1.g4 opening is losing?

[Zenmastur]

But, as we all know, SF (or any other engine) is not perfect (very, very far from perfection IMO), so it does not matter what the SF output is or isn't.

If you actually believe that why do you bother with posting on this board. If you can't believe anything any engines says, why are all these people bothering to write these engines and then wasting time play matches with them and posting on this board?

And just getting SF to announce mate in X is obviously a useful indicator in a practical game, but does not prove a mate in the way a TB does. I've often watched SF announce mate in 30, and then it keeps calculating, and suddenly it's gone to mate in 50, and then it keeps calculating, and it retracts any mate announcement. That means the mate in 30 was (according to its longer-calculating self) just wrong. I'm sure others have seen this behavior too.

Yes, I've seen that happen as well. But when you take a line of play that terminates in mate such as one that SF might output the first thing I do after a few more iterations, is to record it, and then do a reverse analysis on it. If it really is a mate then the last move should be a mate in 1. One move before that should be a mate in two. ect. ect. all the way back to the root position. While it may not be a proof tree per se any person with the will and good enough hardware could turn it into a proof tree if they want to spend a good amount of time on it. It would, of course, be better if someone would write some software to automate the process but I'm not holding my breath.

Regards,

Zenmastur

[jp]

You'll see incorrect evals in the middlegame, and won't even consider investigating some winning moves.

The last time I checked there are only three scores in chess. Winning, losing, and drawn. So every evaluation that isn't one of those, is by definition, wrong. This includes ALMOST all evaluations any engine spits out. But that's not the same as being useless.

But being useful in CC chess is not the same as being able to provide theoretical results of positions 1 move from the starting position, which is the claim.

Static evaluation (including endgame tables) determine everything.

So what you're saying is, if I have ANY endgame table bases and I can search deep enough I can play perfect chess. Was that the point you were trying to make?

No, having just any TBs is not good enough to play perfect chess (unless by "any" you mean "every"). You need the 32-man TBs and your search.

[jp]

But, as we all know, SF (or any other engine) is not perfect (very, very far from perfection IMO), so it does not matter what the SF output is or isn't [for getting theoretical results for opening positions].

If you actually believe that why do you bother with posting on this board. If you can't believe anything any engines says, why are all these people bothering to write these engines and then wasting time play matches with them and posting on this board?

Because being very far from perfection does not mean they aren't "very, very good" or "very, very useful", etc.

It also does mean there's the possibility that they will keep improving vastly, which is a good thing, isn't it?

Your statement is like saying that if we cannot achieve perfection in any activity, we should not do it at all.

[Zenmastur]

Yes, of course the human half does something, but surely we can all agree that centaurs are imperfect? "Make up for the shortcomings of the flawed evaluation" doesn't mean the result is guaranteed perfection. It's more likely to have guaranteed imperfection. No one can check every single branch, so there's no way this process can guarantee perfection.

There is no way to guarantee perfection in ANY endeavor. So, what you are saying is this, and every other endeavor is a waste of time because WE CAN'T GUARANTEE PERFECTION???

When it's too simple to be true, it probably isn't true!

Do you have any "real" arguments to make?

Again, no one can guarantee perfectly accurate and exhaustive tagging. This is assuming a magic tagging power that does not exist outside of TBs.

What makes you think the table bases you keep talking about are PERFECTLY accurate. Have you personally checked ever single entry? Of course you haven't! And neither has anyone else. Of course you will claim that they have been checked by a program. But how do you know the program doesn't have an error in it. The answer is you don't. Nor do you know if the compiler and linker used to create the program are error free. What about any libraries used. Are they with out errors? And what about the CPU's used. We know they have errors ( at least the ones made by INTEL) etc. etc.

If your entire argument is based on perfection I would say your argument lacks PERFECTION in a big way. There is no guarantees of perfection in anything. Including your oracle the table bases.

Regards,

Zenmastur

[jp]

Yes, of course the human half does something, but surely we can all agree that centaurs are imperfect? "Make up for the shortcomings of the flawed evaluation" doesn't mean the result is guaranteed perfection. It's more likely to have guaranteed imperfection. No one can check every single branch, so there's no way this process can guarantee perfection.

There is no way to guarantee perfection in ANY endeavor. So, what you are saying is this, and every other endeavor is a waste of time because WE CAN'T GUARANTEE PERFECTION???

No, I'm not saying that at all. Please read what I (and you) wrote after that post (http://talkchess.com/forum3/viewtopic.p ... 85#p828785). I am saying the opposite. You asked (in the later post) why, if I think they don't achieve perfection, I bother or engine writers bother, etc. As I replied, acknowledging that SF is very imperfect does not mean it is not very good and very useful.

When it's too simple to be true, it probably isn't true!

Do you have any "real" arguments to make?

Yes, in the sentences below the one you quoted.

Again, no one can guarantee perfectly accurate and exhaustive tagging. This is assuming a magic tagging power that does not exist outside of TBs.

What makes you think the table bases you keep talking about are PERFECTLY accurate.

If your entire argument is based on perfection I would say your argument lacks PERFECTION in a big way. There is no guarantees of perfection in anything. Including your oracle the table bases.

We cannot guarantee that. We are being generous in assuming existing <8-man TBs have no errors, e.g. in the programs that generated them. You don't seem to be getting the logic straight. You are giving more reasons why we cannot know with certainty theoretical results 1 move from the opening. Yet you are trying to argue the opposite: that we can know with certainty some of them.

We certainly do not need to achieve perfection to talk about our achieved imperfection.

I wonder if there is a larger or a small percentage of such positions going up to 8, 9, 10-piece EGTBs. I'm guessing smaller. We could take random 5-piece and 6-piece EGTBs positions to see where programs without EGTBs misevaluate compared to EGTBs and it would probably be the same ratio as when going up to more pieces.

No need to test random positions. There are tables that will tell you how many positions are won and in how many moves, how many of these are cursed wins, losses or cursed losses. Simply add them up and divide by the total positions in that file and you will have your answer. Easy peasy. For pawnful endgames it would be “nice” to have a break down by how advanced the pawns are. This however, would be a lot of work, but less than it took to create the tables, and it would be quit useful.

No, because as Jim says in a later post, we need to be able somehow to filter those positions. There are many TB positions that will skew the statistics and therefore that we do not want to include. How do we filter these out of the file?

[Uri Blass]

Here's a 20men position where black wins by force:

3r2k1/1Br2ppp/1p6/p5P1/2bNP3/n1P1BP2/7P/R5K1 b - -

If you don't know how do I know no software or hardware would help you

I doubt you know. I have no doubt you think you know.

There won't be any proof tree because SF doesn't produce those and I don't have any software that does. Just SF analysis that either gives a mate or some that ends in TB score.

But, as we all know, SF (or any other engine) is not perfect (very, very far from perfection IMO), so it does not matter what the SF output is or isn't.

And just getting SF to announce mate in X is obviously a useful indicator in a practical game, but does not prove a mate in the way a TB does. I've often watched SF announce mate in 30, and then it keeps calculating, and suddenly it's gone to mate in 50, and then it keeps calculating, and it retracts any mate announcement. That means the mate in 30 was (according to its longer-calculating self) just wrong. I'm sure others have seen this behavior too.

1)No

It does not mean that the mate in 30 was just wrong(I do not know and maybe stockfish has memory problem and does not memorize the mating line and prune it at bigger depth).
Of course hash collision is also possible but I never saw a single case of reproducable analysis of stockfish that show a wrong mate because of hash collision.

In order to prove the mate in 30 was just wrong you need to prove that there is no mate in 30.

I would like to see a position when stockfish claim mate in 30 when there is no mate in 30.
Maybe somebody can try all the mate positions from tablebases to see if there is a single position when stockfish without tablebases claim mate that is shorter than the shortest mate(when we assume for the discussion that tablebases are right about the shorter mate).

2)I would like an engine to be able to generate a tree that prove mate after claiming mate.
Stockfish does not do it but basically it is possible to do it even if the size of the tree may be very big.

Note that a tree does not have to include all options and for
me it is possible to have a tree to prove mate in 2
like 1.Qh6(node 1) threat 2.Qg7#(node 2) no defence when black has 30 legal moves but no legal moves prevent Qg7#
or in case there is a single defence something like
1.Qh6(node 1) threat 2.Qg7#/(node 2) only defence 1...Bf8(node 3) 2.Qxh7#(node 4)

The tree also may use tranposition and basically every node in the tree should have a number.
It is possible to search at node number 756 and claim mate in 4 because it is already proved that node number 3 that has exactly the same position is mate in 4 and in this case I would like to see in node number 756(mate in 4 see node 3).

The idea is that the human can check the tree and see why the program claimed mate in 30 even if it means checking millions of nodes and it will take years to check it for a single human and the human does not need to memorize previous analysis when it check node number 756 because there is a comment that explain what is the earlier node to look at it.

[jp]

1)No

It does not mean that the mate in 30 was just wrong (I do not know and maybe stockfish has memory problem and does not memorize the mating line and prune it at bigger depth).
Of course hash collision is also possible but I never saw a single case of reproducable analysis of stockfish that show a wrong mate because of hash collision.

In order to prove the mate in 30 was just wrong you need to prove that there is no mate in 30.

2)I would like an engine to be able to generate a tree that prove mate after claiming mate.
Stockfish does not do it but basically it is possible to do it even if the size of the tree may be very big.

Not provably wrong, but not trustworthy. Even if we trust the lower-depth SF and the higher-depth SF equally (which seems a very generous thing to do), that still raises (too much) doubt about its mate-in-30 announcement.

Yes, if we can get it to generate a tree, that would help. Maybe without this coding we could catch one of these examples and save it before it disappears, because it'll at least give the PV.

Note that a tree does not have to include all options and for
me it is possible to have a tree to prove mate in 2
like 1.Qh6(node 1) threat 2.Qg7#(node 2) no defence when black has 30 legal moves but no legal moves prevent Qg7#
or in case there is a single defence something like
1.Qh6(node 1) threat 2.Qg7#/(node 2) only defence 1...Bf8(node 3) 2.Qxh7#(node 4)

Yes, this is perhaps even better, but to generate this the program needs to be quite "intelligent", because it needs to "know" a bit about human reasoning.

[Dann Corbit]

2)I would like an engine to be able to generate a tree that prove mate after claiming mate.

Chest can do this and even print it out for you, but if the mate is long you are going to need a lot of paper.

[Ovyron]

Not sure what to add. Hope my point is clear to others, and that someone else can expressed it better.

It seems your point is that since you're unable to use engines to do something, that you assume nobody is.

Here's a thought experiment: suppose that I have in my possession 10 chess positions, for all of them, Stockfish at high depth, or whatever, shows scores higher than 2.00. But only one of them is won, the rest are draws.

I hand them to Zenmastur, and ask him if he can spot the only winning one. Can he do it? Well, it'd probably take him minutes, as the drawn variations make no progress, and the winning one explodes, and the more you analyze it, the worse scores get. At what point does he think he has found it? When the score hits 6.00? When the score hits 8.00?

If he was playing 10 games with those positions, he'd have seen from the distance those positions, and avoided the 9 drawn lines, and gone with the won one, and won all ten games, no problem, regardless of Stockfish's eval.

So what happens if I hand the same positions to you (Zullil)? Do you just raise your hands in the air because you can't tell? Because you can't prove if any position is drawn or not? Because some win could be hanging from a won position that Stockfish scores as drawn or some draw hangs from a drawn position that Stockfish scores as winning? And you just give up?

But there was a way to do it, within minutes, wasn't it?

I'm happy that most people don't know about this, though, if they did, they could see a lost position in the distance and avoid it, and then I'd not be winning any more games, so I continue playing because if you put enough pressure about finding the best moves in an opponent, they'll still slip, and I'll still win. And I'll do it much more often if I learn better openings to play (instead of 40 move mainlines that ICCF GMs play against each other), the analysis of 1.g4 had been very instructive in that front (the draw width of chess is much smaller than I thought, if it doesn't cover 1.g4, so all you need to do it bring your opponent to the edge of it, and push them.)

[zullil]

Not sure what to add. Hope my point is clear to others, and that someone else can expressed it better.

It seems your point is that since you're unable to use engines to do something, that you assume nobody is.

Here's a thought experiment: suppose that I have in my possession 10 chess positions, for all of them, Stockfish at high depth, or whatever, shows scores higher than 2.00. But only one of them is won, the rest are draws.

I hand them to Zenmastur, and ask him if he can spot the only winning one. Can he do it? Well, it'd probably take him minutes, as the drawn variations make no progress, and the winning one explodes, and the more you analyze it, the worse scores get. At what point does he think he has found it? When the score hits 6.00? When the score hits 8.00?

If he was playing 10 games with those positions, he'd have seen from the distance those positions, and avoided the 9 drawn lines, and gone with the won one, and won all ten games, no problem, regardless of Stockfish's eval.

So what happens if I hand the same positions to you (Zullil)? Do you just raise your hands in the air because you can't tell? Because you can't prove if any position is drawn or not? Because some win could be hanging from a won position that Stockfish scores as drawn or some draw hangs from a drawn position that Stockfish scores as winning? And you just give up?

But there was a way to do it, within minutes, wasn't it?

I'm happy that most people don't know about this, though, if they did, they could see a lost position in the distance and avoid it, and then I'd not be winning any more games, so I continue playing because if you put enough pressure about finding the best moves in an opponent, they'll still slip, and I'll still win. And I'll do it much more often if I learn better openings to play (instead of 40 move mainlines that ICCF GMs play against each other), the analysis of 1.g4 had been very instructive in that front (the draw width of chess is much smaller than I thought, if it doesn't cover 1.g4, so all you need to do it bring your opponent to the edge of it, and push them.)

No, that's not my point.