Interesting position for catching out cheaters :D

I would play Rxf7 anyway, even if I do not see the winning line... A move like Rxf7 just screams to be played... Tbh most of my sacs are or unsound or very unsound but sometimes opponents see ghosts and then I win

@Sarg0n said in #36:

Three different outcomes out of many. It looks easy but I would find it highly suspicious. Those non-linear combinations are difficult for humans IMHO.
The first game where if Queen were to capture Bishop will result in an instant checkmate for White.
QxRe8. Rook can't capture Queen as Bishop is on d5 aligning the King. King can only move to g7 and then Queen f8 is checkmate.

Let's just say the Queen refuses to capture the Bishop it will have only one safe square and that is c7. Again, giving me the opportunity to capture the Rook on e8. Queen g7 would be a disaster and will once again lead to checkmate after QxRe8 giving a check and Queen has to block the check with f8 which, will once again result in checkmate.

On your second game. Bishop can now capture Rook e6 with the idea that if Queen was to capture the Bishop on e6, Rook can now capture Knight on b7. The game will now be equalized and game will be on an equal playing field.

I have not used any computer analysis for either of these scenarios. These are just common sense observations that any experienced non titled chess player will see from the game positions you've posted.

There is nothing difficult about these scenarios for any 2000+ rated chess player. They should see it instantly. In fact, I'd be over the moon if I was presented with these positions to play from. I'll know the momentum is in my hands.

@GoLdEnFLAME said in #42:

The first game where if Queen were to capture Bishop will result in an instant checkmate for White.
QxRe8. Rook can't capture Queen as Bishop is on d5 aligning the King. King can only move to g7 and then Queen f8 is checkmate.

Let's just say the Queen refuses to capture the Bishop it will have only one safe square and that is c7. Again, giving me the opportunity to capture the Rook on e8. Queen g7 would be a disaster and will once again lead to checkmate after QxRe8 giving a check and Queen has to block the check with f8 which, will once again result in checkmate.

On your second game. Bishop can now capture Rook e6 with the idea that if Queen was to capture the Bishop on e6, Rook can now capture Knight on b7. The game will now be equalized and game will be on an equal playing field.

I have not used any computer analysis for either of these scenarios. These are just common sense observations that any experienced non titled chess player will see from the game positions you've posted.

There is nothing difficult about these scenarios for any 2000+ rated chess player. They should see it instantly. In fact, I'd be over the moon if I was presented with these positions to play from. I'll know the momentum is in my hands.

Playing the outcome is not difficult. It’s the combination beforehand.

Rxf7 totally falls within my realm of moves. Mostly, because I wouldn't have considered the best defense by black. So trying to win it later on would surely put me into (even more) time trouble. It would be a move-by-move thing afterwards.

@Anon581 said in #4:

Thank you for the thinking introspection arguments, a rare thing. also thanks the op for a rational version of a thread about the cheater concept. I did not really consider the position information or the specific SAN moves chess points, but the description without the SAN was already self contained. You post was readable and having a reasoning that could fly by itself (a compliment). I might need some other time to actually make my own mind about the chess, and the various human qualities of the moves you assessed in that way.

But about the amount of thinking that you project on the hypothetical cheater. Perhaps engine manipulation out of sight of lichess automatic things is faster than blitz time scales (or accumulated emergency that position may also have as part of the position challenge information by op, neglecting that last bit). But it seems that the judgment about the a good enough same outcome more human move, than the first PV, required a human chess assessment of that. I might have collapsed part of your reasoning. I should reread. but I leave this possibility still. anyway. if wrong you would tell me.

I might have inverted which move was more human.. But in general. The ultimate cheater hypothesis, that knows as much chess or chess psychology versus engine "psychology"** seems like an arms race with some possible difficult motivation compatibility.

And the op is right there might be such positions likely to separate the behaviors.. but then, the ultimate cheaters also knows that.. and then will still find more efficient to keep learning how to cheat, the training difficulty of which might be proportional to that chess study that would have perhaps been reward all the way to rating consequence, if chess itself not enough reward.

** well that is not yet a discipline, no introspection data made public on demand per position analysis, off-topic plug, sorry, just popped up, kind of relevant though, just saying human versus not human, might need more characteristics per positions (not just game sequence of engine scoring). If we could characterize engine behavior per position with as much information as you could bring forth about human thinking, we might have more subtle classification notions, at position level.

Finding such positions. that are possibly also a possible basis for diversity metric among engines. as for from humans is one engine can serve as calibration for how far are each engine to each other.. give more meaning to engine pool based ELO measures and having a pool characterization. one could subsample the very games of the tournaments for such positions.. or develop on that idea that some positions might have more divergence of behavior. end of extrasolar interlude. Best would still be full sub-tree of node explored. but that might not be popular or socially acceptable.

edit: not just node visited.. node considered for any further exploration classification internally (bunch of move based heuristic mixed with leaf evaluation history** including the subset of those that did get any leaf evaluation function actually computed passed all the other dismissal tests (the other nodes above) in the single position tree search (root with score). This is modern exhaustive search common basic model. (unless I can get some correction, I might keep that story going... and going... we have tried to get some understanding of that from source code, I think the burden of sharing with userland is not on us. but userland should start building some critical thinking backbone about machine programs they use to shape their own chess thinking at each position... steam out of ears!)

** (pre-move ordering, move ordering axis for AB pruning strict, and many others in the iterative deepening direction, to be overly precise for those who might be bathing in there passing by here, please respond, btw).

op. sorry for the half post tangential manifesto iteration. one idea leads to another and voilà a whole mutli-point post has sprouted. thanks for the opportunity. with a concise yet interesting open question.