Poll
Would you like the ranking system for November to be only by number of wins?
Yes, I would like the ranking be decided only by number of wins 6
No, I would not like the ranking be decided only by number of wins 9
Total Votes: 15
You must be a logged-in member to vote
New ranking System for the Computer Pool! VOTE!
Posted: 21 October 2015 07:55 PM  
Administrator
RankRankRankRankRank
Total Posts:  395
Joined  2011-11-18

For November, we are proposing a new ranking system in the computer pool.

The ranking will be decided only by number of wins, not by rating.

You can check HERE to see what this month’s ranking would be if we had ranked only by wins.

Please VOTE and let us know if you’d like this ranking system for November 2015.

Profile
Posted: 21 October 2015 10:43 PM   [ # 1 ]  
Member
Rank
Total Posts:  1
Joined  2015-09-28

I see a problem with the number of wins,  one could have a very low rating and have computer playing a lot of games that were not against top players and could win trny.  Perhaps you can up the number of win’s to qualify for prize but let rating be the final determination.  What you dont want is someone not to play all month until the end then come in when others have established rating then their provision rating gets them a higher rating.  Then they just sit on that high rating with only a few wins.

Profile
Posted: 22 October 2015 12:35 AM   [ # 2 ]  
Member
Rank
Total Posts:  6
Joined  2015-05-29

That already happens man. This is not about low-rated players or top-rated players. This is about avoiding draws as having any benefit in the pool - and instead having wins be the only thing that matters. That is the way that it should be. Top players like myself will also face weaker opponents and get wins also. There is absolutely nothing wrong with wins being the only thing that counts. This method also helps to avoid disconnections ruining a players rating… these things do happen, and it hurts.

I was the one who proposed this wins idea. Everybody is using the same two engines in the pool. Getting rid of draws will force ourselves to win games… and give an incentive for people to play here on a daily basis instead of doing the bare minimum number of games and days and then attempting to sit on his or her rating. Give this a shot guys… like in any sport… the only thing that matters is winning… draws are simply too prevalent here… and they can either reward one’s rating or hurt it…

Profile
Posted: 22 October 2015 12:36 AM   [ # 3 ]  
Member
Rank
Total Posts:  6
Joined  2015-05-29

The pool is completely random… you will need to face top players regardless…

Profile
Posted: 29 October 2015 10:43 AM   [ # 4 ]  
Member
Rank
Total Posts:  3
Joined  2015-09-05

well only wins is a bit simple, against a stronger player a draw with black in fact is good,
(and a won game very good) and thus the ‘Elo’ rankings should be calculated; in fact the system
with 0.5 for every won game still is a bit simple, i would propose something as the following:
(in addition to the normal rating calculations, like done this month i.e. October):
After some discussion maybe possible for December, or maybe already November:

A) 0.5 (extra) points for a draw with Black (0 for White, yes, this can cost rating points for White..)
B) 0.5 up to 1 (extra) points for a win with White (similar as now, i.e. done in October)
C) 1 up to 1.5 (extra) points for a win with Black (instead of 0.5 as it’s now)

Not sure yet about the value of parameters B and C
and ofcourse the whole thing can also be scaled (eg 3/5 x A,B,C)

For discussion, on the basis of the October results sofar we could get
an estimate how such a system would look like (the idea is that
it’s better reflecting real strength (based on the inherent advantage
with White),  while also giving bonuses for playing more often,
but maybe not so much as in October (here it was beneficial
to run 24/7, in my system there’s maybe a better balance.

 

 

Profile
Posted: 30 October 2015 02:20 AM   [ # 5 ]  
Member
Rank
Total Posts:  6
Joined  2015-05-29

Great point values bookbuilder; I like them a lot. We need to reward points for drawing as black, winning as black, and winning as white. We should give 1 point in some general sense at least - and make sure that time, strength, and effort is effectively rewarded. Rewarding wins should still be the most crucial part of this system. It is one thing to draw, but winning against the best players in the pool is something else entirely. Therefore, getting hundreds of wins should naturally give a person who plays here 24/7 a good chance at first place. This is what Rayjr has done - and I came close myself. I’m looking forward to seeing changes for November.

We need to reward wins… in computer chess, it is wins that separate the best from the weakest.

Profile
Posted: 30 October 2015 08:35 AM   [ # 6 ]  
Member
Rank
Total Posts:  3
Joined  2015-09-05

thought about it more, relative values 0.5 extra for draw with Black, 1 white win, and 1.5 Black win seems reasonable.
(a win with Black is much more difficult).
So yes bonus for wins but then with the above params
playing more or 24/7 often like rayrj will be rewarded excessively;
so i would suggest scaling factors, eg 2/5 times the above params,

This (2/5) would give (as example)
+0.2 for draw with Black
+0.4 for win with White
+0.6 for win with Black.

Similar to last month but more difference (and more fair representation)
for black/white intrinsic differences (at this level of computer chess_.

Profile
Posted: 30 October 2015 10:17 PM   [ # 7 ]  
Member
Rank
Total Posts:  6
Joined  2015-05-29

Much more difficult?? Winning with black is next to impossible at this level of play… even inferior opponents can hold for draws relatively easily. Everybody here is using Komodo 9.2. We might as well call this Komodo fest, with the exception of a few Stockfish’s. This is not a human tournament. Why are we rewarding draws??? Winning is the only thing that should matter. I have yet to lose a single game versus any inferior opponent… yet we reward draws day in and day out. People can say what they want about not liking wins only… I don’t get why there has been such a dispute against it. Are people here scared that they won’t win enough???

Draws are a natural part of computer chess when the only two engines facing each other are SF and Komodo. We can have the best books, and the best hardware…. it doesn’t matter. Winning is what champions do… winning is what should be rewarded. Not pathetic draws that mean nothing.

Want to win money??? Then win more matches than most… either increase the point spread to at least 1 to 1.5 per win… give points for drawing as black… I would say a full point actually… or call this pool pretty much the pool of luck…. because there is luck involved in this pool. Sometimes you get the right openings, the right opponents… win streaks occur… or you don’t.

Just fix the pool so that wins mean absolutely the most… then things will look as they should. As I say in chess… draws are for cowards… winning is for champions who actually are the best. Inferior hardware can draw but they almost never win versus those of us who are the best.

Profile
Posted: 30 October 2015 10:27 PM   [ # 8 ]  
Member
Rank
Total Posts:  6
Joined  2015-05-29

Wins only is not simple… it isn’t about it being simple. This pool should be about who the top three contestants are… we are talking about money… ratings in a pool like this are not completely accurate… users such as ipfreel or jpsingh are not bad opponents… and they are not as bad as their rating suggests… if you are a top contender then winning will be natural for you… and it will show in the win coloums… wins only dismisses draws completely. It is absolutely the right approach. This pool should not be based on complicated jargon… it is about effectively dealing with boring dull draws that can ruin the pool.

Take today for example.. I haven’t lost a single game… not one… I I’ve actually won a lot of matches though… about 7…. guess what??? My rating is now lower than what it was before.. why? Because of draws… basically every game at my rating level is a fight for win… if I lose I lose big… if I draw I still lose points… this is ridiculous.

Nothing here is rocket science, and I am certainly not stupid. The draw syndrome is a problem here… put the rewards where they belong. Draws should not be rewardable.

Heck… why don’t we discount draws completely?? But keep the ratings so that it is based solely on wins and losses. Enough with these stupid draws….

Profile
Posted: 05 November 2015 03:30 AM   [ # 9 ]  
Member
Rank
Total Posts:  3
Joined  2015-09-05

ok well, so apparently you are in favor of upscaling the parameters i mentioned,
then we would get:

0.5 for draw with Black
1.0 for win with White
1.5 for win with Black

as extra points, besides the rating system.

NB other point:
something which i find highly annoying, especially with higher time controls,
are such drawn endgames as RK vs RK which occurs rather frequently
(or sometimes QK vs QK); when both sides would have Nalimov this
would quickly result in a draw by the 50 move rule, but because of the
syzygy bases nowadays many engines dont use Nalimov anymore.
What i would propose is either obligatory use of Nalimov (instead of sygyzy)
or (better idea, if possible) let the serve adjudicate such positions as a draw
(in human chess also not a bad idea maybe, at least for RK vs RK).

Profile
Posted: 08 December 2015 09:16 AM   [ # 10 ]  
Member
Rank
Total Posts:  13
Joined  2015-10-26
bookbuilder - 29 October 2015 10:43 AM

well only wins is a bit simple, against a stronger player a draw with black in fact is good,
(and a won game very good) and thus the ‘Elo’ rankings should be calculated; in fact the system
with 0.5 for every won game still is a bit simple, i would propose something as the following:
(in addition to the normal rating calculations, like done this month i.e. October):
After some discussion maybe possible for December, or maybe already November:

A) 0.5 (extra) points for a draw with Black (0 for White, yes, this can cost rating points for White..)
B) 0.5 up to 1 (extra) points for a win with White (similar as now, i.e. done in October)
C) 1 up to 1.5 (extra) points for a win with Black (instead of 0.5 as it’s now)

Not sure yet about the value of parameters B and C
and ofcourse the whole thing can also be scaled (eg 3/5 x A,B,C)

For discussion, on the basis of the October results sofar we could get
an estimate how such a system would look like (the idea is that
it’s better reflecting real strength (based on the inherent advantage
with White),  while also giving bonuses for playing more often,
but maybe not so much as in October (here it was beneficial
to run 24/7, in my system there’s maybe a better balance.

Giving more point for drawing with black is not good.  Why?

1) It is too easy to set the engine to draw (and they already draw most of the time because they are all of even strength).

2) It favors engines that are online 24/7 and that are playing a huge number of games, so they will have many more draws that will then help have a better score, even with a weak engines.

Imo, you don’t decide who is the best by how many draws he get, but how many wins he gets, or how many more games he did play.

The rating is still the best way to determine the best engine among its peers for a month, although I questioned sometime why the ratings we have are not higher then what is listed.

Once establish, the rating self-correct itself because draws among engines of the same strength have no effect on the rating, and only the wins helps get a higher rating.

And there is no clear proof (yet) that black, by playing second, is at a disadvantage.

The only real disadvantage is for engines that have no proper opening book.  This is mainly make some engine seems like they perform better then others, because they have better opening line then others.

 

Profile