< Earlier Kibitzing · PAGE 2 OF 2 ·
|Jul-10-19|| ||Sally Simpson: ***
I'm with you all the way on the hype these things get, I rejoice in being a Luddite but Alpha is shifting in a new direction.
click for larger view
..when it offered the d5 pawn it's fishing for a bad move, usually these things shun such moves because they are blinkered and only expect the best move.
The reason I'm interested in this move is because the position repeats two moves later.
click for larger view
(difference being the Rook is e1).
And Alpha does not play play 29. Re4 it played 29.Bc1. It's as if it's saying: 'OK he is not going to fall for that, let's try something else.'
That is why I was asking if S.F. etc would offer the same pawn.
K.P. baulks when I mention traps and two move tricks but you csmath appear to be a tournament player, you know this is our bread and butter, especially at the lower levels. At last a computer we can relate to. It's pointless us following a computer bet move v best move lines because that is not what we going to see.
This Aplha is stimulating and I've never said that before about a computer. (I said something like it once in an advert for a Texas Instrument machine in the 1980's in a P.C. World advert but then I was taking the money, selling my soul if you will.)
I too am also sceptical of the Alpha and Carlsen claims which have not (yet) come from Carlsen but from punters and the players he is leaving in his wake as the they struggle to come to grips that when Carlsen is on song and hot he can make the rest look mediocre.
There was no sign of it 6 months ago at the W.C. match. Carlsen always has this ability, he is currently hot and in his own words 'lucky'.
The alpha hype, which I'm open-minded enough to say a lot of is justified, is being nicked by the other programs. Recently Komodo advertised it's model (Mark13) now has 'Alpha Technology.' https://en.chessbase.com/post/annou...
Can you give that link to this game. (I must get in touch with that lad. He wants a picture of me and pronounces my name 'Gee-Off' (rimes with pee-off, maybe he means it. :) ) https://www.youtube.com/watch?v=nJ2...)
Hopefully picking up the Sadler book on Alpha today. Thanks.
(appears there are two ways to spell 'sockdollager' maybe it's one of those that has UK and American spelling. However in the past I myself have used it with one 'L' (OOPS!) https://www.redhotpawn.com/chess-bl... so I will revert to one 'L'.
Edward Winter has two spellings but not about the 'L' he is more concerned with Sockdolager or Sockdologer C.N. 7360 )
|Jul-10-19|| ||keypusher: <Gee-off> Here you go.|
Nice game against Hynoch!
|Jul-10-19|| ||Sally Simpson: ***
Thanks again, me and that lad went the same way. I'm chuffed to tea breaks. There is life in the old dog yet. I can still sniff out a neat variation.
When I was watching it I realised it was quite simple. Bd7 was the star move which he asked the viewer to turn off and find. (I was sweating on that bit...is it Bd7. Is the same idea there but with a niftier sequence. NO..I'm right!)
He had a different winning sequence if the Rook on e6 is not taken. (I baled out with the old Reinfeld trick; 'White is winning.' looks like many ways to win) when I posted I expected someone to find a bust v the whole line, as I say, too me was fantasy.
Went to pick up book but told 'maybe' tomorrow.
Looked around for a preview and found https://www.newinchess.com/media/wy...
Looks the chess does not start till page 131 chapters 1-5 taken up with how it thinks etc etc and etc.. I'm not all that interested how it works.
And it's expensive. 90% of my library is 2nd hand. Edinburgh has over 50 2nd shops selling books and I have contacts all over Scotland who call me when they get a new batch in.
She...Mrs Ge-off will only allow me £20.00, if where I am going tomorrow has loads of good 2nd hand books I'll give the book a miss...it's bound to turn up 2nd hand somewhere...and anyway now that I myself have figured out Alpha Zero I don't need it! (later I'll look for another Alpha game with another 'pie in the sky' line. )
Thanks for link, I'm still glowing that it's right, people will think I copied it. I'd never knowingly do that but in years to come I'll say he copied from me (revenge for Ge-off! :) )
|Jul-10-19|| ||keypusher: <Sally> Glad you like this game. I think <Game Changer> discusses at some point why A0 is likely to pick no-harm no-foul pawn sacs like the one in this game.|
Anyway, how about the game discussed at the link below, which I've been (trying to) fling in your face for some time. Not in database. Annotations mostly nicked from <Game Changer>.
RESISTANCE IS USELESS. YOU MUST SUBMIT TO THE HYPE!
|Jul-10-19|| ||scholes: Thing is while leela and stockfish are at similar strength levels compared to humans.
But leela is nearly 100 elo stronger than stockfish in opening analysis. That's the reason leela is able to compete with stockfish despite being nearly 1000 elo weaker in tactics. |
For example see the link below. Leela on cpu beat sf10 in opening analysis.
|Jul-10-19|| ||john barleycorn: <scholes: Thing is while leela and stockfish are at similar strength levels compared to humans. But leela is nearly 100 elo stronger than stockfish in opening analysis. That's the reason leela is able to compete with stockfish despite being nearly 1000 elo weaker in tactics. ...>|
That is in accordance with my observations and analysis.
|Jul-10-19|| ||keypusher: <Gee-off>
I like A0 and the book a lot, and I've said many mean things to you over the years despite generally enjoying your posts. So I'm happy to get you a copy if you can supply me an address. You can respond at my forum here. Happy to delete the address after I see it.
|Jul-10-19|| ||MrMelad: There’s a lot of hype around deep learning in general and reinforcement learning in particular all over the world in every single algorithmic / computer vision / AI domain you can think of and for good reasons. From autonomous cars to robots and medical devices everyone is talking about deep reinforcement learning.|
Deep reinforcement learning techniques was shown in multiple academic peer reviewed studies as very efficient and in finding optimal strategies in games, from Atari games to Starcraft, chess and go.
Magnus Carlsen said in a recent interview that he was inspired by AlphaZero.
Keep your facts straight then form whatever opinion you want.
|Jul-11-19|| ||Sally Simpson: Hi MrMelad,
<"Magnus Carlsen said in a recent interview that he was inspired by AlphaZero." >
I've heard about this but never seen a link to when he said it. I added (yet) when I spoke about it.
I have seen this: (you can tell by the lack of beard when it was filmed.)
He says he skimmed through the book yesterday and found it, after a long pause and with a smile on his face, 'inspirational.' (which is pretty startling for a 'skim through.')
Is this where everyone have got 'Inspirational.' from. A one word throw away, with a grin and a shrug of the shoulders on a cherry picked part of the interview.
The full interview is here: https://www.newinchess.com/game-cha...
press the '> button' twice and you will see it has been used as a plug for the book. (they have not edited it like above...OOPS!)
Early in the interview he was asked what he did yesterday. He answered he played football and that was it. (what happened to skimming though a book and getting inspired...)
Carlsen added he was thinking during his game how would Alpha Zero have played this position. He closed with I am not 'Alpha Zero' so played something else. So much for being inspired.
(How about I've been world champion since 2013 without Alpha why change now? )
Then we get this exchange.
Q: "How close do you think you can get to playing like Alpha Zero?"
Carlsen: "Not very close."
Inspired or in awe.
|Jul-11-19|| ||Sally Simpson: ***
Thank you for your kind offer. I could eaily afford the book and if there is nothing better I may get it. My objection is paying all that money for what looks like 50% of the stuff would never get read.
I recall that game and that thread. I stopped going there as I up to my neck in jargon and I had upset AylerKupp when I was actually defending him about being a troll.
I like rattling cages...sometimes I learn things or change opinions but never no malice.
Yes that game. You should submit it especially if Sadler and Co. thought it was instructive enough to go in the book.
I do recall this bit:
".... Black is not planning to place the knight on its natural c6 square, but on its best square: d5. There it is invulnerable to attack from a pawn. "
And stopped. Alpha has discovered and given to the waiting world Knight outposts - Hallelujah.
Sometime later somewhere I posted something - it's hard to keep track of all my gibberings-
about all Alpha is doing is re-inventing the wheel nothing more.
But I was wrong regarding Alpha. Especially with my no hope comment because it does have hope. (fishing is hoping to catch a...stockfish.)
And I will look at other games. Trap is a dirty in GM circles but Alpha has no shame. (no shame with hope coupled with the ability to see deep traps is a wonderful mixture. Players of my ilk play 'Hope Chess') and 27 Re4 was a trap.
They should have put that in the book because it's a wonderful argument and sceptic buster (maybe they saw it but mentioning the word 'trap' might effect sales...Ah good I getting back to the sceptical me.)
Thank you again, very kind, but I've been married to Mrs Ge-off for 42 years I can easily smuggle another £20.00 from the joint account without her knowing.
She knows me well so set a limit. If I took £100 I'd spend the lot on chess junk.
Right off now to the book stall. Thank you again for the kind offer...and you have not been mean to me. Being mean to me is accepting my sacrifices and then taking me to a lost ending. You should get into the spirit of the game and sac-back. I do!
Forum insults, huh! happens everyday. Like Carlsen - see post above. I shrug my shoulders and move on. I only get a tad worried if I think I've actually upset someone in error. People here need to take chill-pill before replying or posting.
|Jul-11-19|| ||MissScarlett: <Miss Sally> is inured to being mistaken for a charity case.|
|Jul-11-19|| ||MrMelad: Hi <Sally>, thanks for your comments. |
In the same manner Carlsen benefited from studying Lasker/Capablanca/Fischer/Kasparov he may have benefitted from AlphaZero and Leela. Analogous to your example he may sometimes think to himself “what would Kasparov do” and realize he is not Kasparov and play his own move instead. How is this different?
But more than that, maybe playing through an AlphaZero game he got reaffirmation for his own ideas? Maybe he recognized strategies he wasn’t sure about? I don’t know but I think it is likely. “Inspired” is a really strong word.
People always say “if Fischer had access to computers he would be a better player” and I think it is a logical argument, so why is having access to games of the best chess player of all time not improve the best ever human player of all time?
|Jul-11-19|| ||Sally Simpson: ***
Carlsen is/was a gifted child prodigy. He had a huge leg up with some excellent coaches including Simen Agdestein and later Kasparov but the natural talent was always there. He is without doubt up there with Kasparov Karpov and Fischer and he was before Alpha.
I do not think 'Skimming' actually qualifies anyone, including Carlsen, to say the day after he remembered he skimmed it, it was inspirational. (you could see him struggling for a word.) and everyone, including the ad-men have seized it.
If Carlsen was in the middle of an out of form run then it would have been dropped. (can you imagine the inspirational jokes from the punters here...remember here are the people who buy the books)
No doubt other players were asked about 'The Book' and maybe gave an in depth analysis having actually read it. But they are not Carlsen.
Also if it is so good why are the other players not getting inspired?
Possibly because Carlsen is and has been for the past 10 years the better player.
Had the book in my hand today.
https://www.redhotpawn.com/imgu/blo... (put on Beatles t-shirt to cheer up harrylime.)
Did not buy it, skimmed through it...(not a charity case Miss. S. did not want K.P. to waste his money on a book I'd never do justice to.)
Was told by good authority it was going to be the book of the year. I expect so. You cannot buy the one word publicity 'inspirational' from a World Champion. (though we were wondering how much the girl had been paid to ask the question.)
The 'We' in this case was a bunch of experienced players (fellow sceptics hanging around the bookstall) who have been burned in the past by the computer hype in chess books. Put that down to previous books on computer games and books written by computers.
Houdini's book on Aronian Best Games - look at it:
What a waste of paper. Who is going to go through that and be inspired. The whole book is like that.
Anyway...I found myself telling them this Alpha is different, which got a few strange 'is this Geoff talking' looks.
Wait till they find out 'Game Changer' is actually 'Geoff Chandler' and it's my best games.
Mathew and Natasha are touring the country giving lectures about the book. Don't know if they are coming to Scotland. If so I'll go...Then I'll get the book, get it autographed and flog it for an inspirational fee on E-Bay!
More interested in Demis Hassabis https://en.wikipedia.org/wiki/Demis... the guy behind Alpha.
Apparently he was playing at a congress, saw all the heads bowed thinking and thought what a waste of human brain power and quit chess.
So behind it is an out-of-the-box thinker. These guys are good for mankind. Hopefully he will channel his extraordinary mind into saving the planet from ourselves rather than fannying about trying to solve games.
|Jul-11-19|| ||Sally Simpson: ***
Sorry mate, did not buy the book. (see above).
Instead I got 2nd hand 'Soviet Chess' by D.J.Richards. (skimmed though it!....looks good and I expect soon to be inspired by it.)
Chess Terminators 'the rise of the machines' by Ray Keene.
Like Keene's books (even though he made a copy and paste blunder on the last W.C.book. Why? he can write entertaining stuff when he wants to.)
Got that because it was dirt cheap and as I'm soon to become an Alpha champion better check out the also rans.
Prefer to do it with a book and board rather than go over a game on the screen. Am going to print off that game you link to (submit it) and a few of the Alpha games.
Also might help with the jargon. What does any of this actually mean: AlphaZero (Computer) (kibitz #631) - suspect AylerKupp is a marooned alien who cannot get home.
,,,and a Fischer Cup and a chess pen which the guy who sold it to me it says was actually used by Karpov in the Moscow Marathon. I believe him...there is no ink left.
Going back in a day or two to unload all the good 2nd books doublers I have for credit - may swap it for 'The Book'. I have 3-4 Fischer's 60MMG's and doublers of other good books I picked up for £1.00 to 50p over the years.
|Jul-11-19|| ||keypusher: <MissScarlett>|
OK. I leave to your conscience whether to tell Mrs. Simpson-Chandler that your foolish pride is either costing her 20 BPS or you a good book. The offer stands, anyway. I'd even be willing to buy the copy for her instead of you.
I'm going to take issue with a couple of your comments, politely.
<".... Black is not planning to place the knight on its natural c6 square, but on its best square: d5. There it is invulnerable to attack from a pawn. "
And stopped. Alpha has discovered and given to the waiting world Knight outposts - Hallelujah.>
That's not fair. I think Sadler and Regan sort of struggled to figure out who their audience was for the book and in particular what level of chess player they should be writing for. Sadler seems to have settled for dispensing sound, basic strategic advice/commentary together with mostly computer-generated analysis. Hence comments like "A0 puts the knight where it can't be attacked by a pawn."
(Minor point: I don't think d5 for the knight counts as an outpost, since it's not protected by a pawn. Wikipedia agrees, and for all we know FSR wrote the article.
One thing I liked about the book is that Sadler pretty much always explained what was going on, and avoided the trap of shoveling in long engine lines that leave an amateur wondering at every move why any of a half-dozen alternatives couldn't have been played instead.
IMO the hype in the book is not too bad, and in particular Sadler writes somewhere that A0 didn't come up with anything completely new, but just synthesized existing attacking ideas with great skill. (A0 does seem to have decided that the minority attack is ineffective as well as boring -- see AlphaZero vs Stockfish, 2018. That's the sort of conclusion I'd expect you to get behind.)
I think MC has praised A0 more than once (not going to try to back that up with links). One thing you realize if you watch a few interview videos is that Carlsen gets asked the same questions over and over, which probably goes some ways towards explaining his occasionally glazed look on camera.
It's not just that he's been successful this year, it's that he's been notably aggressive, sometimes in what seem to me like rather AlphaZero-ish ways. He even played an A0 innovation, albeit in a rapid game. W So vs Carlsen, 2019.
I'm not trying to oversell this. Carlsen has been a great player for a long time. I do think A0 has had some influence, probably not too much, on how he's playing these days.
If you're interested in how A0 plays, there are links to a bunch of the ones I liked in my forum, if you scroll down. I'm quite partial to this game (not discussed in <Game Changer>. No doubt the opening has a lot to do with that. AlphaZero vs Stockfish, 2018
<More interested in Demis Hassabis https://en.wikipedia.org/wiki/Demis... the guy behind Alpha.>
There's a long interview with him in <GC>. As a kid he seems to have been very good at lots of games, not just chess, which is what helped get him interested in an computer that could be used in multiple settings. He's interested, or says he is interested, in using AI for good in the world, not just to make human chessplayers (and programmers) feel stupid. If I had his brains and Google's money I'd be mining Bond movies for pointers on how to be a supervillain.
|Jul-11-19|| ||thegoodanarchist: <csmath: Those that are promoting AZ have gone as far as to claim that Magnus Carlsen improved by using it even though Magnus did not mention it>|
This is the first I've heard anyone had claimed that Carlsen had improved by playing A0. Do you have a source?
IIRC Carlsen <has> said some nice things about AlphaZero, and a lot of people (including me) have speculated that his aggressive play this year shows A0's influence.>
On Riga Grand Prix website it is reported that
<Daniil Dubov, who is famous for his win in the Rapid Championship and a praise from Magnus Carlesn (the Champion said that he has two heros, Alpha Zero and Dubov) is giving its second try to preveail in the event.>
|Jul-11-19|| ||scholes: One thing that leela chess zero and alphazero have shown that it is possible to play very aggressive against very strong taticians without taking undue risks. |
But there's no proof Carlsen is training with leela. In fact Carauna used leela for his preparation for world Championship match.
Since then leela has grown much stronger only.
|Jul-11-19|| ||AylerKupp: <<Sally Simpson> Carlsen added he was thinking during his game how would Alpha Zero have played this position. He closed with I am not 'Alpha Zero' so played something else. So much for being inspired.>|
Maybe he was just being pragmatic. He knows that he can't calculate like Alpha Zero or, for that matter, any other computer engine. So if he has to make a choice between a line that provides a winning advantage but requires precise calculation and there is a significant likelihood that an inaccuracy will result in a disadvantage versus a different line that only provides a significant advantage but does not require precise calculation and has a small risk of not being correctly played, the pragmatic thing to do is to select the second line, particularly if one is short of time. You of all people should agree with that approach.
<I could eaily afford the book and if there is nothing better I may get it. My objection is paying all that money for what looks like 50% of the stuff would never get read.>
Well, I think that the true measure of a book is what you get out of it from the 50% of the stuff that you do read, not from the fact that you might only read 50% of the stuff. And, who knows, you might actually read the other 50% if you like the 50% that you do read.
I think that you will like the book, I did. I think that you would particularly like the 18 games providing historical parallels of how Alpha Zero played its games compared to how the players, 9 of them world champions, played their equivalent game. But, who know, you might even appreciate the short chapters on how Alpha Zero works. And Mathew Sadler is a good analyst.
|Jul-11-19|| ||john barleycorn: honestly, this one got me confused/brought me to my limits:|
<AylerKupp: ... So if he has to make a choice between a line that provides a winning advantage but requires precise calculation and there is a significant likelihood that an inaccuracy will result in a disadvantage versus a different line that only provides a significant advantage but does not require precise calculation and has a small risk of not being correctly played, the pragmatic thing to do is to select the second line, particularly if one is short of time. You of all people should agree with that approach. ...>
< but requires precise calculation and there is a significant likelihood that an inaccuracy will result in a disadvantage> what does it take to figure that out?
<versus a different line that only provides a significant advantage but does not require precise calculation and has a small risk of not being correctly played> ah ha.
How do you recognize a line that provides a winning advantage but requires precise calculation? And then give it up for a line that provides significant advantage?
<Sally Simpson> once mentioned he gave up the booze (which I do not believe) but when did you start drinking hard?
|Jul-11-19|| ||keypusher: <jbc> It's simple enough. In AK's scenario, Carlsen is choosing between a really sharp line that looks winning but if he screws up or there's a hole in his analysis, he's probably going to lose, versus a line where he has a clear advantage, but there may not be a win there, even with best play.|
|Jul-11-19|| ||john barleycorn: <keypusher: <jbc> It's simple enough. In AK's scenario, Carlsen is choosing between a really sharp line that looks winning but if he screws up ...>|
Not quite. it is not about lines that look winning. The complete nonsense gets obvious here " a line that provides a winning advantage but requires precise calculation and there is a significant likelihood that an inaccuracy will result in a disadvantage". How can that be a winning line?
|Jul-11-19|| ||AylerKupp: <<john barleycorn> How do you recognize a line that provides a winning advantage but requires precise calculation? And then give it up for a line that provides significant advantage?|
I don't recognize it but GMs certainly do. Time and again when I'm watching live commentary of games, by GMs, they allude to the player choosing a less promising line than the computer's suggested best line because it also looks <likely> to win although they can't calculate it all the way to the end. But it looks simpler and safer.
It's all about risk management. You can play a risky line that looks like it will win in the short term or a less risky line where the win is likely in the long run but not certain. And, of course, in a complex line there is a higher probability that your calculation was not correct, particularly when you are short of time.
It's similar to accepting a draw in the last round of a tournament when you are in first place, have a ½ point lead, a better position in your game, and your primary contender has an inferior position and is likely to get no more than a draw. Why jeopardize your first place in the tournament by trying for a win (and possibly losing) even though you have an advantage in your game?
I'm surprised that you don't seem to consider practicality and risk management as pragmatic approaches in chess.
And, no, I haven't started drinking hard. As you can see from my forum's header I'm mostly interested in wine, although I will have a strong drink on occasion. In fact, I've lowered my alcohol intake in the last 1 ½ years when I started my diet.
|Jul-11-19|| ||Sally Simpson: Hi K.P.
To be fair I thought about the 'Hallelujah' post BEFORE seeing this game and as a...ahem...fellow author I can relate to Sadler that he will have to explain obvious to us details for the none or casual players who are interested. I was just having my usual rant.
And yes Nd5 may not be a true outpost. If the Black e-pawn goes then the e3 pawn can kick the Knight whcih is not protected.
Regarding the minority attack, you are correct. They are great on paper and very easy to note up a successful one where Black just rolls over and lets it happen. Good to show a student a middle to end game plan in action.
Chernev gives possibly the most famous minority attack in history in his Most Instructive Games. Larry Evans vs H Opsahl, 1950 A very good game. great book. I once gif'd the famous cover to make the people move. https://www.redhotpawn.com/imgu/blo...
I know the game Stockfish vs AlphaZero, 2018 I posted there, it's when I first started to look at Alpha games with a different view.
Mathew has always been an entertaining and talented player.
I recall he was away from the game for years, then came back to play in a small tournament in Holland perhaps to see if he still had the knack.
Unfamiliar with current opening trends he open as Black with 1...a6 2...h6
click for larger view
C van Oosterom vs Sadler, 2010
He won that, so decided he did not need to book up and off he went to Oslo and came first! Oslo Chess International (2011)
Looking at his games...Maybe he inspired Alpha.
I want to find my own Alpha games to look through but I ran through without blinking AlphaZero vs Stockfish, 2018
and stopped...I'll post there.
Following a conversation today I understand Demis Hassabis is maybe going to use Alpha technology to delve into medical issues. (it's where I got my please the planet from.) Think I'd prefer Stockfish for that. Alpha would sac my left arm to save my right leg.
|Jul-12-19|| ||FSR: <keypusher> Not this time. If you look at the revision history of the Outpost article in Wikipedia, you'll see that Krakatoa (me) played no role. https://en.wikipedia.org/w/index.ph...|
|Jul-13-19|| ||MrMelad: Hi <Sally>, thanks again for your comments.|
I agree with you about Carlsen’s talent, I was <inspired> after he won group C or B at Wijk Ann Zee. I remember thinking he was the most talented player since Morphy. :)
<Also if it is so good why are the other players not getting inspired?>
Are you sure they aren’t? Maybe they just don’t get asked.
<Houdini's book on Aronian Best Games - look at it:
What a waste of paper. >
That’s hilarious! Can’t believe it’s a real book.
<Anyway...I found myself telling them this Alpha is different, which got a few strange 'is this Geoff talking' looks.>
Alpha and Leela really are different. While stockfish is a (state of the art) chess algorithm it is based on manually (or humanly) defined evaluation method that rely mostly on brute-forcing as many positions as possible to asses it’s eventually scalaric score.
AlphaZero is different in the sense that no one taught it strategies or tactics, it was only given the rules of the game and played with itself 40 million games to <learn> his own evaluation method. Even if it weren’t so strong it would be interesting to study what it had learned.
It also makes for a totally different playing style than traditional engines as it rely mostly on pattern recognition and not strictly on ply depth. It can evaluate complex positions and openings from a strategic point of view, where as stockfish opening strategies result from human programming and some arbitrary numeric balance.
Another interesting element of reinforcement learning is that it can recognize similarity between different positions (states) in terms of optimal decision making. So somewhat similar to a human it has experience memory and long term memory that helps it recognized that similar (and not necessarily identical) decisions leads to optimal strategies.
In chess terms I’m guessing it translates to long term positional sacrifices, and we can indeed see in AlphaZero and Leela games amazing piece sacrifices which gets a whole lot more impressive when played successfully against the mighty stockfish.
And it can also lead to superior opening play as it can use it’s experience to “understand” the position in deeper strategical manner.
Or at least this is my take on it..
< Earlier Kibitzing · PAGE 2 OF 2 ·