GameKnot: GameKnot related, What does this mean?

GameKnot related: What does this mean?

From

Message

nathanman22
05-Jun-11, 15:34

What does this mean?

Tournament rules change: if you play in GK tournaments, you are probably aware of the rule that the final 1% of any remaining unfinished games in a tournament section are automatically declared as draws in the 1st and 2nd rounds. It is designed to speed up the tournaments, as the final 1% of the games routinely take months to complete, holding up the entire tournament. Naturally, simply declaring games as draws is not ideal.

The good news is that we have now implemented an automatic adjudication system that uses a chess engine to analyze the current position on the board for such games and decide on the likely outcome. Please note that this adjudication only affects the tournament table (and who will advance to the next tournament round) — the games remain active and can be finished normally, and will affect the players ratings and stats accordingly at the time they are finished (but not the tournament table any more). This change comes into effect today and will be applied to any new tournament rounds reaching 99% completion as of today.

Good luck and thank you for choosing GameKnot as your premier chess battlefield!

nathanman22
05-Jun-11, 15:35

confusion

Does this mean that the games will remain active and the tournament will not move forward at 99% until these particular games are finished, or that these games will be declared wins rather than draws for the player currently at advantage. How do they determine what games will not be considered draws or are all games determined to be won by either one party or the other?

Any other questions and clarifications are welcome. Thanks.

-Nathan

levellerlevvie
05-Jun-11, 23:23

As I see it ...

... they will calculate a certain amount of moves in the future like any chess-engine would (20 - 25) to determine the most likely outcome of that board. Probably they will set a threshold deciding whether it will be a draw or a win .. for eg 3 points ahead or more will win the game, less then 3 points will result in a draw.

Well I think it's better then all draws anyway so this change is fine by me

kingdawar
06-Jun-11, 07:02

This might lead to comical situations and declared outcomes in the casual section... but ok, it won't affect the progress of the games themselves.

rookie879
06-Jun-11, 08:29

Well hopefully they just don't go by material, especially in endgames where the side of lesser material has a better chance for a draw than the middlegame.

Hypothetically say a game ended in a tournament where White had a Queen and a King and Black only had a King and a pawn. Now, as lopsided as it may seem, there are some positions where White can't force the win and the game ends in a draw. Hopefully the CPU will be able to distinguish this and award a draw accordingly

nathanman22
06-Jun-11, 08:41

annotation of a game I won

Speaking of tournament games, I played a very interesting one that recently finished. I am looking for an experienced player to annotate to help me learn how to improve my gameplay. Is there anyone who could help me with this endeavor? I'd really appreciate it.

game

Nathanman22

maca
06-Jun-11, 10:49

This feature doesn't really touch me personally because I don't play in the GK Tournaments anymore, but I'm not sure how well it is going to work in the practice. At least I'd like to hear a few more details about the logic applied in deciding the outcome of the games. The problem is that while there are clear "stalling" situations in the games played between the lower-rated players that the engine is probably able to evaluate correctly, the differences in score become smaller and smaller as the players get stronger and stronger. Is there a fixed threshold of score, above which the game is won, and below which the game is declared a draw?

I really don't have confidence that the GK engine is able to evaluate games between higher-rated players correctly, given that its strength was evaluated to be above 2000 when it was implemented. Of course, a very complete tablebase might help to evaluate "tricky" endgame situations, as the one described by rookie879. But, that's only for endgames, and tablebases assume perfect play from both sides... Regardless of how the system is executed, I foresee it to cause probably as much as discussion as the stalling issue it was developed to remedy. Even if the computer finds a victory in a position, one can always ask whether the human player would have found that same solution.

Regards,
MaCa.

liam1260
06-Jun-11, 11:16

I had the same quetsions when I forst read the notice.

Are all games going to be 1-0 or 0-1?

If draws are to be declared what would the criteria be. If it is based on Material/Positional points then I would hope the Causal and Intermediate sections have a very generous spread 5 or even 7 points before a win is declared. 'cos I have played in both Levels over the recent months and the opportunity for a player to blunder is quite high.

baronderkilt
07-Jun-11, 08:12

MACA

I think you are right in questioning the strength/ability of the GK Chess Engine, at least as it is used to provide post-mortem analyses. I and another player who is over 2100 here were both questioning that and agreed that in a couple of our games (EG He saw mine computer analyzed too) the computer missed the mark enough to make considerable difference in the outcome of that line given. (As much as a 1/2 point worth I believe.)

It not being an evaluation problem of that program however, rather was the Horizon effect causing it to fall short of the more correct evaluation had it looked a move or two further. That being the case, then presumably it could be told to look deeper in evaluating these games for adjudication. But there still always arises the problem of Just How Deep is Deep Enough. (And still the Question also, Just How Strong is Strong Enough, if it only rates out at 2000ish )

Perhaps there could be an Appeal mechanism for a player to use if they really felt the Adjudication was a 1/2 point off the mark ?

Gameknot.com
07-Jun-11, 09:33

Because the chess engine has to analyze every single move in a game for post-game analysis (i.e. 80 positions total for a 40-move game), we can only allocate 10 seconds of thinking time (for premium memberships, 5 seconds for free accounts). Which generally allows the engine to analyze positions 14-20 moves ahead/deep. When adjudicating the last 1% of games in a tournament, only one position needs to be analyzed, so the engine is allowed to use up to 60 seconds, which means its playing strength is increased significantly. And because the vast majority of such games are already in the endgame with only a few pieces remaining on the chessboard, the chess engine is routinely able to analyze 30-40 moves ahead/deep. If anyone appeals the adjudication decision, we'll gladly re-run the analysis with even more increased thinking time (an hour?).

nathanman22
07-Jun-11, 09:37

gameknot

I think this is a great idea. I think it solves the problem of stalling and prevents those who are stalling from being promoted unfairly. I appreciate this move as I think it just is another step in making gameknot greater! I love this site. Now if you could only offer lifetime memberships again so I could get one! (:

-Nathan

levellerlevvie
07-Jun-11, 10:07

Gameknot ...

... any thoughts on how it will be determined whether the analysed position is a draw or a win? I assume not all boards will result in a winner, right? So ...

For eg after 1 min of thinking time the engine gives white a +3 advantage. Is that enough to declare white as winner? Or will it be a draw?

Also ... When a decision is made after analysing ... will the players be able to see the result of the analysis? I'm not sure I would like the situation where my opponent is declared the winner while I'm not smart enough to see why ... The analysing score would at least give an idea how much difference there was.

Anyway ... allthough I support the implementation ... some aspects of this feature are worth debating about I think

Gameknot.com
07-Jun-11, 12:31

The draw threshold is currently set at +1 for the Grandmaster section, +1.5 for the Expert section, and +2 for all lower sections. We might adjust/tweak it in the future of course, as necessary.

No, other than the adjudication decision (i.e. win/loss/draw) and at which move it was made, no other information is provided to the player. After the game is over, you can certainly contact us for more info, or simply request a post-game analysis to find out the computer evaluation at that move (it will be the same chess engine doing the evaluation).

jotheblackqueen
07-Jun-11, 14:52

King, Bishop and Pawn v bare king.

Hi Nathan,

I have just taken a look at your game in which your opponent's king copped Old Matey in the corner in the above mentioned endgame. Here is an important rule to remember - everyone. If your opponent has an h pawn or an a pawn then the game is a draw by stalemate or insufficient mating force if your opponent's bishop is of the OPPOSITE COLOUR of the squares a1, a8, h1 and h8. If the bishop is the same colour then the defending side with the bare king will cop Old Matey as shown in the final position of your game.

The a pawn and h pawn theory also crops up in K + Q v K + Passed Pawn on the 7th rank. If the pawn is an a pawn, a c pawn, an f pawn or an h pawn on the 7th rank and the defending king is able to defend the pawn then the game is drawn! Here is an instance of that happening. Be interesting to see what gameknot would have made of the closing stages of that game!

gameknot.com

In the cases of the drawn endings the stalemate factor comes into play. It does not with the b, d, e or g pawn on the 7th rank scenario so if you, the defending side, have the passer on the d file etc. then alas, your position is lost although it may well pay to test your opponent out in these circumstances.

Anyhow, congratulations on your checkmate and many thanks for showing us this piece.

Hope you have a nice evening.

Bye for now.

Joanne

kingdawar
07-Jun-11, 14:53

Install/integrate tablebases for the GK analyzer, and it should work ok.

lostkarma
12-Jun-11, 15:36

Deleted by lostkarma on 14-Sep-11, 18:10.

liam1260
12-Jun-11, 16:10

Is this new rule, about wins awarded within the last 99%, come in from immediate effect like ANY current rounds that have yet to reach 99%?

Or current Round one and Round Two competitions?

Or future Tournaments that have yet to start?

Thanks in advance

baronderkilt
13-Jun-11, 11:16

Excellent GK~!

That is reasuring. Makes sense to me. And there must be like 80 bazillion positions in an hour. But I did have this one position come up in my last GK tournament that i never did get a grip on, and might well have put me into the last 1%, but I pressed on, stumbling thru the rest of the game, trying to make some sense of it. I'd sure like to get maybe 20 minutes on it. mmm ... 22 minutes?

thereaper1
14-Jun-11, 18:15

Psshhh thats an easy position. I see a mate in 250.

baronderkilt
14-Jun-11, 22:42

Yes , yes !

I am sure you are right, Reaper. And I know it is not a horizon problem since I can see well beyond 300 moves ahead. The trouble is, all my pieces are gone before move 200 ! (and Still gone beyond move 300 ) . So its obvious I am simply making a mistake somewhere, and it may well be the first move. I've tried everything but h4, do you suppose ... ?!

« Back to forum