How can people think they can detect engine use?

Paradox's idea that positions have some inherent difficulty strikes me as a common sense idea, and I agree with him that's a way for experienced chess players to spot cheaters.

I mean, say you play with some 8 y.o. kid : would you expect him to be able to play a rook endgame with the same patience and skill than a veteran player ? Or would you expect any chess player to find a combo such as this one ?

One way would be to use a chess engine to rate the value of each possible move. The more moves there are that are close in value to the "optimal" move, the more difficult the position would be to assess.

If there is only one or two good moves, then that is a much easier proposition.

That's one possible way, I guess

yeah the notion of "inherent difficulty" is definitely an useful practical idea, but how do you put it into numbers?

@imirak, when there are many moves very close in value, that most likely just means that it's not a sharp position and as such there's most likely no single most optimal move. any move that is a part of a sound plan could be considered "best enough".

The more the top move changes from a increasingly deepening two ply search depth, as compared with the moves that end up scoring second best that they have to be compared to, and how much these second best moves change (don't forget all of the equal evaluations of the best and second best moves that have to be considered as you look further and further ahead), all affect the difficulty of the position evaluation for a human because, that it is more lines they have to not only look at but, to remember and then compare also.

It isn't just plausible that it can be done this way, I have already been doing it and seeing the metric values for positions.

Puzzles should absolutely be assigned a static difficulty rating. It is the only way to correctly rate them. If you really think having a consistantly fluctuating tactics trainer system that has puzzles with ratings getting shoved further and further down by repeated attempts, puzzle abandonment and bot use, then I think you should really ask yourself if it is easier for a first time solver to solve it happens to be rated at 1400 or 1500 when they first attempt it ? Does the number it constantly changes to, affect the actual difficulty encountered by the first time solver ? 

Why would you need to ? If a 1200 guy plays a 7-moves combo without blinking, do I need to know if the combo is 2200, 2400 or 2800 level to get suspicious ?

This isn't necessarily true. It can be in the case of three equal checkmates that are obvious 2 move scenarios but, when you have to look at three different lines, 8 moves (16 plies) deep that have multiple moves that are forcing, threatening, etc, before you finally see the one that wins a pawn, even though the evaluations may be the same scores at 4 (8 plies) moves deep.

You guys are missing my point entirely. You cannot say that a chess position is inherently difficult - the difficulty of a chess position only makes sense with reference to agents (chess players). You simply cannot ask the question "is it difficult to find the best move here?" without assuming that there is a chess player who will be contemplating the move. There is no intrinsic difficulty in this case - there is only relational difficulty. If you disagree with this then you simply have an incorrect notion of what intrinsic means; it means pertaining to a thing in and of itself, belonging to a thing by its very nature. Difficulty is not something that can belong to a chess position in and of itself - difficulty requires relation to an agent.

But, besides all this philosophy, the idea of measuring a chess position's intrinsic difficulty is of course absurd anyway, for altogether separate reasons. How could you measure this difficulty any other way than simply seeing how many people were able to do it? And then your "intrinsic" difficulty is actually just a relational difficulty, and you're no better off than using ELO.

Hogwash, basically.

@hicetnunc you'd need it to "argue with the authorities" to have the cheater be actually banned. :p of course, for me and my intuition to be "pretty sure" that someone is a cheater, the standards are much lower... and i mostly just keep it to myself, because i'm lazy and i don't like to accuse lightly. mostly because i'm lazy though.

@paradox that actually sounds like it could work

Inherent difficulty is intuitive and easy to understand.  Rigorously quantifying it is a waste of time.

Of course there are gray areas.  Sometimes weak players play strong moves.  But there are also black and white areas, as hicetnunc says, a weak player will not spot complex tactics in a short time (and probably not even after a lot of thought).

Out of curiosity (I don't know any philosophy really) can you give an example of some intrinsic properties present in something without an observer to make that judgement?

Of course weak players often play strong moves.

What makes them weak is that they more often play weak moves

quantifying it could be a very nice step towards new and more efficient systems of cheating detection.

i also believe that "weak" players (i use the word "weak" in a relative sense, there are some areas where i am weak and so this also applies to myself) are just completely unable to spot some things no matter how much time you give them, because they just don't have the knowledge of what they should be looking for.

Density - Intrinsic
Weight - extrinsic (or relational)

In neither case is an observer required - it's not about observers, per se, it's about having to explain or define some property of an object with reference to something else. Density is explainable by the physical construction of an object itself; weight is only explainable by reference to the gravitational force in the immediate vicinity and varies accordingly. Difficulty is only explainable through reference to an agent, and varies according to the agent's skill 

Sometimes opponents begin a game in a form related to their elo and when the puzzle gets difficult, they become super-chess players and breeze their way through complications. This is immedietly noticable for players 2000 elo and below.

Mass ? But I don't know - I'll wait for the expert's answer as you do

@spiker i guess that the word we actually are looking for is "complexity". then everything makes more sense

Whether you think it is a functional system isn't in debate for me.

Why don't you want it to work ? This is the far more interesting topic now...

...are you afraid to know if your opponent's play shows they regularly seek a complexity of position more complex than a top GM, while showing a better level of performance in those positions, yet somehow managing to have lower T-scores than the benchmarks commonly used to detect cheating, via intentionally playing weaker moves once an advantage is reach ?

My system shows all of that...

I mean really, don't you want to know when you've been cheated by a weaker player pretending to be stronger, just not so strong they get caught easily ?

Actually, yes, you're right.  Kind of a silly question I guess, sorry.

I didn't think intrinsic was a bad choice of words, but I guess you're right that it wasn't correct.

Quantifying the complexity or difficulty of finding the appropriate move in a given chess position is indeed a dead end. The only way to accomplish this is to see how many players can do this, and then see how well those players rank against each other. But I'm just describing ELO, which is the best that can be done. ParadoxOfNone has already ruled this out as a basis for his calculations. Which leaves absolutely no basis whatsoever, in my opinion.