Originally posted by wormwoodOK, those last paragraphs I fully agree with though. CT gives positions and basically says "here Toga evaluates the position +1.75, find that move", and indeed, the question arises "so what, great for Mr. Toga, congratulations to him".
but take a Good Problem, one about which a human can say "now, here's the idea... and this is why this doesn't work... this looks promising, you get up material, BUT you lose the ending so it's a no go..." etc, and you're actually learning something.
we need more [b]Good Problems, and less random crap that rybka happens to evaluate +5.27. (and that goes for CTS just as well)[/b]
Originally posted by richardchesstempoChanging a few parameters probably won't do the trick for me.
Something I could look at doing would be to require a larger gap between the best and second best move for tactics that win larger amounts of material (which would then reject this problem), as I think this would address some of the concerns you have. The issue is I have to balance yield of hard problems against such concerns, so I'll need to decide if the trade-off would be worth it here.
Originally posted by wormwoodI presume you're being intentionally obtuse, as that would seem to fit.
that would mean all problems ended up into a mate. they don't.
this whole conversation illuminates perfectly why relying on engines is such a damaging idea for improving players. you engine guys don't even understand what we're saying, but instead run back to 'rybka evaluates this +5.27' every single time, thinking that represents some kind of 'truth' in the position.
You know something - people can be versed in engines AND in chess. Just think of that - wow!
A computer's evaluation in a tactical position is, generally, as good as it's going to get. No-one is pretending the computer is always going to be right on all occasions, and of course there are always going to be arbitrary lines (1.75 in this case). What of it? Do you think we're incapable of understanding a position because we also own engines?
Especially on the harder problems in CT, human feedback IS welcomed, and various high level problems have been disabled because various high-rated users have shown or said that an endgame is easily winning. No, the computer isn't always right but it's right a lot more than anything else anyone has.
With regards to your definition of tactics of there always being one completely superior line - rubbish. Take some kind of normal kingside position and all moves could be equal except there may be Nxh7 or Bxh7, both of which win to different degrees. This is still tactics, even though more than one move wins, because it's two sacrifices. CT throws out problems where there are going to be more than 3 (I think) 'winning' moves, because then it's not going to be tactics really.
But, often in tactics there is a win or an even bigger win. I do agree that it is undesirable (and I have moaned about it in the past) when there is a +11 and a +9, but firstly the problem doesn't fail you for the +9, you get to try again, and secondly such problems are in the very small minority.
Originally posted by marvellosity...and yet again you're back to 'explaining' your decisions by rybka evaluation.
But, often in tactics there is a win or an even bigger win. I do agree that it is undesirable (and I have moaned about it in the past) when there is a +11 and a +9, but firstly the problem doesn't fail you for the +9, you get to try again, and secondly such problems are in the very small minority.
and about my attitude, it was you who started calling me a troll. I've never spoken to you before, and hadn't even heard about you, but you still saw fitting to try to insult me. such things instantly forfeit the right to complain about a bloody nose.
I think this discussion is rather pointless
1-We all can safely agree that computers are not always right.
2-Computers have a high degree of correctness when the point is tacticall.
3-Tactics servers use computers to evaluate their positions and come up with a solution.
4-A small amount of these are faulty. They are removed.
Also:
wormwood doesnt like CT, due to faulty problems
richardfromchesstempo doesnt agree - he says those are removed
someone appears out of nowere and says that wormwood is not correct
heinz keeps trolling around
...and so do I
We all agree with this right? Then, can anyone tell me what the problem is, exactly?
6 edits
Originally posted by orion251)wormwood accuses Richard of "lifting up" CTS problems and adding ambiguous continuations.
I think this discussion is rather pointless
1-We all can safely agree that computers are not always right.
2-Computers have a high degree of correctness when the point is tacticall.
3-Tactics servers use computers to evaluate their positions and come up with a solution.
4-A small amount of these are faulty. They are removed.
Also:
wormwood doesnt ...[text shortened]... so do I
We all agree with this right? Then, can anyone tell me what the problem is, exactly?
2)I correct him on that this is simply wrong, and ask for evidence.
3)Wormwood keeps going with no evidence whatsoever.
4)Richard comes and corrects him, asking for evidence again.
5)He still goes on with no evidence whatsoever.
6)heinzkat does the traditional anti engine trolling. ---edit: allegedly 🙂.
7)marvellosity refutes him.
8)wormwood appears again and at least makes sense.
9)marvellosity makes more sense (in my opinion).
the only things that were worthy of discussion were
A)if tactics can involve more than 1 winning moves
B)the reliability of engine evaluations in tactical sequences.
my opinion:
A)simply, yes, and I find it more useful for improvement.
B)they are much more accurate and objective than human evaluations in tactical sequences. In 99.9% of cases of complaints in CT, it's just a matter of humans being able to comprehend them.
collect ALL positions where a human grandmaster has evaluated as "clearly winning", and then collect ALL positions where toga has evaluated as 1.75 and let the chess god play all games in both sets to the end. I don't have the slightest doubt that the latter would simply crush the former in percentage of games ending in mate. and since chess gods appear like twice in a century and when they're gone we're stuck with either humans or engines, I say go with engines.
2 edits
Point 6) that's not at all the underlying thought of what I have been saying. 😕 I ain't that bloody rascal Greenpawn you know. Point A) is what my main issue is with CT. Point 6) is more of a side-observation. 🙂
Point C) would be that the "random" positions are not suited for engine-bulldozering.
Originally posted by philidor position10) you still don't know what you're talking about. richard does, I do, you don't. so you just shut it.
1)wormwood accuses Richard of "lifting up" CTS problems and adding ambiguous continuations.
2)I correct him on that this is simply wrong, and ask for evidence.
3)Wormwood keeps going with no evidence whatsoever.
4)Richard comes and corrects him, asking for evidence again.
5)He still goes on with no evidence whatsoever.
6)heinzkat does the tra ...[text shortened]... and when they're gone we're stuck with either humans or engines, I say go with engines.
2 edits
Originally posted by philidor positionNo it does not, I do believe the engine's evaluations are as "perfect" as you can get it in such positions (for practical purposes). I am not comfortable with the way they are used on CT though. CTS has some BetaX engine that poops out pretty nice positions and "logical" move sequences. CT has a Toga engine that just poops out Toga analysis lines, right? The "best" of them is picked for the problem sequence. The user is asked to "play like Toga from a random position". That is one of my issues.
C goes back to B.