1.g4 opening is losing?

Discussion of anything and everything relating to chess playing software and machines.

Moderators: bob, hgm, Harvey Williamson

Forum rules
This textbox is used to restore diagrams posted with the [d] tag before the upgrade.
jp
Posts: 1254
Joined: Mon Apr 23, 2018 5:54 am

Re: 1.g4 opening is losing?

Post by jp » Wed Feb 12, 2020 10:16 pm

Ovyron wrote:
Wed Feb 12, 2020 10:10 pm
jp wrote:
Wed Feb 12, 2020 10:02 pm
Backtracking like this is a slightly different topic. Even with all the RAM he wants ("engine forgetting" solved), he won't necessarily be able to backtrack to mate at the root, because backtracking will discover different defensive moves (which may or may not still give up mate, but the engine won't necessarily know).
That's why I said "mainline". Mainline means the absolute best moves from black to defend, as you go back in the positions the engine won't change her mind about them because the alternatives lose faster, so it'll still show the mating line up to the root (if the mate in the TT survives.)

This is in fact what I'm doing with his position, to exhaust all black defenses, with the help from a learning engine (that saves important stuff to hard disk, so I can do it with just 128MB RAM for the TT) and it works (I can show a mate score in the root that guarantees black can't delay mate more than it shows, but it's possible white has a faster mate, because you never know if a position that Stockfish says it's just 6.00 actually explodes and mates faster...)
Okay, but I'm assuming we're thinking about a real CC analysis generally. There's no way to know with certainty what the best line is from the defensive side in a general CC position. You can't literally exhaust all defences, even if SF is doing that (according to that code comment) at the times when it makes a mate announcement.

User avatar
Ovyron
Posts: 3827
Joined: Tue Jul 03, 2007 2:30 am

Re: 1.g4 opening is losing?

Post by Ovyron » Wed Feb 12, 2020 11:12 pm

jp wrote:
Wed Feb 12, 2020 10:16 pm
Okay, but I'm assuming we're thinking about a real CC analysis generally.
No, we're talking about engines forgetting lines to mate, those only happen in won positions, and those have a mainline (best attack vs. best defense) so that any white deviation mates slower and any black deviation mates faster.

I do think all this is irrelevant, because if you're on one of those defending, you better resign to save time, and if you're attacking, the slowest moves that mate are enough, so the best thing to do is play the first win that you find and don't worry if it's optimal, because in real CC analysis, generally you want to aim for a position like this as the attacker, or avoid a position like this as the defender, if the position is on the board the game is over and how's it won doesn't matter.

jp
Posts: 1254
Joined: Mon Apr 23, 2018 5:54 am

Re: 1.g4 opening is losing?

Post by jp » Wed Feb 12, 2020 11:24 pm

Ovyron wrote:
Wed Feb 12, 2020 11:12 pm
jp wrote:
Wed Feb 12, 2020 10:16 pm
Okay, but I'm assuming we're thinking about a real CC analysis generally.
No, we're talking about engines forgetting lines to mate, those only happen in won positions
No, that is not true. The engine announces mate in a position you've gone to e.g. 40 plies down the line. (We assume this is correct. It can also be forgotten. We are trying to backtrack from 40 plies down up to the root position.) There could be a blunder that allows mate anywhere before those 40 plies, e.g. one ply back, with the blundering side not in a losing position before it blundered into forced mate.

If you go near the tail of SF's line, this is not just possible but realistic. The moves at the tail are often weird. Of course, if you play through the whole line, this may be obvious.

Ovyron wrote:
Wed Feb 12, 2020 11:12 pm
those have a mainline (best attack vs. best defense) so that any white deviation mates slower and any black deviation mates faster.
A winning position may have a mainline they way you describe it, but that' may be no use to us if we don't know with certainty what it is.

Zenmastur
Posts: 781
Joined: Sat May 31, 2014 6:28 am

Re: 1.g4 opening is losing?

Post by Zenmastur » Thu Feb 13, 2020 12:32 am

Ovyron wrote:
Wed Feb 12, 2020 11:12 pm
jp wrote:
Wed Feb 12, 2020 10:16 pm
Okay, but I'm assuming we're thinking about a real CC analysis generally.
No, we're talking about engines forgetting lines to mate, those only happen in won positions, and those have a mainline (best attack vs. best defense) so that any white deviation mates slower and any black deviation mates faster.

I do think all this is irrelevant, because if you're on one of those defending, you better resign to save time, and if you're attacking, the slowest moves that mate are enough, so the best thing to do is play the first win that you find and don't worry if it's optimal, because in real CC analysis, generally you want to aim for a position like this as the attacker, or avoid a position like this as the defender, if the position is on the board the game is over and how's it won doesn't matter.
It IS relevant because you can use a very similar procedure on a non-mating positions. If you find a good score deep in the tree and it's the best line of play you can back it all the way to the root even if it's a 100 moves away. I know because I have done it. Of course there is a limit, and that limit is defined by the TT size and the TT replacement algorithm. This is how I was reporting -2.40 evals when komodo was reporting -1.20. All you need to find is one good line and then back it up. Generally what happens is the opponents line will start changing trying to find better moves. But if your line is good there are few if any "better" moves. Even if it doesn't propagate all the way back to the root, let say it stops 3-moves from the root, the opponents best move will have changed at move 3. If he hasn't seen this line we will likely go for the older and worse line that leads to the position you found and a much better score for you. Even if it doesn't reach the root the results have value because they change the opponents best move close to the root. You only need three to four of these to win a game. If you do this every move and there is a 1 in 10 chance your opponent doesn't see the proper continuation and the game is 50 moves long you have a very good chance of winning. It's not a sure thing but it definitely helps in the long run. Most of my CC games are won this way.
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

jp
Posts: 1254
Joined: Mon Apr 23, 2018 5:54 am

Re: 1.g4 opening is losing?

Post by jp » Thu Feb 13, 2020 12:37 am

Zenmastur wrote:
Thu Feb 13, 2020 12:32 am
All you need to find is one good line and then back it up. Generally what happens is the opponents line will start changing trying to find better moves.
Yes.
Zenmastur wrote:
Thu Feb 13, 2020 12:32 am
But if your line is good there are few if any "better" moves. Even if it doesn't propagate all the way back to the root, let say it stops 3-moves from the root. The opponents move will have changed. If he hasn't seen this line we will like go for the older and worse line that leads to the position you found and a much better score. Even if it doesn't reach the root the results have value because they change the opponents best move close to the root. You only need 3 to four of these to win a game. If you do this every move and there is a 1 in 10 chance your opponent doesn't see the proper continuation and the game is 50 moves long you have a very good chance of winning. It's not a sure thing but it definitely helps in the long run. Most of my CC games are won this way.
How do you keep all the lines, including the older and worse line, as you go backwards towards the root? Do you just save them manually, or is this automated somehow?

Zenmastur
Posts: 781
Joined: Sat May 31, 2014 6:28 am

Re: 1.g4 opening is losing?

Post by Zenmastur » Thu Feb 13, 2020 1:10 am

jp wrote:
Thu Feb 13, 2020 12:37 am
Zenmastur wrote:
Thu Feb 13, 2020 12:32 am
All you need to find is one good line and then back it up. Generally what happens is the opponents line will start changing trying to find better moves.
Yes.
Zenmastur wrote:
Thu Feb 13, 2020 12:32 am
But if your line is good there are few if any "better" moves. Even if it doesn't propagate all the way back to the root, let say it stops 3-moves from the root. The opponents move will have changed. If he hasn't seen this line we will like go for the older and worse line that leads to the position you found and a much better score. Even if it doesn't reach the root the results have value because they change the opponents best move close to the root. You only need 3 to four of these to win a game. If you do this every move and there is a 1 in 10 chance your opponent doesn't see the proper continuation and the game is 50 moves long you have a very good chance of winning. It's not a sure thing but it definitely helps in the long run. Most of my CC games are won this way.
How do you keep all the lines, including the older and worse line, as you go backwards towards the root? Do you just save them manually, or is this automated somehow?
For overnight runs it's automated. During the daytime I do it interactively and then late at night I set up for the overnight runs. Be advised that what is outlined above is only one step in a multistep process and the most important steps have been omitted. i.e. if you just do this in some random fashion you will only succeed in wasting a bunch of computer time for little if any gain. It takes quite a bit of experimentation to get the procedures "tuned" so they work efficiently and then you have to apply them in an efficient manner. It's easy to take a "tuned" procedure and then waste a bunch of time with it by applying it in inappropriate situations.
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

User avatar
Ovyron
Posts: 3827
Joined: Tue Jul 03, 2007 2:30 am

Re: 1.g4 opening is losing?

Post by Ovyron » Thu Feb 13, 2020 5:00 am

Zenmastur wrote:
Thu Feb 13, 2020 12:32 am
It IS relevant because you can use a very similar procedure on a non-mating positions.
Nope, this procedure I'm doing for calculating a mating line is about x40 times slower than my normal procedure for finding wins (where I don't care how slow the wins are, just to tell apart draws for wins.) If a win would exist that was much faster but it'd take me too long to find it I'd just mark it as a draw (like you do to "cursed wins") and just make sure the winning lines win (but never about how quick.)

The method I'm using is completely useless for games (because it trades accuracy for speed) and I won't be using it ever again (because I hate grinding faster mates). What is that? I solve an endgame to a mate in 18 and have to keep grinding it to a mate in 13? Who in their right mind does that? A mate in 18 has always been enough, the time taken to grind those extra 5 faster moves is unacceptable.

I have looked at so many positions that I'm now able to suggest moves that mate faster than what Stockfish can see (it reports some random rook move as a mate in 23, I play a4!! and it goes quickly to a mate in 17...), don't expect to see this method in our game (if you decide to play till mate, I'll play the slowest mate that still wins to prove a point :D )

BTW, my top line has been a mate in 35 for 1.5 days now, so unless there's an actual mate in 29, I've been done for a while (but I have no way to know), so prepare your 1.g4 defense already. What is taking me so long is not to find the mate, it's to make sure there's no faster ones (to a point; I'm fine with missing some mate in 30 if 35 suffices. If it's on 29 and I couldn't find faster than 35, you've got a point.)

Zenmastur
Posts: 781
Joined: Sat May 31, 2014 6:28 am

Re: 1.g4 opening is losing?

Post by Zenmastur » Thu Feb 13, 2020 7:34 am

Ovyron wrote:
Thu Feb 13, 2020 5:00 am
Zenmastur wrote:
Thu Feb 13, 2020 12:32 am
It IS relevant because you can use a very similar procedure on a non-mating positions.
Nope, this procedure I'm doing for calculating a mating line is about x40 times slower than my normal procedure for finding wins (where I don't care how slow the wins are, just to tell apart draws for wins.) If a win would exist that was much faster but it'd take me too long to find it I'd just mark it as a draw (like you do to "cursed wins") and just make sure the winning lines win (but never about how quick.)
Well, I don't know what procedure you're using but it sure is taking you a while. The only reason I gave you that problem was I was working on the much harder position and that happen to be a sub-solution. I didn't mean it to take this long. I don't know how long it took me but it wasn't long. I think your method isn't tuned for this type of problem. The method I use doesn't care if it's a mate or not, it works the same in either case in about the same time.
The method I'm using is completely useless for games (because it trades accuracy for speed) and I won't be using it ever again (because I hate grinding faster mates).


Go back to you other method. What's the big deal. By the way, the position you were having problems with I solved in less than 4-minutes. I found a mate and refined it to a much smaller number than you had in a minute or so. I worked on it a few extra minutes and shaved off 1 more move before I quit.
What is that? I solve an endgame to a mate in 18 and have to keep grinding it to a mate in 13? Who in their right mind does that?
About once or twice a week I take ICCF games and do that for fun! :D :D :D

You can even do it on games that were drawn if you can find an error in them. In fact I like doing those the best!

It's good practice.
I have looked at so many positions that I'm now able to suggest moves that mate faster than what Stockfish can see (it reports some random rook move as a mate in 23, I play a4!! and it goes quickly to a mate in 17...), don't expect to see this method in our game (if you decide to play till mate, I'll play the slowest mate that still wins to prove a point :D )
It's amazing how bad SF plays down in the tree. :evil: It's a good thing there is a way to mitigate that! :D

There are way too many games played on ICCF where one side is dominating the whole game and then in a few moves it'a a draw or it's been 0.00 the whole game and suddenly one side blunders three times in a row and loses. It's almost like people are purposely giving away points to their opponents. Why do you suppose they would do that? :D

Anyway, have fun with your mate finding! I'll be waiting.

If you don't see a need for better hardware I guess you should just stay with what you have for the rest of your life and let that be a lesson to you! :D :D :D

One thing I am curious about. Do you ever go back to games you have played and see what mistakes you have made? i.e. post game analysis and then look at you game analysis to see where it went wrong?
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

User avatar
Ovyron
Posts: 3827
Joined: Tue Jul 03, 2007 2:30 am

Re: 1.g4 opening is losing?

Post by Ovyron » Thu Feb 13, 2020 3:46 pm

Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
Go back to you other method. What's the big deal.
My other method would have mated in 42. I'd have taken me 1 minute/position and it'd have been annoying that an opponent didn't resign in this position.

The big deal is that this line wouldn't have been inside 0-5 moves from best.

If this line was a subtree of another position I'd have tagged it as "won" and gone back immediately without ever caring about how long white takes to mate.
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
About once or twice a week I take ICCF games and do that for fun! :D :D :D
You wouldn't have time for that if you used it for playing more games. Try to beat the strongest players you haven't yet. Try beating someone like Ozymandias and see if any of your methods work against such a solid player.
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
By the way, the position you were having problems with I solved in less than 4-minutes. I found a mate and refined it to a much smaller number than you had in a minute or so. I worked on it a few extra minutes and shaved off 1 more move before I quit.
Note this may be more related to lacking tablebases than lacking bigger hardware. Since you enjoy optimizing won positions (where optimization is useless and playing the first win you see is good enough) you might want to try and see how the position looks with just 5menTB, maybe it'll take longer so your fun will last longer :D
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
One thing I am curious about. Do you ever go back to games you have played and see what mistakes you have made? i.e. post game analysis and then look at you game analysis to see where it went wrong?
Only for games I've lost. I've only lost 3 games since 2015, one on the LSS where the guy outbooked me in some Silician, along with other games where other people outbooked me (their play is impressively aggressive) but only he was able to win, since then I have adopted a much better opening strategy (what Antares calls the "ICCF optimized '20 Uly", since I used to play more "romantic" openings, but now I play ones without risk). This one, where I couldn't find a win for the opponent in a lost position, but at least for people on the thread, they weren't able to find a win either, until someone with big hardware and tablebases appeared; but I let Komodo lead that game, big mistake! And my last lost game was 1.g4 against Harvey, I didn't know if I should count it, because I'd never play 1.g4 in a "real game", but it'd be unfair to dismiss his effort, so I count it.

There's no time to check my opponents' loses to see where they made a mistake, as I still have 16 games ongoing trying to make them play more mistakes. How many games do you play simultaneously?

Zenmastur
Posts: 781
Joined: Sat May 31, 2014 6:28 am

Re: 1.g4 opening is losing?

Post by Zenmastur » Fri Feb 14, 2020 12:57 am

Ovyron wrote:
Thu Feb 13, 2020 3:46 pm
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
Go back to you other method. What's the big deal.
My other method would have mated in 42. I'd have taken me 1 minute/position and it'd have been annoying that an opponent didn't resign in this position.

The big deal is that this line wouldn't have been inside 0-5 moves from best.

If this line was a subtree of another position I'd have tagged it as "won" and gone back immediately without ever caring about how long white takes to mate.
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
About once or twice a week I take ICCF games and do that for fun! :D :D :D
You wouldn't have time for that if you used it for playing more games. Try to beat the strongest players you haven't yet. Try beating someone like Ozymandias and see if any of your methods work against such a solid player.
I haven't ever played more than 30-40 games at once. That's enough for me. And I don't like playing players that have a game load over a hundred. I once played a person that had over 500 games ongoing. It was a miserable affair. I complained about him to my national rep. Eventually, I think he was banned and they forfeited all his games. IIRC it was something like 582 games.

I don't know Ozymandias. My plan is to play in some norm tournaments.
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
By the way, the position you were having problems with I solved in less than 4-minutes. I found a mate and refined it to a much smaller number than you had in a minute or so. I worked on it a few extra minutes and shaved off 1 more move before I quit.
Note this may be more related to lacking tablebases than lacking bigger hardware. Since you enjoy optimizing won positions (where optimization is useless and playing the first win you see is good enough) you might want to try and see how the position looks with just 5menTB, maybe it'll take longer so your fun will last longer :D
I have already tested that. It's almost always slower to use table bases unless the position is very hard. In some positions it's MUCH MUCH slower to use TB's. The problem is when you have already found a mate and then backup one or two plies it will report a win ( a table base score ) and it can take for ever for it to go back to the actual mating line. So when mate finding sometimes I use them and sometimes I don't. It just depends on the position. I generally start with 6-man and if it starts giving me trouble I stop using them. I recently found a way to get around this issue most of the time.
Zenmastur wrote:
Thu Feb 13, 2020 7:34 am
One thing I am curious about. Do you ever go back to games you have played and see what mistakes you have made? i.e. post game analysis and then look at you game analysis to see where it went wrong?
Only for games I've lost. I've only lost 3 games since 2015, one on the LSS where the guy outbooked me in some Silician, along with other games where other people outbooked me (their play is impressively aggressive) but only he was able to win, since then I have adopted a much better opening strategy (what Antares calls the "ICCF optimized '20 Uly", since I used to play more "romantic" openings, but now I play ones without risk). This one, where I couldn't find a win for the opponent in a lost position, but at least for people on the thread, they weren't able to find a win either, until someone with big hardware and tablebases appeared; but I let Komodo lead that game, big mistake! And my last lost game was 1.g4 against Harvey, I didn't know if I should count it, because I'd never play 1.g4 in a "real game", but it'd be unfair to dismiss his effort, so I count it.

There's no time to check my opponents' loses to see where they made a mistake, as I still have 16 games ongoing trying to make them play more mistakes. How many games do you play simultaneously?
It's enlightening to go back and re-analyze your old games. It's surprising the number of errors you will find. Even in drawn games. Which reminds me, I down loaded all your games a few months ago and analyzed them. I'll have to go looking for them. Now that I know more about your analysis method they might be worth a second look.

My current game load is ZERO. My last game ended in mid December. I had one of those players that took forever to move and wanted to play all the way to mate. I ended the game by forcing a conversion to a 6-man position and then claiming a TB win.

I stated having medical issues about 18 months ago and it was affecting my play so I stopped entering new events. In the mean time, I think I've got my health situation under control so I plan on entering new events. I was think about norm tournaments and maybe a theme event or two. I've always wanted to try one of those. I think they would be fun. Too bad you can't gain rating points from them.
Only 2 defining forces have ever offered to die for you.....Jesus Christ and the American Soldier. One died for your soul, the other for your freedom.

Post Reply