +9.28 advantage for black after 1...f5 says Stockfish, the goal here is to post positions that engines can't solve, or have a very hard time solving, in other words to "troll" the chess engines.

Position #4

The engine troll factor on this one is about 7/10

White to play and win:

It's not solvable for any human.

Solved in about 15 minutes with a little bit of help by SF. Seeing the idea is not that hard. How fast SF solves it may depend on whether or not the "fortress" software is activated. I have no clue how that is on chess.com. This one is very similar to a well-known draw study which starts with Ba4+! followed by closing the position with forcing pawn moves. Actually the hard part is not finding the winning strategy but managing the frustrating queen checks in the beginning and the end. That's the part where engines often outplay the humans!

That's just it, Stockfish doesn't solve it.

Not so sure! We are quite impatient in testing with engines, expecting results in a few minutes. And then there is the fortress software. I usually feed the engines "good human moves" before I give them a go but I also found that they find the good moves themselves with some more thinking time. Before writing off the engines we ought to give them at least 24 hours before passing a verdict.

I am always a bit surprised about how SF works for me. Letting it solve a #5 it often has trouble identifying the first move - but finds it straight away when I suggest it. Komodo does a lot better. Wonder if they aren't implementation errors in the versions we run here!

Belive me, I have let Stockfish search these positions for a very long time on a reasonably fast computer as well as other engines like Komodo, Houdini, and a few others. About the only one it will find is Position #4 so far. I should add that I am not so sure that you have solved Position #2 as you did not post your proposed solution.

I should believe you and you do not believe me? I never take the time to write up long boring solutions. I played it until SF saw the win - it does at some point - and then stepped through its analysis to verify it made progress. The critical step is capturing pawn d7. Things after that point are less interesting as a strategy.. Btw, I am a study composer and expert study solver. I have learned to compose and solve very efficiently with engines and tablebases. I now do both about 100 times faster than 30 years ago. Though 73, I am definitely a cyborg product of 21st century technology!

You did not have to post a lengthy solution, just the first few moves and then you could have pointed out the critical d7 pawn, also there weren't hardly any Queen checks in the beginning that's what made me wonder what you were talking about. I don't think it's too difficult to solve studies with the assistance of an engine, but of course there may be a few rare exceptions, the engines are however indispensable for composing studies to make sure there are no cooks.

By the way you're the one who said "Not so sure!" to me first (I'm not offended nor did I take offense). Keep in mind that I have thoroughly tested these positions with engines and in most cases for 24+ hours to make sure the engines were not able to solve them, or if they were able to solve, it didn't take just a few minutes.

From your first post I couldn't know how much testing you had done. As you know, lots of people post these weird diagrams from the web without great knowledge of their background, let alone testing. Also I saw diagram #1 is pretty close to engine range as its solution is already seen after a few moves are given. So I would expect it's solvable in 24 hrs but may be I was too optimistic. Or may be the engines improved a lot recently.

Yes indeed there are not that many queen moves in the beginning of #2 but I only saw that after I played a while with ..Qg1+ which is impossible because of Qxg1. That is precisely what makes the queen check series so hard to manage. We humans overlook so many dumb things in them like cross checks, captures and interferences. So I decided to stop there straight away to turn to the main dish which I had already identified. Gotta get that reduction to 15 minutes somehow!

I almost never post solutions to any puzzle, certainly not the long ones. I am interested in their content and I like to comment on them when there is something interesting to say. But I solve them for myself, or not. And sometimes with an engine or tablebase, and sometimes not. I use many of the ideas in them as an inspiration for new compositions. Without the chess.com puzzles I wouldn't have made a lot of puzzles in the past years. Which is amazing when you consider how low the average quality of the posted puzzles is. I distrust them so much that I briefly open many puzzles in SF to check it is not all nonsense before I start solving! Puzzle life is hard on chess.com.

Position #5

The engine troll factor on this one is 10/10

White to play and DRAW

Position #6

The engine troll factor on this one is 9.8/10

White to play and DRAW

g3

1

Good job, you solved them all

Very well stated, and nice to meet you, I hope some of these positions will help you in your composing endeavors.

The first Position is one that I used a 5 mover from 1904 by Friedrich Köhnlein which you can see here:

I heavily modified it into one that I was trying to stump engines with which became Position #1, and I of course tested it for cooks and checked to make sure a few top engines couldn't solve it. The funny thing about the original 5 mover is that most engines don't opt for the mate in 5, and instead opt for the FREE ROOK!!!! which results in a mate in at most 19 instead.