Can Engines properly assess openings?

Sort:
drmrboss
Prometheus_Fuschs wrote:

Computers aren't great at deciding what openings are the best, they do help to cull the bad ones and even then, it doesn't mean it cannot be played, nobobdy is good enough to thoroughly refute an opening.

Yes, computer are 3500 elo. Show me the opening where you know about 4000 elo strength!!

Or else, play vs properly configured Stockfish for 20 moves.. And then  stop and show us, Stockfish  played horrible in the opening and you have better position than Stockfish.

 

 

Prometheus_Fuschs
drmrboss escribió:
Prometheus_Fuschs wrote:

Computers aren't great at deciding what openings are the best, they do help to cull the bad ones and even then, it doesn't mean it cannot be played, nobobdy is good enough to thoroughly refute an opening.

Yes, computer are 3500 elo. Show me the opening where you know about 4000 elo strength!!

Or else, play vs properly configured Stockfish for 20 moves.. And then  stop and show us, Stockfish  played horrible in the opening and you have better position than Stockfish.

 

 

A sufficiently better player can beat another player in virtually any opening.

drmrboss
Prometheus_Fuschs wrote:
drmrboss escribió:
Prometheus_Fuschs wrote:

Computers aren't great at deciding what openings are the best, they do help to cull the bad ones and even then, it doesn't mean it cannot be played, nobobdy is good enough to thoroughly refute an opening.

Yes, computer are 3500 elo. Show me the opening where you know about 4000 elo strength!!

Or else, play vs properly configured Stockfish for 20 moves.. And then  stop and show us, Stockfish  played horrible in the opening and you have better position than Stockfish.

 

 

A sufficiently better player can beat another player in virtually any opening.

Show me a legitimate game where a human beat Stockfish.

A claim against beating Stockfish is more than a hoax / scam than Alien .

( A publicy known computer setting, e.g known engine room such as playchess server or streaming). I know Naka won 1 game against Rybka in blitz , but today Stockfish can beat 1:100 time odd against that old Rybka in tcec).

tmkroll

Drmrboss, we were using the analysis on this site and in people's version of Stockfish on their home computers and the let's check analysis on the chessbass cloud which has come from a lot of people running engines on different machines with different depths. The phrase "to a ridiculous search depth" or some such other came up regarding some of that analysis. I might be able to find one of the threads again. It was in the Nxf7 variation where Black needs to counter-intuitively move the already developed Knight backward to block a check instead of developing another piece with tempo. I never saw the engine suggest that move unless you suggested it first and then it gets in the hash table or something and the next time you run the search it looks at that move. When you do make it look at that move it pretty quickly finds the draw. That was where people were talking about very high search depths all coming up 0. No one said anything about nodes. The member who told us that 0-0-0 was actually better said the line was known as an "engine trap." IMO, there was no point regardless, as I said, though because I don't think any engine will take on f7 with the Knight in the first place if you don't make it start in that position.

Prometheus_Fuschs
drmrboss escribió:
Prometheus_Fuschs wrote:
drmrboss escribió:
Prometheus_Fuschs wrote:

Computers aren't great at deciding what openings are the best, they do help to cull the bad ones and even then, it doesn't mean it cannot be played, nobobdy is good enough to thoroughly refute an opening.

Yes, computer are 3500 elo. Show me the opening where you know about 4000 elo strength!!

Or else, play vs properly configured Stockfish for 20 moves.. And then  stop and show us, Stockfish  played horrible in the opening and you have better position than Stockfish.

 

 

A sufficiently better player can beat another player in virtually any opening.

Show me a legitimate game where a human beat Stockfish.

A claim against beating Stockfish is more than a hoax / scam than Alien .

( A publicy known computer setting, e.g known engine room such as playchess server or streaming). I know Naka won 1 game against Rybka in blitz , but today Stockfish can beat 1:100 time odd against that old Rybka in tcec).

I hope you realize I didn't claim any of that and that Stockfish beating humans agrees with what I said.

HolyCrusader5

I think what I am getting here is that engines cannot determine the soundness of some openings accurately, but they can still play opening lines well. An engine's opinion on some attacking lines does not mean that the line is objectively worse. This does not mean a human has a better understanding than Stockfish at openings, but theory as a whole understands opening lines better than an engine may.

drmrboss
Prometheus_Fuschs wrote:
drmrboss escribió:
Prometheus_Fuschs wrote:
drmrboss escribió:
Prometheus_Fuschs wrote:

Computers aren't great at deciding what openings are the best, they do help to cull the bad ones and even then, it doesn't mean it cannot be played, nobobdy is good enough to thoroughly refute an opening.

Yes, computer are 3500 elo. Show me the opening where you know about 4000 elo strength!!

Or else, play vs properly configured Stockfish for 20 moves.. And then  stop and show us, Stockfish  played horrible in the opening and you have better position than Stockfish.

 

 

A sufficiently better player can beat another player in virtually any opening.

Show me a legitimate game where a human beat Stockfish.

A claim against beating Stockfish is more than a hoax / scam than Alien .

( A publicy known computer setting, e.g known engine room such as playchess server or streaming). I know Naka won 1 game against Rybka in blitz , but today Stockfish can beat 1:100 time odd against that old Rybka in tcec).

I hope you realize I didn't claim any of that and that Stockfish beating humans agrees with what I said.

Opening involve virtually 1/3 of the game. 

If there is reasonable condition enigines play bad in opening , then human will claim they can play better than engines in opening with proof of games.

In fact , engines are better than 99.99% of human in any phase of the game!

It means, engines are playing opening properly!

I wont be surprised if there is 1:100 opening line that engines did not play well. But in general , engine moves are trustable in any phase of the game!

 

drmrboss

This topic is like , can a car run properly (faster than a human ) ? 

(answer is human can run faster than a car in certain conditions like in muddy field etc, otherwise no need to proof) 

HolyCrusader5

Engines may play openings better than humans but they are unable to determine whether an opening is sound.

HolyCrusader5

I think that dozens of years of human analysis triumphs what Stockfish says at the starting position at depth 45

Prometheus_Fuschs
drmrboss escribió:
Prometheus_Fuschs wrote:
drmrboss escribió:
Prometheus_Fuschs wrote:
drmrboss escribió:
Prometheus_Fuschs wrote:

Computers aren't great at deciding what openings are the best, they do help to cull the bad ones and even then, it doesn't mean it cannot be played, nobobdy is good enough to thoroughly refute an opening.

Yes, computer are 3500 elo. Show me the opening where you know about 4000 elo strength!!

Or else, play vs properly configured Stockfish for 20 moves.. And then  stop and show us, Stockfish  played horrible in the opening and you have better position than Stockfish.

 

 

A sufficiently better player can beat another player in virtually any opening.

Show me a legitimate game where a human beat Stockfish.

A claim against beating Stockfish is more than a hoax / scam than Alien .

( A publicy known computer setting, e.g known engine room such as playchess server or streaming). I know Naka won 1 game against Rybka in blitz , but today Stockfish can beat 1:100 time odd against that old Rybka in tcec).

I hope you realize I didn't claim any of that and that Stockfish beating humans agrees with what I said.

Opening involve virtually 1/3 of the game. 

If there is reasonable condition enigines play bad in opening , then human will claim they can play better than engines in opening with proof of games.

In fact , engines are better than 99.99% of human in any phase of the game!

It means, engines are playing opening properly!

I wont be surprised if there is 1:100 opening line that engines did not play well. But in general , engine moves are trustable in any phase of the game!

 

There is a difference between being able to play properly and to evalue properly...

 

And I'm talking short openings, <15 moves, beyond that it's more like a middlegame that has been analysized.

drmrboss
HolyCrusader5 wrote:

I think that dozens of years of human analysis triumphs what Stockfish says at the starting position at depth 45

Probably, engines are just a million times faster than human. If you use 1 million times of human resources, it can probably be.

But why you would like to spend?

e.g, A  Lawn mower was inveted to save our time, but 1000 people can mow faster than the machine.

drmrboss
tmkroll wrote:

Drmrboss, we were using the analysis on this site and in people's version of Stockfish on their home computers and the let's check analysis on the chessbass cloud which has come from a lot of people running engines on different machines with different depths. The phrase "to a ridiculous search depth" or some such other came up regarding some of that analysis. I might be able to find one of the threads again. It was in the Nxf7 variation where Black needs to counter-intuitively move the already developed Knight backward to block a check instead of developing another piece with tempo. I never saw the engine suggest that move unless you suggested it first and then it gets in the hash table or something and the next time you run the search it looks at that move. When you do make it look at that move it pretty quickly finds the draw. That was where people were talking about very high search depths all coming up 0. No one said anything about nodes. The member who told us that 0-0-0 was actually better said the line was known as an "engine trap." IMO, there was no point regardless, as I said, though because I don't think any engine will take on f7 with the Knight in the first place if you don't make it start in that position.

What do you mean?

Nxf7 was not suggested but still draw .Do you mean other moves that SF suggest leads to worse than Nxf7? Show me the FEN or position let me analyse with my 4 cores i5 cpu for 3 mins with  6 men TB access. ( approx 1 billion position ).

I wont be surprised if SF missed 1% of opening ( I will forward that 1% for stockfish development fishcooking discussion)

Are you saying other moves that SF suggest leads to a loss or  worse position?

 

 

dpnorman

There isn't a straightforward answer to your question. In general, the answer is yes, in the sense that if a position is winning, the computer will likely figure out why, and if a position is dead, a computer will likely figure out why. But there are also positions in which the computer's horizon effect influences its evaluation, or in which it thinks it can win against a fortress, or in which it gives a certain number just because it has been programmed to like space, two bishops, or some other specific imbalance depending on the engine.

There are openings in which the engine's numbers tend to be a little (perhaps overly) optimistic for one side or the other, and this is most common for instance in the Mar del Plata King's Indian and the Modern Benoni, openings in which a computer will usually give white relatively large static advantages. These openings feature relatively fixed pawn structures in the center (apart from the fact that at times f4+e5 can happen in the Benoni) and in both cases you often have to worry about horizon effect, as black has pawn storms and long-term counterplay. It's entirely possible the computers have a point, since the objective soundness of both openings have at times been questioned by strong theoreticians, but in practical chess black does better than the computer numbers tend to indicate he should. And on the flipside, as someone who regularly plays the French Defense, I often find the computer evaluates dangerous positions as roughly equal, and it isn't always clear if this is because the positions are actually equal (and only difficult to play in practice), or if this is because the computer has a horizon issue or a programmed tendency not to understand how strong the f5 break or some other pawnstorm is until it is shown the way.

Once we realize that the computer is flawed, and that the number it gives isn't necessarily correct (particularly in the very first few moves), we still have to acknowledge that computers are much stronger than us, and that they have and will always have a very profound role in opening theory. At top level everything needs to be deeply analyzed with computers; that's where novelties come from. When the human who's helping the computer do the analysis is also very strong, and knows when it's giving you iffy numbers and when it has a point, then you're really getting somewhere. Otherwise, computers are useful, but don't rely on them for everything, and don't let a number be the reason you play one opening instead of another. 

drmrboss
HolyCrusader5 wrote:

Engines may play openings better than humans but they are unable to determine whether an opening is sound.

Why not?? You can see evaluation values before and after the moves. If evaluations get worse, it is unsound, getting better means sound.

Or else, do you want an engine to explain the moves verbally? In fact it is possible, if chess is a big business that everyone is interested, programmers can do verbal output / explaination of engine moves.

 

They have their own evaluation that guide them to decide the better moves and they always choose the better move after the allocated search time.

drmrboss

If you have a modern graphic card ( 2018, 2019), leela is a better choice than Stockfish for multiple reasons.

1. Leela can tell you, your chance of winning % / probability in every possible position.

 

Is it something that you are looking for?

In fact , even Komodo Authour and Grandmaster Larry Kaufmen prefer Leela for analysis.

And also, in term of quality of game play Leela already surpassed Stockfish in TCEC.

MARattigan
drmrboss wrote:

...

1. Leela can tell you, your chance of winning % / probability in every possible position.

...

 

In every possible position the probability of winning with perfect play is either 0 or 1. With imperfect play it depends in the great majority of cases on who or what the players are and how much they've had to drink.

Does Leela always give a probability of 0 or 1 or does it ask you who is playing and how much they've had to drink? Sounds like an interesting program.

drmrboss
Manatini wrote:

And 10 years ago we had people saying Fritz was great for the opening.

10 years from now we (or maybe I should say you) will look back and see Leela isn't some oracle.

Yes, when those people come back and see Today's stockfish analysis, they will be shocked.

 

Within 10 years, computers are approx 10 times faster. More importantly software are 50 times better.

 

 Tests have been shown that, 1 hour of analysis by fritz 10 in 10 years ago would be equavilent to 1 min of analysis by today Stockfish 10+.

Including hardware and software, 10 hours of analysis by Fritz 10 in those old hardwares approx equalivent to 1 min of analysis by today Stockfish. ( x500 times better).

drmrboss
MARattigan wrote:
drmrboss wrote:

...

1. Leela can tell you, your chance of winning % / probability in every possible position.

...

 

In every possible position the probability of winning with perfect play is either 0 or 1. With imperfect play it depends in the great majority of cases on who or what the players are and how much they've had to drink.

Does Leela always give a probability of 0 or 1 or does it ask you who is playing and how much they've had to drink? Sounds like an interesting program.

Yes, in perfect play by both players, the chance of winning is zero( 50% scores). But there is no perfect player up to so far and all games are played in restricted time control. So white has slightly better chance resulting 53-55% win rate in Leela training. All Leela games are trained against herself, 55% means, if she play as in this position ( e.g white turn in initial position) , she will get 55% scores against herself.

MARattigan
drmrboss wrote:
tmkroll wrote:

There's a line in the Traxler where a lot of people on this forum could play better than Stockfish. We were debating it here a few years back. Stockfish says White is winning until it sees black has a draw by repetition, then its evaluation goes to 0. Stockfish will take that draw but people who read that forum would castle Queenside as black. Eventually Stockfish sees black is better but it takes it a very long time. There's a line in the KG that at least five or six years back Fritz was similar, idk about now. Of course engines will never play into either of these lines if you don't make it do it because their opening books have been programmed by human players who have studied and know they are bad.

Which version of Stockfish you use and how many nodes SF searched for that position?

Do you mean analysis by this crappy chess.com server stockfish? In fact if SF search only a few hundreds nodes per move, her strength will be like 1200, but a few hundred million nodes per move will make her like 3500.

 

Show me the position, and I will analyse in 3 mins and show you how strong  stockfish is. ( Let me see whether SF really played bad)

Try your SF out on this position. It's a well known win for White, but my SF can't play it for toffee.


 As mentioned in a previous thread my SF evaluates the following position, which you yourself posted, as +6.40 no matter how long I leave it running, whereas Black has a very easy draw.

This one even gets +7.34.

 

Throw in an extra piece and it does no better. It evaluates this win for White at 0.16 at depth 30 both before and after it blows it on its second move.

In fact endgames tend to get more complicated the more men there are on the board. The maximum length forced mates with perfect play (no 50 move rule) are something like 28, 43, 127, 262 and 594 for 3,4,5,6 and 7 men respectively.

SF appears to play and evaluate 3 man positions perfectly (if you take +ve, 0 and -ve evaluations to mean wins draws and losses) but it already starts going awry with both evaluation and play with 4 pieces (it can't play KBNK accurately). With 5 or 6 pieces it starts losing half points.

Can you really believe that in spite of that, when it gets up to a 32 piece endgame, it starts to give accurate evaluations?

And what do the evaluations mean anyway - there's nothing in my SF documentation that tells me. Positions after all are either won for one side or drawn; there's nothing in between.