Not logged inRybka Chess Community Forum
Up Topic The Rybka Lounge / Computer Chess / FinalGen
1 2 3 4 5 Previous Next  
Parent - By mexicanstandoff (*) Date 2012-04-13 21:34

>no one is gonna pay $120 for your disassembler


I was hoping for $549, minimum! :lol:
Parent - By pPerez (*) Date 2012-05-19 07:31
Hi

The Version 1.1 of Finalgen has been released, with the following enhancements:
1- A new status "Unsolved" has been introduced to differentiate unsolved positions from non-generated ones. Next moves of unsolved positions are available.
2- New Copy/Paste FEN buttons. Although no GUI modifications were initially planned, many people asked for those buttons, and this is the reason why they have been implemented.

The new version is also free of false alarms:
FinalGen 1.1 passed successfully all scans from Virustotal, Virscan and jotti.virscan on 18/05/2012

http://www.mtu-media.com/finalgen

I would like to thank Richard Vida, without him we would be still talking about viruses (in a chess forum).
Parent - By RFK (Gold) Date 2012-04-10 17:00
:cool:

Sanity prevails!
Parent - By Pia (****) Date 2012-04-11 13:07

>Since so few bytes (max. 8) can't contain any malware at all


Might it be a virus signature? Worms leave it inside files which are already infected to avoid doing it doubles.

There's no malware code in the first version of FinalGen; virus just marked files with its signature but unable
to inject malware code into program by whatever reason (incapability or program is missing required functions or access rights).

Or it might be a different signature of the same virus.......
Parent - - By Dhanish (***) Date 2012-03-21 16:31

> who needs IdeA?


The difference between the first and second move scores is very less, compared to IdeA, so it looks IdeA is superior here. On the same PC as the above tests, after about an hour, I got:
Parent - - By Dhanish (***) Date 2012-03-22 01:08

>> who needs IdeA?


And overnight :
Parent - By ernest (****) Date 2012-03-23 11:19

>>> who needs IdeA?
> And overnight :


OK, a bit longer with infinite analysis:

at depth 33 (35 min on my dualcore)
New game - Houdini 2.0c w32 Nlm  
8/4R3/p5kp/2r5/8/3K4/P7/8 b - - 0 1
Analysis by Houdini 2.0c w32 Nlm:
1. -/+  (-0.86): 1...Rd5+ 2.Ke3 Ra5 3.Rc7 Ra3+ 4.Kf4 Rxa2 5.Rc6+ Kh5 6.Kg3 Ra3+ 7.Kf4 a5 8.Rc5+ Kh4 9.Rc1 a4 10.Rc6 Kh5 11.Rc5+ Kg6 12.Rc6+ Kg7 13.Rc7+ Kf8 14.Rh7 Rh3 15.Kg4 Rh1 16.Kf5 Ra1 17.Rh8+ Kf7 18.Ra8 Ke7 19.Ra6 h5 20.Ke5 a3 21.Ra7+ Ke8 22.Ke6 Kd8 23.Kd6 Rd1+ 24.Kc5 Rd3 25.Kc4 Rf3
2. =/+  (-0.59): 1...Kf6 2.Rh7 Rh5 3.Rh8 Rh2 4.Kc3 Kg7 5.Rd8 Rxa2 6.Rd7+ Kg6 7.Rd6+ Kg5 8.Rd5+ Kg4 9.Rd4+ Kg3 10.Rd3+ Kg2 11.Rd5 Rf2 12.Rh5 Rf6 13.Kb4 Rb6+ 14.Ka4 Kg3 15.Rh1 Rf6 16.Kb4 Kg2 17.Rh4 Rc6 18.Ka5 Kf3 19.Rh1 Rg6 20.Kb4 Rc6 21.Ka5 Rg6
3. =/+  (-0.53): 1...Ra5 2.Re2 Kg5 3.Ke4 Kg4 4.Rg2+ Kh3 5.Rg1 h5 6.Kf4 Ra4+ 7.Kf3 Ra3+ 8.Kf4 Kh2 9.Rd1 h4 10.Kg4 h3 11.Rd2+ Kg1 12.Rc2 a5 13.Kh4 a4 14.Kg4 Rd3 15.Rc1+ Kf2 16.Rc2+ Ke1 17.Rh2 Kd1 18.Kf5 Ra3 19.Kf4 Kc1 20.Rh1+ Kc2 21.Rh2+ Kb1 22.Ke5 Rf3 23.Rd2 Kc1


at depth 40 (13 hours on my dualcore)
New game - Houdini 2.0c w32 Nlm  
8/4R3/p5kp/2r5/8/3K4/P7/8 b - - 0 1
Analysis by Houdini 2.0c w32 Nlm:
1. -/+  (-0.87): 1...Rd5+ 2.Ke3 Ra5 3.Rc7 Ra3+ 4.Kf4 Rxa2 5.Rc6+ Kh5 6.Kg3 Ra3+ 7.Kf4 Ra4+ 8.Kf5 Ra1 9.Rc2 Ra5+ 10.Kf6 Kg4 11.Rc4+ Kg3 12.Rc3+ Kf2 13.Rc2+ Ke1 14.Rc1+ Kd2 15.Rh1 Rb5 16.Rh2+ Ke3 17.Rh3+ Kf4 18.Rh4+ Kg3 19.Ra4 a5 20.Ra3+ Kg4 21.Ra4+ Kh3 22.Ra3+ Kh2 23.Ra2+ Kg1 24.Ke7 h5 25.Ra4 Kg2 26.Kf6 Kf2 27.Rd4 Rb4 28.Rd5 h4 29.Rf5+ Kg2 30.Rxa5 h3 31.Ra2+ Kg3 32.Ra3+ Kg4 33.Ra2 Kf4 34.Ke6 Rb5 35.Ra4+ Kg3 36.Ra3+ Kg2 37.Ra2+ Kg3 38.Ra3+ Kg2
2. =/+  (-0.67): 1...Ra5 2.Re2 Kf5 3.Rc2 Ra3+ 4.Ke2 h5 5.Kf2 a5 6.Kg2 a4 7.Rc5+ Kg6 8.Rc6+ Kg7 9.Rc2 Kf6 10.Rf2+ Ke5 11.Re2+ Kf5 12.Rc2 h4 13.Kh2 Ke5 14.Re2+ Kd4 15.Rf2 h3 16.Rb2 Kc5 17.Re2 Rc3 18.Rf2 a3 19.Rf5+ Kb4 20.Rf2 Kc4 21.Rf4+ Kd5 22.Rf5+ Ke4 23.Rf2 Rd3 24.Re2+ Kd4 25.Rf2 Rc3 26.Rf4+ Kd5 27.Rf5+
3. =/+  (-0.65): 1...Kf6 2.Rh7 Rh5 3.Ke4 Rh2 4.Kf3 Kg6 5.Rc7 Rxa2 6.Rc6+ Kg5 7.Rc5+ Kh4 8.Rc4+ Kh3 9.Rc5 Ra3+ 10.Kf2 Kg4 11.Rc4+ Kf5 12.Rc5+ Ke4 13.Ke2 Kd4 14.Rh5 Re3+ 15.Kd2 Re6 16.Ra5 Rd6 17.Ra4+ Kd5 18.Ra5+ Ke4+ 19.Ke2 Kd4 20.Kd2 Rc6 21.Rh5 Ke4 22.Ke2 Rf6 23.Ra5 Kd4 24.Kd2 Rc6 25.Rh5
Parent - By Dadi Jonsson (Silver) Date 2012-03-23 21:42
Here you spend too much time on each position. I used six instances of Rybka 4.1 (single core), depth=20, preferred side=White, and got the following in a couple of hours. I started by seeding the tree with infinite analysis and increased the depth in the final prolongation stage to 21.
Parent - - By Dadi Jonsson (Silver) Date 2012-03-23 21:36
With seven pieces on the board and depth 6 you are effectively ignoring the tablebases. In my opinion this run was a waste of time.

> if this can be improved, making tablebase trees in Aquarium will prove practical.


IDeA isn't designed for that, and effective generation of tablebases requires specialized tools. The Aquarium team had already developed tools for generating tablebases (and posted the results here on the forum). It only took them a few hours to generate a full 7-piece tablebase, so it was quite fast. However, it seems that they only generated a few tablebases and then shelved the project.
Parent - By Mark Eldridge (****) Date 2012-03-24 00:12
Can you use full 6-piece tablebases in  IDeA
Parent - - By Barnard (Bronze) Date 2012-03-24 04:11
Hi Dadi

>The Aquarium team had already developed tools for generating tablebases (and posted the results here on the forum). It only took them a few hours to generate a full 7-piece tablebase, >so it was quite fast


>(and posted the results here on the forum).


please,can you link me to the thread were the Aquarium team posted the results here on the forum?i want read it

thanks in advance
Parent - - By Vegan (****) Date 2012-03-28 01:00
get microsoft security essentials if you need a program that is respectable, free for XP and up
Parent - - By Barnard (Bronze) Date 2012-03-28 01:22
thanks...but why i must need that program?
Parent - - By Vegan (****) Date 2012-03-31 00:58
to protect from hostile web sites, malware and more

the MS one is free, some others like McAfee etc are crap by comparison and they want $$ to use it
Parent - By Barnard (Bronze) Date 2012-03-31 01:04
thanks for the info,but i was asking Dadi the link to the thread were they posted about the tablebases they generated,so i dont know what having to do the link to the thread in that forum to 'microsoft security essentials'
Parent - - By Dadi Jonsson (Silver) Date 2012-04-01 23:41
Parent - - By Barnard (Bronze) Date 2012-04-02 18:11
thanks Dadi,im gonna take a look
Parent - - By Dadi Jonsson (Silver) Date 2012-04-23 09:44
I just heard that the Aquarium team has resumed the 7-piece endgame tablebase project. More later.
Parent - - By Barnard (Bronze) Date 2012-04-23 18:39
Hi Dadi

good news!do you know if when they will finish their work,they will release their work to the public?
Parent - - By Victor Zakharov (*****) Date 2012-04-23 21:39
Unfortunatelly this time the choice of ending was not wise: KRPP-KRP. The longest mate is 269 moves only.
Parent - - By Barnard (Bronze) Date 2012-04-23 21:51
thanks for the explanation,but im not interested if the longest mate is in 529 or 639 or in 25.394 moves,for me,it isnt practical,and hasnt any interest

im just interested if Aquarium team will generate the complet 7 men tablebases,and if they generate it,if they will put the complete tablebases available to the general public
Parent - - By Victor Zakharov (*****) Date 2012-04-23 22:12

> im just interested if Aquarium team will generate the complet 7 men tablebases,


They will try. Maybe in December

> and if they generate it,if they will put the complete tablebases available to the general public


They will try. Maybe in May
Parent - - By Barnard (Bronze) Date 2012-04-23 22:15

>They will try. May be in May


in may,of what year?because to arrive may,only 7 days needed

>They will try. May be in December


just curious,why if they generate the tablebases in 'x' time,wait 9 months to make it available to the general public?

p.s.how can you give stimate dates about that?are you part of the Aquarium team?
Parent - - By Victor Zakharov (*****) Date 2012-04-23 22:22 Edited 2012-04-23 22:24

> in may,of what year?because to arrive may,only 7 days needed


I consider that in May: 31+7 = 38 days needed

In December. The year is not specified. But with the current speed - 2012 is not just a fantasy

> p.s.how can you give stimate dates about that?are you part of the Aquarium team?


I was not (am not) far from the generation process.
Parent - - By Barnard (Bronze) Date 2012-04-23 22:32
well,i was counting the first day of may,but you are also correct,the latest day of may also is may :smile:

so if the year is not specified,the tablebases can be endly generated on 2013,or 2014,hasnt a finish date,right?

i asked if you are in the Aquarium team,that was my specific question,because other people can be generating tablebases
Parent - - By Victor Zakharov (*****) Date 2012-04-23 22:54

> so if the year is not specified,the tablebases can be endly generated on 2013,or 2014,hasnt a finish date,right?


Have you impression about the complexity of this work. One table is 130 GB in average. And there are 1001  7-men endings (525: 4+3; 350: 5+2; 126: 6+1). Yes, yes, 6+1 tables are useless, 4+3 are most interesting, but this is still 70-80 TB. Supporting such disk array raises as technical so financial questions.

So nothing can be defined here.

So "maybe" is dominated in the answers.

> i asked if you are in the Aquarium team,that was my specific question,because other people can be generating tablebases


Aquarium team generates tablebases. I am a part of the team. This is exact.
Parent - - By Barnard (Bronze) Date 2012-04-23 23:05
in that case,yes,the 'maybe is a good answer :smile:

yes,the 4+3 are the most interesting,of course,but the 5+2 are also interesting;just imagine the endgame KQ vs KRBPP or KQ vs KRRPP

in total,from your 130 Gb each table,the 4+3 aprox. 68 Tb and the 5+2 aprox. 45 Tb...meaning about 30 hard drives (if one user want to have it at home,with 4 Tb hard drives,not online storage)...

thanks for the 'exact' answer about the Aquarium team...are the team generating on the 'nalimov' format,or in a propietary format?
Parent - - By Victor Zakharov (*****) Date 2012-04-24 18:25
There is no Nalimov format for 7-men.

So any 7-men tables will be in a propietary format.
Parent - - By Barnard (Bronze) Date 2012-04-24 18:32
are there any chances to have the tablebase generator?

and another question,the nalimov tablebases can be converted to the propietary format?for example,to save time converting and not generating the whoie 6 men nalimov tablebases to the propietary format,and with the 6 whoise set,start generating some 7 men tablebases
Parent - - By Victor Zakharov (*****) Date 2012-04-24 21:23
I don't think the generator will be publically available in the nearest time. It is very specific for hardware it uses.

6-men tablebases are generated (it took less than 5 days) and now they are in the process of full comparison with Nalimov data (it is longer process than generation). Converter is not in the plans. Size ratio of the current tables (let's name them Lomonosov tables) is 1.6 times better than Nalimov's. Partially it is due better data organization and partially this is due better compression (2MB solid blocks used instead of 8KB for Nalimov). Current compression is aimed to provide smaller size not the fastest access. Speed issues will be resolved later.
Parent - - By Barnard (Bronze) Date 2012-04-24 21:35
thanks for you explanation Victor

well,one developer (i dont remeber his name now) gave us a tablebase generator for 7 pieces,but to generate 7 pieces tablebases,we must generate before at least the corrsponding 6 minor pieces,and that is too much time consuming in a old laptop like i have (and i think is time consuming in general in all the computers that people will have at home)

but if Aquarium team can generate the full 6 men tablebases in less than 6 days (generated and compressed),i suposse they are using a powerful system to do it,and of course,a very good designed software (very good designed to generate and compress)

edit:what hardware is using the Aquarium team to generate the tablebases?
Parent - - By Victor Zakharov (*****) Date 2012-04-24 22:01
May be I hurried with the details. The methods and hardware will be described later when data will be officially available. Maybe in May.

Anyway thanks for the interest.
Parent - - By Barnard (Bronze) Date 2012-04-24 22:16
thanks to you for the explanations Victor :smile:

i will wait until the full generation of the tablebases to know it
Parent - - By Victor Zakharov (*****) Date 2012-04-26 08:39 Edited 2012-04-26 13:21
Possibly this drawing will help to have a better impression about KRPP-KRP generation. Not all dependencees are shown otherwise it will be a mess.
http://dl.dropbox.com/u/72038782/krpp-krp_final.png
(700 KB)
Parent - - By Barnard (Bronze) Date 2012-04-26 21:52
very complex,from what im looking...

again,thanks from the info Victor

one question,if one person wants generate the endgame Krpp Krp,and thinking in a 'normal computer' (for example,a quad),how much time can take it to generate?aprox.

and the previous tablebases needed (i mean the minor tablebases like Krp Krp,Kbp Kbp,etc...because to generate the 7 tablebase,the person previously must generate some of 6 and maybe 5 tablebases),how much time is required in a quad computer?aprox.
Parent - - By Victor Zakharov (*****) Date 2012-04-27 10:22 Edited 2012-04-27 11:53
It is not a task for a 'normal computer'.

1) You need more than one
2) You need plenty RAM
3) You need fast network
4) Linux system is preferable
5) You need huge and fast disk array

Even in these conditions generation (and performing dependent tasks) is not an easy walk. Only today we looked at real data and found that the number for the longest mate in 269 moves reported earlier is wrong. It was reported from minor ending after non-essential promotion. The real number is 434 halfmoves. In the following position Black can't avoid mate in 217 moves
8/k1P5/8/8/1R5P/7p/r7/1K6 b - -

The moves are evident and can be checked with 6-peice tables 1...h2 2.Kxa2 h1=Q 3.c8=R! ...  218. Rc1#
Parent - - By Barnard (Bronze) Date 2012-04-27 20:40
i cant understan something:

with the tablebase generator that Aquarium team developed,one person need that lot of things that you told me...but with the tablebase generator that other guy developed and gave us (for the public) a few months ago,just with one computer and with a maximmum of about 6Gb of RAM,you can generate it...and that is why i cant understand why the tblebase generator developed by Aquarium team needs that lot more requeriments than the other tablebase generator
Parent - - By Volodya (***) Date 2012-04-28 09:41 Edited 2012-04-28 09:43
I guess it's because of different priorities. You cannot have a tablebase generator that is both fast and lightweight (well, maybe you can, but I've never heard of such a thing and I would be very interested in looking at it). 'That other guy' has made his generator fit to work at a single PC, which is great (many people can use it), and I suppose it's a brilliant work, but the generation speed is relatively low then. Aquarium team has chosen another way: maybe the generator can be used only by the team itself, but it gives results quickly, and I suppose the results will be accessible to others in some way.
Parent - - By Barnard (Bronze) Date 2012-04-28 13:01
Hi

thanks for the reply

i suppose you are right,but i cant asure it since i cant generate the tablebase (it requires some minor tablebase,and i havent hard drive space and pc requeriments to do it)

about the guy,i named it 'other guy' because i dont remember his name,im really sorry

well,i suposse that to acces we ''in some way'' the tablebases will be online,the only way,but here comes the question of the million,will be free of charge? :wink:

edit:i mean acces 'online' positions like shredder page has to consult,not buying,for example,hard drive (or Dvd's) with some preinstalled tablebases
Parent - - By keoki010 (Silver) Date 2012-04-28 14:50
Jerry, ChessOK gives you access to 6-man Tablebases now. You can access them here in the forum, online at the site, or from Aquarium.
Parent - By Barnard (Bronze) Date 2012-04-28 15:04
thanks keoki,i know i can

i mean that if when 7 men tablebases will be availabe (all generated from the Aquarium team),will be available for as to acces for free like now we can access for free the 6 man tablebabes
Parent - - By Victor Zakharov (*****) Date 2012-04-28 15:07
I beleive the 'other guy' is Sergey Yankovich. His work looks interesting. But several years more are essential to make his method practical and usable for ordinary users.

I can say nothing about kind of access to the tables. No any decisions yet. I can only say that software and hardware tools for RPP-RP hosting will be ready in May, 2012.

To be honest although this work was not done for making money - some income can really help for its progress.
Parent - By Barnard (Bronze) Date 2012-04-28 15:28
yes,he is,now i remeber his name,too much thanks

why do you think several years are needed to make his method usable for ordinary users?

that is pretty soon! :smile:

well,i think Aquarium team has the right to make some bucks,why dont make a web page with a 'donate button',giving free acces to the people who donate when the full tables are created?
Parent - - By Highendman (****) Date 2012-04-29 13:16
I wish you success in finding a business model around this! You are one of the most innovative (if not THE most) in the computer chess scene. You IDEA tool changed the game - some people don't know it yet, but it has. But I'm sure that with Chessbase being the guerrilla in the market it's not easy to conquer your place. I do hope you will find a way to generate / publish / provide access to 7-men tablebase with a decent market share growing for you- personally I find this super exciting. Kudos on another innovative move on your side!

Cheers,
HEM
Parent - By Nick (*****) Date 2012-04-29 16:51

>  IDEA tool changed the game - some people don't know it yet, but it has.


+1 and this 7-man EGTB generation effort is one of the finest innovations I have seen in the genre for some time, I commend Victor and his team on their project and we're looking forward to their results.

> Chessbase being the guerrilla in the market


I wonder if Frederic would prefer to be seen as a Guerilla or a Gorilla? In any case, I trust you are well, Shahar :grin:

Best regards,
Nick
Parent - By Highendman (****) Date 2012-04-29 12:56
Hi,
This is super cool! I can't wait for the official news, when you have them. May I make a wild guess? Are you using GPU code and not just CPU code? Because a week to generate the 6 men tablebase is very impressive.

I'm sure whenever you'll be ready to announce what you're doing with the 7 men tablebase - it will make a lot of news!
Good luck with your efforts - I'm sure it's an enormous task to coordinate it / store it all!

Best,
HEM
- - By Pia (****) Date 2012-04-06 14:31
Where FinalGen can go wrong, a little.
Rf1 is better according to Nalimov and black delays mate by another 4 moves.

Parent - By Fulcrum2000 (****) Date 2012-04-06 16:58
Yep, finalgen solves to quickest way to promote a pawn, not quickest way too mate. But that's just a technicality ;-)
Parent - By Barnard (Bronze) Date 2012-04-06 20:54
finalgen downt work wrong,it works perfectly;finalgen tells the distance to promotion,and Nalimov tablebases tells distance to mate

so in general terms,is the same if you compare apples and oranges;both are fruits,but both are differents...you cant say finalgen works 'a little worng because it doesnt tell the shortest way to mate',because what finalgen tells is the shortest way to promotion
Parent - - By Dhanish (***) Date 2012-04-07 02:39

> Where FinalGen can go wrong, a little.
> Rf1 is better according to Nalimov and black delays mate by another 4 moves.


Here FinalGen at least proves the win. But sometimes it simply says White does not lose, which is very disappointing. If it were able to access six men bases, such results may not occur. Perhaps it can make use of some free engine to tell the result when converting to six men positions!
Up Topic The Rybka Lounge / Computer Chess / FinalGen
1 2 3 4 5 Previous Next  

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill