Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / My eternal current Top 5
1 2 3 4 5 6 Previous Next  
Poll Uploading games (Closed)
after every gauntlet 7 28%
after a new book enters 0 0%
Weekly - Monday 2 8%
Weekly - Sunday 10 40%
Every 2 weeks 2 8%
Every Month - starting 1st July 1 4%
Every Month - starting 25th July (=1 month after started) 0 0%
Do not upload at all 3 12%
Parent - - By Jonas (****) Date 2009-07-03 23:00 Edited 2009-07-03 23:03
yes, but that only makes sense if you test 2 books. as soon as you run a gauntlet against more than 1 opponent learning is bad again because it affects the result. so id have to run book a vs book b and then reset both then run book a vs book c and reset and so on. that is simply impossible as sad as this is.
so learning off is the best solution in my opinion.

i think deleted is right with the post he made directly after nelsons 2nd post there

Deleted: Turning on book learning adds variability with a positive bias for books that have played more games. Even worse, it makes the order that the games are played important (its much better to play against the most recent books first). Meanwhile, it adds nothing to our understanding of the books strength.

that is also my opinion
Parent - - By Bobby C (****) Date 2009-07-04 01:15 Edited 2009-07-04 01:27
I do not think there is a problem running a test this way with book learning. You get 150 games and the same result as running a Round Robin tournament but the books do not get learning from previous matches and they only get to learn during their current match. There might be another way to configure this better but this get's might point across the best I can explain. I know that for instance in the first gauntlet that MyBook 15.1 get's more learning because it is playing more books but the other books have the advantage of learning only against MyBook which helps because it is better to learn only against one then learning at the same time against many books.

Gauntlet #1 - 50 games

MyBook 15.1
Titielmago 6.1
Hurricane 3
Poker 1.3
Venezuela 1.3
Bertran Blitz 1.0

Then reset books back to default and test
Gauntlet #2- 40 games

Titielmago 6.1
Hurricane 3
Poker 1.3
Venezuela 1.3
Bertran Blitz 1.0

Then reset books again and test
Gauntlet #3- 30 games

Hurricane 3
Poker 1.3
Venezuela 1.3
Bertran Blitz 1.0

Then reset books again and test
Gauntlet #4- 20 games

Poker 1.3
Venezuela 1.3
Bertran Blitz 1.0

Then reset books again and test
Gauntlet #5 - 10 games

Venezuela 1.3
Bertran Blitz 1.0
Parent - By Jonas (****) Date 2009-07-04 02:01

>I know that for instance in the first gauntlet that MyBook 15.1 get's more learning because it is playing more books but the other books have the advantage of learning only against MyBook which helps because it is better to learn only against one then learning at the same time against many books.


yes, thats exactly the problem! when wanting to test serious, you cant give advantages and therefor different disadvantages to a book and then estimate its fair. and thats why learning is off. also running round tournaments whenever a new book gets tested doesnt work because the order is random. but in my opinion it would be more fair than the method you give. your example is bad because running a round robin you could also reset every 2 rounds and you have to reset exactly the same time because you only have to reset after a book played white and black once against each book. still would be unfair because it played a random order of books inside the 2 rounds.
conditions have to be fair from start.. so id always have to test 1 on 1 and then remove all the learning. its the only fair method if i allow book learning but im not able to do that. there should be a program that does that for you. until that, learning is off.
the good thing is a book cant stay in my top 5 if not most of the lines are stable. imo titielmago 6.1 is the best public book out there atm. no book ever scored much higher 50%  against it (only 5,5-4,5 was possible until now) but it was able to score higher against weak books and it showed to stay in my top 5 for a long time and it had a positive score against 15.1.. to hold that record and stay in my top 5 it has to be good in many lines. not just one main line.
Parent - - By Bobby C (****) Date 2009-07-04 01:31 Edited 2009-07-04 01:53
BTW - Attention everyone, I just checked Jonas's first 610 games for doubles and found 21 doubles making the percentage for his first section about 3.5%. This was using Nelsons method of finding doubles. I checked his first tests of over 6700 games and found 211 which makes the percentage about 3.1%, those tests were ran with book learning and resulted in a lower percentage of doubles. I think that when he releases more games without book learning the percentage will raise to more then 3.5% making the difference much more noticeable.

I have ran a double check on my 3000+ engine test games and found 6 doubles using the same method. All my tests have been with book learning in the GUI on, if book authors specify no learning then that is turned off in the book settings or in windows.
Parent - - By Jonas (****) Date 2009-07-04 02:41
dont worry, there is another method i created and thought about that might work quite well when learning is on:
run gauntlets and always remove the result of the runner. then reset. and run gauntlet with the next book. The runner always uses learning OFF. always only use the results of the opponents for the opponents
they always learned against one book that wasnt affected due to learning from other books. and for its result is deleted anyway it doesnt care that it got a bad result.
example

Gauntlet #1 - 50 games
A
vs:
B
C
D

ignore result of A...write down results of B C D and then reset books back to default and test
Gauntlet #2- 50 games
B
vs:
A
C
D

write down result of A and add result of C and D to their previous result...
and so on

what sedat does is exactly the other way round: he only uses the total result of the runner which uses learning on which is a completly random result because it depended on the order of the books it played. then it gets worse: he runs the next gauntlet and doesnt reset between that, so its the second order thats counts even more. so the books kinda run crazy. although he ignores as much results as in this method and for that needs the same time as if he would use this method. thats why i critizise his results alltough he is #1 tester. you can call it shit what i say or whatever you like its still the truth. his testing method is inefficient and therefor slow but still unfair. as hard as that sounds.
imo, if learning is on you have 2 possibilities:
round robin: unfair, efficient
or my method: inefficient because of ignoring results and need more games but therefor fair.

why i dont use this method is that i want to test much faster. it doesnt work with my top 5 because the next book would be E and id have to run each book after another against this. i want to run gauntlets only against the best books and add all results i get from that. the method im using atm is exactly 5 times faster because i only have to run 1 gauntlet because im able to use all previous results and another bit faster because i dont have to ignore any results.
Parent - By Bobby C (****) Date 2009-07-04 04:58

> what sedat does is exactly the other way round: he only uses the total result of the runner which uses learning on which is a completly random result because it depended on the order of the books it played. then it gets worse: he runs the next gauntlet and doesnt reset between that, so its the second order thats counts even more. so the books kinda run crazy. although he ignores as much results as in this method and for that needs the same time as if he would use this method. thats why i critizise his results alltough he is #1 tester. you can call it shit what i say or whatever you like its still the truth. his testing method is inefficient and therefor slow but still unfair. as hard as that sounds.


I think I will let him respond to this. When he get's back from vacation I am sure he will see it.
Parent - - By Jonas (****) Date 2009-07-04 02:44

>BTW - Attention everyone, I just checked Jonas's first 610 games for doubles and found 21 doubles making the percentage for his first section about 3.5%. This was using Nelsons method of finding doubles. I checked his first tests of over 6700 games and found 211 which makes the percentage about 3.1%, those tests were ran with book learning and resulted in a lower percentage of doubles. I think that when he releases more games without book learning the percentage will raise to more then 3.5% making the difference much more noticeable.


lol, almost the same % of doubles. i expected having much more with learning off. thank you for that statistic.
Parent - - By Jonas (****) Date 2009-07-04 02:51
i have no problem with +0,4% more doubles because my results are therefor fair.
Parent - By ForrestGump (***) Date 2009-07-04 23:10
I agree with Jonas, in this type of tournament, "learning off" is absolutely logical and fair option...
Parent - By Bobby C (****) Date 2009-07-04 04:55
The thing is that the percentage of doubles in your test games are higher then they should be, learning on or off. I to expected a much higher percentage with learning off but I never expected as high of a percent with learning on. The percentage I am finding with my tests with learning is less then 1% so I expected the same with yours.
Parent - - By Sedat Canbaz (*****) Date 2009-07-18 13:55
Hello Jonas,

and i also know that he is the #1 book tester
Thank you...but i think you are better Tester than me :)

then it gets worse: he runs the next gauntlet and doesnt reset between that,
so its the second order thats counts even more. so the books kinda run crazy.

Hmm...first of all i always test the books with Author's special recommendations.
So,if the Author's recommendation is with Learning ON,then i can not test the book in other way...

For example,there are also such books (Storm,Ultra Blitz ...) which have been tested without learning,the reason the Authors's recommendations are without learning.

From my own experience,i can say that the Learning ON,is the best way to get high performance on playchess or on book testings.

At least when the book learning is on,then the engine:
1.does not repeat the loosing line,its trying to find and play another variation,of cause if it's available another opening variation.
2.is performing much better,the reason its playing only the strongest/winning line
Note:to see the real performance of the strong books its needed minimum 100-200 games per player

Btw,i respect you as a Tester,you put a lot efforts,for this i am really so thankful to you...
but without this i can't...i think you need to change your My eternal current Top 5's rules

I recommend you:
1.Use the original settings per each book which are recommended by the Book Authors
2.To make the tournament harder and more interesting...give a dead line,where many new books will meet and play first time between...
3.You need to run more than 40-50 games per book player,for a better conclusion it's needed at least 300-400 games
4.When you create elo rating list please post more statitsics,for example the number of the games,draws...
5.And last (for quality games) its will be great if your change your hardware,at least Q6600 2.40 GHz will be much better than the current one...

Best,
Sedat
Parent - By Jonas (****) Date 2009-07-19 01:46 Edited 2009-07-19 01:54
its really ok how you test.. will say more soon... its just you test something different
we need to have both tested
Parent - - By Bobby C (****) Date 2009-07-19 03:59

> and i also know that he is the #1 book tester
> Thank you...but i think you are better Tester than me <IMG alt=":-)" src="/mwf/epx.png" class="sic sic_sml_pos"/>


Please Sedat, you do not have to be so modest. You could have mentioned to him that you have been testing books since the days when cavemen were the pawns, tyrannosaurus-rex was the queen and stegosaurus's were the rooks.

> So,if the Author's recommendation is with Learning ON,then i can not test the book in other way...
> At least when the book learning is on,then the engine:
> 1.does not repeat the loosing line,its trying to find and play another variation,of cause if it's available another opening variation.
> 2.is performing much better,the reason its playing only the strongest/winning line.


This is why book authors should not be counting his results as accurate. They can not be considered by me to be serious results that reflect the kind of conditions a book would be operating under if I were using it (i.e. playchess, flyordie, etc...)

> I recommend you:
> 1.Use the original settings per each book which are recommended by the Book Authors
> 5.And last (for quality games) its will be great if your change your hardware,at least Q6600 2.40 GHz will be much better than the current one...


Big time +1, in the world of making books there are many who do not understand the importance of the simple details. What hardware is used, exact book settings and testing environment are very important. How can Jonas be online and posting hundreds of posts the way he does with only one computer running book tests? If I am running a book test on my PC there is nothing else being done on the computer. I have seen many games lost in drawn positions because the engine the book was using had to fight for more CPU time then the other engine.
Parent - - By Jonas (****) Date 2009-07-19 12:58
i dont test while doing anything else on my cpu.

> 5.And last (for quality games) its will be great if your change your hardware,at least Q6600 2.40 GHz will be much better than the current one...


i agree with that, and i plan to buy a Q6700, windows-7 and more ram. until that i use higher time controls to get usefull results. after i got those 3 things ill start my top 10

>> 1.Use the original settings per each book which are recommended by the Book Authors


i do... but i dont test with learning because it destroys results

i love to quote deleted here:
Deleted: Turning on book learning adds variability with a positive bias for books that have played more games. Even worse, it makes the order that the games are played important (its much better to play against the most recent books first). Meanwhile, it adds nothing to our understanding of the books strength.

and i have to add that it can also add a negative bias for books that have previously played less recent opponents.
Parent - - By Bobby C (****) Date 2009-07-19 18:46

> i do... but i dont test with learning because it destroys results


No it does not. Learning off destroys results. Look at your famous 10-0 score for Yograj vs MyBook, in the real worl that would never happen with book learning the way the books are used on playchess. What good is your test?

> i love to quote deleted here:
> Deleted: Turning on book learning adds variability with a positive bias for books that have played more games. Even worse, it makes the order that the games are played important (its much better to play against the most recent books first). Meanwhile, it adds nothing to our understanding of the books strength.
>
> and i have to add that it can also add a negative bias for books that have previously played less recent opponents.


It is simple, book learning off does not show the strength of a book because the book plays the SAME exact losing lines over and over and if a book starts drawing it willnot try something else to win. To say something like "Well I am testing the book the way the author sends it to me" is total bullshit. When you test a book without learning you are only getting an accurate result for the very first game and every other game after that is worthless.
Parent - By Jonas (****) Date 2009-07-20 01:38

> i do... but i dont test with learning because it destroys results


it happened one time only...
Parent - - By Sedat Canbaz (*****) Date 2009-07-19 22:07
but i dont test with learning because it destroys results

The book learning should be OFF:
-only in case of running many different kind of engines using same opening book

There is no doubt that the book learning should be ON:
- for maximum performance on playchess,on book tournaments...

Regards,
Sedat
Parent - - By Banned for Life (Gold) Date 2009-07-19 23:07
Chessbase book learning can change a book very quickly if learning is maximized. By the time your tournament is over, the book with book learning on may have little in common, other than the moves, with the book that was sent in. Someone may download the book based on your tournament result, only to find out that the book really sucks because it isn't the same book that did well in your tournament. Meanwhile the book in your tournament is changing from day to day. With book learning, the book that wins today is not the same as the book that lost yesterday.

Book learning also makes your tournament path dependent, i.e. the tournament results are significantly influenced by the order you play the games in. As a result, the tournament can't be reproduced elsewhere. Most people probably wouldn't consider a non-reproducible test to be particularly useful.

Maximum performance for tournaments, be they engine tournaments, freestyle tournaments, or correspondance chess tournaments will never be achieved by using a book that has been tuned with the Chessbase learning function. The Chessbase learning function is just a toy and is used only in the minor leagues. Using books with Chessbase learning is fine for picking up new move possibilities, but as far as putting a tournament book together is concerned, this has to be done line by line (the method that Eros told you he uses), or with rocket science (as illustrated by our own NH).
Parent - - By Jonas (****) Date 2009-07-20 01:42
as i said learning is a method to improve book strenght but its by far the worst
and learning is bad for testing
f.e. in my book tour 1 and 2 where i used learning on i recognized a significant difference after running gauntlets of the books against confusion 1.0 and moonrider... the previously tested books got much worse. almost every strong book i tested afterwords performed better than the previously tested books. just because they had a bad influence of leraning
Parent - - By Sedat Canbaz (*****) Date 2009-07-20 08:28
almost every strong book i tested afterwords performed better than the previously tested books
Yes...:) but i am quite sure that the book learning off is not the reason.
Its very simple and easy to explain why the new books are performing better:
-The new incomer books are very well optimized against the top 5 books of your test

As i recommended before...make the tournament more difficult,harder,more interesting...
Give a dead line for testing the new books and then you will see who is the real expert of creating strong books.

Best,
Sedat
Parent - By Jonas (****) Date 2009-07-20 12:13
you dont understand
i was talking of my tests that i made before i created this top 5
and at these tournament i used learning ON. and there i had that problem.
now that i use learning off everything is normal
Parent - By Bobby C (****) Date 2009-07-20 06:07

> Someone may download the book based on your tournament result, only to find out that the book really sucks because it isn't the same book that did well in your tournament. Meanwhile the book in your tournament is changing from day to day. With book learning, the book that wins today is not the same as the book that lost yesterday.


What are you smoking? It must have been you who in the beginning convinced poor Jonas to run his tests in a manner that makes them completely useless. To say that a book would perform well in Sedat's test but suck in real life is retarded(Depending on the amount of time that passes). Testing a book that is meant to be used with learning without learning is the most backwards way of thinking I have come across yet. Explain to me why a book losses 10-0 without book learning but with book learning scores 75%, and both of these books in mention are specifically designed for using learning? Are you one of those guys who has an inside team of kibitzer's working their asses off collecting games for you? You must be because it is obvious that you want to poison the public game collection by convincing rookie book testers to test meaningless hundreds of games without book learning, only the first game matters and all others after that are doubles or worthless to a blitz bookmaker.

> Book learning also makes your tournament path dependent, i.e. the tournament results are significantly influenced by the order you play the games in. As a result, the tournament can't be reproduced elsewhere.


This type of thing happens and is the best reason to think that no learning is better, however, games resulting from testing with learning are more useful to me then games without learning because of the wider range of lines. Learning produces a much better quality of information to me about a book then no learning. The results from both ways and all ways of testing are different to each person, no way is perfect and it never will be. To me it is about the games more then the result, if you want to know who's book is better you have to play 1v1 with learning if the books can not be tuned in between games or without learning if you can stop and fix mistakes during the match.

> Maximum performance for tournaments, be they engine tournaments, freestyle tournaments, or correspondance chess tournaments will never be achieved by using a book that has been tuned with the Chessbase learning function.


You can consider me one of those guys that will agree that .ctg sucks and it's learning sucks too, but to me in most blitz game situations where you are not monitoring your book from game to game then learning is better then no learning. If I am not there personally watching my book play then I can not be there to instantly correct mistakes it might have made, this is for blitz books and not correspondence or freestyle books. If you expect a book to have the best result without book learning then there is no reason to tune in alternate moves because they are never played until a book learns to play them.
Parent - - By Moonrider (**) Date 2009-07-03 10:55
Jonas, please do not forget StrongMoves.ctg, I have sent an e-mail yesterday

Regards
Parent - - By Jonas (****) Date 2009-07-03 11:23
ok.. could take a while i have alot of books to be tested atm.
Parent - - By Moonrider (**) Date 2009-07-03 11:47
Thank you, publish the Top5 here is very nice
elo a table of all tested books on your website, I would welcome!
Parent - By Jonas (****) Date 2009-07-03 11:59
will upload an elo table every week before i upload the games
Parent - By Jonas (****) Date 2009-07-03 00:06
poll closed. ill upload the games weekly on sunday. thanks to all who voted for that.. it saved me alot of work.
Parent - By Jonas (****) Date 2009-07-03 10:40
Venezuela 1.3 entered:

ETB-Tour Gauntlet-13  2009

venezuela 1.3 - Poker 1.3           5.0 - 5.0    +1/-1/=8    50.00%
venezuela 1.3 - Bertran Blitz 1.0   5.0 - 5.0    +0/-0/=10    50.00%
venezuela 1.3 - titielmago 6.1      5.5 - 4.5    +1/-0/=9    55.00%
venezuela 1.3 - Hurricane 3         7.0 - 3.0    +4/-0/=6    70.00%
venezuela 1.3 - Mybook 15.1         5.0 - 5.0    +0/-0/=10    50.00%


adding Venezuela results:
1.   Venezuela 1.3          27,5/50
2.   Mybook 15.1            27,0/50
3.   Bertran Blitz 1.0      26,5/50
4.   Titielmago 6.1         25,0/50
5.   Poker 1.3              24,0/50
6.   Hurricane 3            20,0/50


removing Hurricane 3 results:
NEW TOP 5
                            1           2           3           4           5
1.   Mybook 15.1            **********  5.0 - 5.0   4,5 - 5,5   5,5 - 4,5   6,5 - 3,5   21,5/40
2.   Venezuela 1.3          5.0 - 5.0   **********  5.5 - 4.5   5.0 - 5.0   5.0 - 5.0   20,5/40
3.   Titielmago 6.1         5,5 - 4,5   4.5 - 5.5   **********  4.5 - 5.5   5.5 - 4.5   20,0/40
4.   Bertran Blitz 1.0      4,5 - 5,5   5.0 - 5.0   5,5 - 4,5   **********  4,5 - 5,5   19,5/40
5.   Poker 1.3              3.5 - 6,5   5.0 - 5.0   4,5 - 5,5   5,5 - 4,5   **********  18,5/40
Parent - - By Jonas (****) Date 2009-07-03 20:44
tested Om Gambit Book

ETB-Tour Gauntlet-14  2009

Om Gambit - Poker 1.3            3.0 - 7.0    +0/-4/=6    30.00%
Om Gambit - Bertran Blitz 1.0    5.0 - 5.0    +2/-2/=6    50.00%
Om Gambit - Titielmago 6.1       4.5 - 5.5    +2/-3/=5    45.00%
Om Gambit - Mybook 15.1          5.5 - 4.5    +3/-2/=5    55.00%
Om Gambit - Venezuela 1.3        2.5 - 7.5    +0/-5/=5    25.00%
Parent - By Jonas (****) Date 2009-07-04 10:54
Salvo 14 entered:

ETB-Tour Gauntlet-15  2009

Salvo 14 - Poker 1.3               3.5 - 6.5    +0/-3/=7    35.00%
Salvo 14 - Bertran Blitz 1.0       6.5 - 3.5    +4/-1/=5    65.00%
Salvo 14 - titielmago 6.1          5.0 - 5.0    +1/-1/=8    50.00%
Salvo 14 - Mybook 15.1             6.0 - 4.0    +5/-3/=2    60.00%
Salvo 14 - Venezuela 1.3           5.5 - 4.5    +1/-0/=9    55.00%


adding salvo 14 results:
1.   Salvo 14           26,5/50
2.   Mybook 15.1        25,5/50
3.   Venezuela 1.3      25,0/50
3.   Titielmago 6.1     25,0/50
3.   Poker 1.3          25,0/50
6.   Bertran Blitz 1.0  23,0/50


removing Bertran Blitz 1.0 results:

NEW TOP 5
                            1           2           3           4           5
1.   Titielmago 6.1         **********  5.0 - 5.0   4.5 - 5.5   5,5 - 4,5   5.5 - 4.5   20,5/40
2.   Salvo 14               5.0 - 5.0   **********  5.5 - 4.5   6.0 - 4.0   3.5 - 6.5   20,0/40
2.   Venezuela 1.3          5.5 - 4.5   4.5 - 5.5   **********  5.0 - 5.0   5.0 - 5.0   20,0/40
2.   Mybook 15.1            4,5 - 5,5   4,0 - 6,0   5.0 - 5.0   **********  6,5 - 3,5   20,0/40
5.   Poker 1.3              4,5 - 5,5   6,5 - 3,5   5.0 - 5.0   3.5 - 6,5   **********  19,5/40
Parent - - By Jonas (****) Date 2009-07-05 00:23
Bouddha lean entered:

ETB-Tour Gauntlet-16  2009

Bouddha lean - Poker 1.3         5.0 - 5.0    +1/-1/=8    50.00%
Bouddha lean - titielmago40      7.0 - 3.0    +4/-0/=6    70.00%
Bouddha lean - Mybook 15.1       7.0 - 3.0    +4/-0/=6    70.00%
Bouddha lean - venezuela 1.3     5.5 - 4.5    +1/-0/=9    55.00%
Bouddha lean - salvo 14          5.5 - 4.5    +1/-0/=9    55.00%


adding its results:

1.   Bouddha lean           30,0/50
2.   Salvo 14               24,5/50
2.   Venezuela 1.3          24,5/50
2.   Poker 1.3              24,5/50
5.   Titielmago 6.1         23,5/50
6.   Mybook 15.1            23,0/50


removing MyBook 15.1

New Top 5
                            1           2           3           4           5
1.   Bouddha lean           **********  5.0 - 5.0   5.5 - 4.5   5.5 - 4.5   7.0 - 3.0   23,0/40
2.   Poker 1.3              5.0 - 5.0   **********  5.0 - 5.0   6,5 - 3,5   4,5 - 5,5   21,0/40
3.   Venezuela 1.3          4.5 - 5.5   5.0 - 5.0   **********  4.5 - 5.5   5.5 - 4.5   19,5/40
4.   Salvo 14               4.5 - 5.5   3.5 - 6.5   5.5 - 4.5   **********  5.0 - 5.0   18,5/40
5.   Titielmago 6.1         3.0 - 7.0   5.5 - 4.5   4.5 - 5.5   5.0 - 5.0   **********  18,0/40
Parent - - By Bouddha (****) Date 2009-07-05 07:12
Yes, I did it ! :-)

Than you for the testing !

regards
Parent - By Jonas (****) Date 2009-07-05 09:11
Congratulations!
you're welcome
Parent - By Sunny (***) Date 2009-07-05 15:53
My Book 15.1 Removed. I hope Yuri(Kosmodrom) don't mind.
Parent - - By epithet Date 2009-07-05 21:07
out of those is titielmago40 the only public one?
Parent - By Jonas (****) Date 2009-07-06 00:02
sorry this is titielmago 6.1 i forgot to rename the engine file but it used 6.1
yes, its public
Parent - By Jonas (****) Date 2009-07-05 11:54
tested flyordie:

ETB-Tour Gauntlet-17  2009

FlyorDie - Poker 1.3         4.0 - 6.0    +0/-2/=8    40.00%
FlyorDie - titielmago 6.1    5.0 - 5.0    +2/-2/=6    50.00%
FlyorDie - venezuela 1.3     3.0 - 7.0    +1/-5/=4    30.00%
FlyorDie - salvo 14          4.0 - 6.0    +1/-3/=6    40.00%
FlyorDie - Bouddha lean      5.0 - 5.0    +2/-2/=6    50.00%


didnt get enough points to enter top 5.
Parent - - By Jonas (****) Date 2009-07-05 12:14
Download Games Part 2:
http://www.multiupload.com/MEOI49L72T

Elo Rating List:

Yograj 2.4 V1.1        3383
Bouudha lean           3382
Venezuela 1.3          3367
Salvo 14               3351
Bertran Blitz 1.0      3342
Poker 1.3              3330
Titielmago 6.1         3319
Hurricane 3            3313
Fx7                    3310
Bombook 1.1            3308
Troy 2                 3306
MyBook 15.1            3300
FlyOrDie               3291
Om Blitz Book 2.0      3289
Fx3                    3287
Bouddha                3282
MyBook 15              3280
OM Gambit Book         3268
Perfect 2009           3259
Bulldozer 2.7          3257
Moonrider III          3207
Parent - - By Moonrider (**) Date 2009-07-05 13:08
Very nice!
btw: StrongMoves Book we publicly tonight.
Parent - By Moonrider (**) Date 2009-07-05 17:54 Edited 2009-07-05 18:06
Parent - - By Jonas (****) Date 2009-07-05 13:11
note: this elo rating list is unfair like all other book elo rating lists
Parent - - By Moonrider (**) Date 2009-07-05 13:16
maybe better to use a table, win, loss, draw, and poinst! but then all books need the same number of games.
Parent - - By Jonas (****) Date 2009-07-05 13:19
a total table with score % would be even more unfair because they played different opponents..
Parent - By Moonrider (**) Date 2009-07-05 13:24 Edited 2009-07-05 13:42
Jonas, therefore it would be good if each new book plays against all the previously existing (10 game), then we have a fair ranking
top5 So no problem, can continue to exist.
Parent - - By Moonrider (**) Date 2009-07-05 13:46
currently 21 books in the list, always 10 games against the books, then you have 200 games for each book, new books added to the number of games is getting bigger. The more games the better the test!
Parent - - By Jonas (****) Date 2009-07-05 23:17
i dont want results against weak books change the positions. but im thinking about upgrading this to top 10
Parent - - By Moonrider (**) Date 2009-07-06 09:25
Top 10 would be a step forward, simply because more games. Currently there are a lot of material, a Top 20 league would be great! then the weakest books are always out.
Parent - By Uly (Gold) Date 2009-07-06 09:57
Or top 16.

Fan of powers of two ^
Parent - - By Jonas (****) Date 2009-07-07 12:24
20 and 16 is too much imo. i would have to decrease time control too much and games played against each book too to get all books tested.
the problem with top 10 or 8 is that i wouldnt be able to create a full table because it would be too much work since i have to create them manually from the gauntlet results. i could only upload total scores and thatwould be alot less interesting.
Parent - - By Uly (Gold) Date 2009-07-07 12:32

> it would be too much work since i have to create them manually from the gauntlet results.


Can't you export all the games as PGN and use Scid to create a table?
Up Topic The Rybka Lounge / Computer Chess / My eternal current Top 5
1 2 3 4 5 6 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill