|
bitcoin mining but for correspondence chess
|
# ? Jan 27, 2021 08:06 |
|
|
# ? Jun 12, 2024 15:08 |
|
Hand Knit posted:Databases have always been permissible as part of correspondence chess, since there was really no way to prevent them from being used. Similar reasons are leading organizations to make engine use permissible. At this point, it's just part of what correspondence chess is. Sub Rosa posted:ICCF and LSS both allow engines. ICCF is the FIDE recognized authority on correspondence chess for what that's worth. Just to be clear, lichess and chess.com do consider engine use cheating, regardless of time controls. chess.com posted:In Daily Chess (turn-based games with several days per move), you may consult any resource which is not engine-based. This includes books, opening databases (including the Chess.com Explorer) for [https://www.chess.com/openings|opening moves], and thematic games (though not their engine analyses). Tablebases are NOT allowed. You may not consult an engine, or another human, to provide an opinion on the opening database, tablebases, self-preparation or analysis that would relate to a particular game-in-progress on Chess.com. lichess posted:We define this as using any external assistance to strengthen your knowledge and, or, calculation ability to gain an unfair advantage over your opponent. Some examples would include computer engine assistance, opening books (except for correspondence games), endgame tablebases, and asking another player for help, although these aren’t the only things we would consider cheating. If they are moving towards allowing engines in daily games, I haven't noticed it. Maybe they could add a 'computer-aided' as a separate category but it would surprise me greatly if they simply allowed it in Daily games. It would effectively kill them because who'd play mostly human moves and only some computer moves? If you can use the computer then you'll use it every turn because your opponent will too. And that's not what the vast majority of players would consider "playing chess".
|
# ? Jan 27, 2021 16:15 |
|
Where does using Chess.com's Analysis Tool while in game fall? I'm talking about the one that let's you move the pieces around to visualize lines. It's not available in blitz but I use it all the time in correspondence.
|
# ? Jan 27, 2021 16:23 |
|
former glory posted:Do official correspondence games have separate categories for human only play and ones that include engines? In the US you can find rated CC that doesn't allow engine analysis offered by USCF, but I've no experience with it. I've heard that cheating is very common in USCF CC and it's very hard to prove your opponent is using an engine considering the time control if you do suspect your opponent. And if you can't get an arbiter to agree with you ever, it's sort of actually practically allowed? Which generally is why I prefer playing where it is explicitly allowed. former glory posted:I don't see how it doesn't turn into an escalating war of people throwing money at stockfish instances. It's much more about having a lot of meta knowledge about chess engines. NNUE has changed things, but if we go just before we had Stockfish vs Leela as an example. Someone just doing analysis with one instead of both would be at a disadvantage. Easy to select a move if they agree, but when they disagree it's like asking a man with two watches what time it is. So in practice you end up doing things like having engines check other engine's candidate lines, not just blindly playing Stockfish's top move at any arbitrary depth no matter how many cores you throw at it. And all that is before recognizing that deciding which network to use with Leela is important, or that Stockfish is benchmarked for performance at lower time controls so gains in strength in lower time controls may not only not mean that it isn't stronger for infinite analysis, it could be weaker. So in practice there is a world of variant engines based on the more commonly known ones that may actually be stronger.
|
# ? Jan 27, 2021 16:59 |
butros posted:Where does using Chess.com's Analysis Tool while in game fall? This is one where I prefer to work on my mental memory muscles, since I want for this to be helpful for OTB tournaments at some point in the future. Speaking of which, my club is considering a move to downtown since our core has been gutted by the collapse of the tar sands market and huge square footage can be had for pennies, which will make it a short walk once Covid times are done. Thanks for clarifying the point on engine use OP.
|
|
# ? Jan 27, 2021 16:59 |
|
Honest question from a bad player: What is the point of choosing moves proposed by an engine/multiple engines? Aren't you just making worse moves that way? Curating moves doesn't make a lot of sense to me, since the engine is that much stronger than any human player.
|
# ? Jan 27, 2021 17:15 |
|
I suppose the practical answer is if they were worse moves I would be losing games to my opponents instead of winning them?
|
# ? Jan 27, 2021 17:19 |
|
What quality of technology do you need to play high level correspondence chess? I ask because I just saw that Pino Verde is a CC IM. And, like, if he's one. Then why shouldn't I put in the time?
|
# ? Jan 27, 2021 17:21 |
|
Sub Rosa posted:I suppose the practical answer is if they were worse moves I would be losing games to my opponents instead of winning them? So you'd be winning against an engine?
|
# ? Jan 27, 2021 17:23 |
|
I have one IM norm myself, and as soon as a game I'm winning finally ends I will have my second, and my first SIM norm. Two norms is all you need on ICCF, but I won't meet the number of games in norm winning events requirement until I get my third. I'd suggest in terms of technology a reasonably modern PC with an Nvidia graphics card. The more cores in the CPU the better. I have an AMD 3900x so 12 cores. But until a couple months ago I had the 3600 with 6 cores and it hasn't made that much of a difference. You want an Nvidia GPU for CUDA for lc0, though NNUE may mean that Leela is less important in the meta. Then optimally you will have 16TB of disk space to have a local copy of the seven-piece syzygy tablebase. I feel pretty certain I've won games because I had it and my opponent only had a six-piece, and it's something I would say is much more impactful than having a few more cores or the latest graphics card.
|
# ? Jan 27, 2021 17:33 |
|
Yeah the thing about the 7-game data base immediately stuck out to me. A desktop computer is probably out of my reach for the time being, but I'll keep it in mind.
|
# ? Jan 27, 2021 17:42 |
|
teppichporsche posted:So you'd be winning against an engine? I will have won 4 and drawn 6 in a Cat 4 event (which means average rating is between 2326-2350). And players are unlikely in my opinion to reach this rating by blindly copying Stockfish's move.
|
# ? Jan 27, 2021 17:43 |
|
Here is a position from a game I won: Black to play and win. Here is Stockfish's top five move suggestions at depth 52: The move I played is not on the list.
|
# ? Jan 27, 2021 18:01 |
|
I would guess 39...Bxh4 40.gxh4 Rxf4 with the idea that white will not be able to extricate the bishop on a1 and black will be able to create the persistent threat of Rxa2 preventing white from untangling.
|
# ? Jan 27, 2021 18:12 |
|
Correct. As a matter of fact Rxa2 was the final move before white resigned.
|
# ? Jan 27, 2021 18:15 |
|
I'm curious what your computer use process was for getting there.
|
# ? Jan 27, 2021 18:18 |
|
Very cool. I would not have thought of Bxh4 that as I would be more inclined to conserve material but I can see how you're actually trading the bishop for three pawns and opening up the H and G files for your pawns to push
|
# ? Jan 27, 2021 18:22 |
|
I knew white's bishop was entombed which in general made me think about bishop sacs. I saw that white's rook would be stuck on the h file protecting the h pawn, and with my rook safe on g4 white would be able to make no improvement at all while I took as many moves as needed on a king walk. So I put the move on my analysis board to see what the engine's continuation would be. Basically starting with the presumption that I must be missing something about why the move was dubious, I asked the engine to refute it. Instead of refuting it, the engine immediately realized it was winning.
|
# ? Jan 27, 2021 18:30 |
|
Nice finish to Giri-Wojtaszek 49.Rxh7 1-0 (49...Kxh7 50.Rd7+ Rf7 (Kg8 51.h7+ Kh8 52.Nxg6#) 51.Rxf7+ Kg8 52.Re7 Rb6 53.Kxf3)
|
# ? Jan 27, 2021 19:55 |
Something that is becoming very apparent from my last two games is that I have to start thinking about chess being more a long battle of attrition than shocking moves of penetrating insight. This realization also comes from the puzzles I have been doing. If I have an opportunity to trade one piece for two pieces, unless the result leads to a mate thread against its probably the best move.
|
|
# ? Jan 27, 2021 20:00 |
|
Trading one piece for two is closer to a shocking move than to a long battle of attrition.
|
# ? Jan 27, 2021 21:16 |
|
Sub Rosa posted:Here is a position from a game I won: Interesting stuff, thanks.
|
# ? Jan 27, 2021 22:05 |
|
Sub Rosa posted:
Thanks for sharing this with us. I'm completely fascinated that it works this way. I was skeptical, thinking that the competition would essentially be a secondary Top Chess Engine Competition, but I see now how that's its own thing. You're mastering the tools of a craft! teppichporsche posted:Honest question from a bad player: What is the point of choosing moves proposed by an engine/multiple engines? Aren't you just making worse moves that way? Curating moves doesn't make a lot of sense to me, since the engine is that much stronger than any human player. My understanding from Sub Rosa's posts is it's a different level of competition, but involves human ingenuity and intuition as much as over the board chess, just different aspects. It's like you give two guys a block a wood and a knife and ask them to whittle out a figurine. And then in the other room, the woodworkers are allowed to bring all manner of lathes, CNC machines, full CAD design, etc, to make the same sort of figurine. In both cases, you're going to have one guy's figurine look better than the other guy's, but there's going to be a whole world of a difference in precision. Freaking cool.
|
# ? Jan 27, 2021 22:10 |
brb repurposing my Xenon cluster to play chess
|
|
# ? Jan 27, 2021 22:21 |
Hand Knit posted:Nice finish to Giri-Wojtaszek what was up with 19. Bxb6?
|
|
# ? Jan 27, 2021 23:35 |
|
Bilirubin posted:what was up with 19. Bxb6? Whtie has pawns on a3, b4, and c3. The N on b6 plays an important role controlling a4 and c4, effectively neutring white's queenside (and white would love to use the c4 square). Also, with black's pawns on f6 and e5, white would love to wrest control of the light squares. Note how Giri breaks through first by manoeuvering one rook through e6 and then focusing down the pawn on d5.
|
# ? Jan 27, 2021 23:43 |
Hand Knit posted:Whtie has pawns on a3, b4, and c3. The N on b6 plays an important role controlling a4 and c4, effectively neutring white's queenside (and white would love to use the c4 square). Also, with black's pawns on f6 and e5, white would love to wrest control of the light squares. Note how Giri breaks through first by manoeuvering one rook through e6 and then focusing down the pawn on d5. Makes sense. I also really liked the rook dance that happened to avoid attacks and gain the desired position. At that level it looks like an entirely different game yet its exactly the same in another way. Thanks for posting these highlights from the Tata I find them quite interesting
|
|
# ? Jan 28, 2021 01:06 |
|
Sub Rosa posted:Here is a position from a game I won: Thats wild, i saw the bishop sac right away, not because i knew it was good, but rather "what the heck else am i supposed to do here?" and like you said, stockfish didn't see the move at all. but as soon as you play the move, stockfish is like "holy poo poo, this is a good move for black!" that's a heck of a blind spot, stockfish!
|
# ? Jan 28, 2021 01:43 |
|
Helianthus Annuus posted:and like you said, stockfish didn't see the move at all. but as soon as you play the move, stockfish is like "holy poo poo, this is a good move for black!" Yeah I don't really get that. If Stockfish is analyzing the full tree of moves to 20+ moves out, then how is it that it can't see the value of a move beforehand but then recognizes it right away afterward? Does it automatically ignore certain lines that appear to lead to a material disadvantage? You won't see Stockfish miss a queen sac that leads to a mate in 10...
|
# ? Jan 28, 2021 02:34 |
|
Carbolic posted:Yeah I don't really get that. If Stockfish is analyzing the full tree of moves to 20+ moves out, then how is it that it can't see the value of a move beforehand but then recognizes it right away afterward? Does it automatically ignore certain lines that appear to lead to a material disadvantage? You won't see Stockfish miss a queen sac that leads to a mate in 10... Tying everything together, that’s a prime example of why correspondence chess isn’t just “let the engines make all of your moves”
|
# ? Jan 28, 2021 02:48 |
|
Carbolic posted:Yeah I don't really get that. If Stockfish is analyzing the full tree of moves to 20+ moves out, then how is it that it can't see the value of a move beforehand but then recognizes it right away afterward? Does it automatically ignore certain lines that appear to lead to a material disadvantage? You won't see Stockfish miss a queen sac that leads to a mate in 10... A mere six move variation is in to millions of possible sequences. One of the most important features of an engine algorithm is "pruning" evaluative branches, usually by those branches being some threshold worse than the top lines after so-and-so number of moves. This creates a kind of horizon effect, where if the "point" of a line hits too far in the future an engine just won't "see" it from that far out. One of the points of playing through a line, then, is to see if, as you get further along, you get close enough for the engine to see the end of the line. You can actually see this happening on a smaller level on chess24's inbuilt eval sometimes, especially when they're hosting a ton of games like in the Olympiad. Playing through an analysis line will have the eval fly back and forth between +2 and -2 based entirely on who has made the most recent move.
|
# ? Jan 28, 2021 03:02 |
|
Carbolic posted:Yeah I don't really get that. If Stockfish is analyzing the full tree of moves to 20+ moves out, then how is it that it can't see the value of a move beforehand but then recognizes it right away afterward? Does it automatically ignore certain lines that appear to lead to a material disadvantage? You won't see Stockfish miss a queen sac that leads to a mate in 10... you can see more examples of this on youtube with people playing weird and novel gambit opening lines. sometimes stockfish’s evaluation swings around wildly as moves are made it wasn’t expecting. i think this video shows it off https://www.youtube.com/watch?v=vnX6Y9QNeTo
|
# ? Jan 28, 2021 05:03 |
dupersaurus posted:Tying everything together, that’s a prime example of why correspondence chess isn’t just “let the engines make all of your moves” No but if you are using it to try out several options to see how things go several moves forward that deficiency is overcome largely. I mean I get its a thing that people are doing and whatever you do you but to me it just removes part of what chess is to me which is working to overcome the limitations of your own brain. e. or as HK put it much better Hand Knit posted:A mere six move variation is in to millions of possible sequences. One of the most important features of an engine algorithm is "pruning" evaluative branches, usually by those branches being some threshold worse than the top lines after so-and-so number of moves. This creates a kind of horizon effect, where if the "point" of a line hits too far in the future an engine just won't "see" it from that far out. One of the points of playing through a line, then, is to see if, as you get further along, you get close enough for the engine to see the end of the line.
|
|
# ? Jan 28, 2021 06:11 |
|
From our earlier discussion about evaluations and horizon effects. Here are screenshots of a computer's first choice line from chess24's inbuilt engine. The evaluation bar is on the left. Check out how wildly it swings.
|
# ? Jan 29, 2021 17:15 |
|
Watching this pup today
|
# ? Jan 30, 2021 21:37 |
|
Verisimilidude posted:
He seems the type to put an extra queen or two on the board when you're not looking. Don't let him!
|
# ? Jan 30, 2021 23:59 |
|
Verisimilidude posted:
Look at that idiot trapped behind his own pawn structure.
|
# ? Jan 31, 2021 02:48 |
Verisimilidude posted:
lol puppers thinks its a knight
|
|
# ? Jan 31, 2021 06:18 |
|
Bilirubin posted:lol puppers thinks its a knight Officially referring to knights as puppers from now on. Ran into this clever boy last night: Seems legit, Jesus! I was a little suspicious when every move he played against me turned out to be Stockfish's best (we were both ranked ~1200), but this archive was just the icing on the proverbial cake. Also I'm FacepalmingPanda if anybody wants to add me. Modal Auxiliary fucked around with this message at 18:10 on Jan 31, 2021 |
# ? Jan 31, 2021 18:08 |
|
|
# ? Jun 12, 2024 15:08 |
I have no idea what you're trying to say with that image.
|
|
# ? Jan 31, 2021 20:17 |