Stefan Buecker wrote on 01/26/12 at 09:25:08:
Many thanks, Vass, Hacker and Vladimir. Now I've read about it, Houdini's Aquarium seems indeed promising, and as someone who spends lots of time with analyzing I should certainly try it. (I didn't try the first version of Aquarium, because in some blogs a bit too many flaws were discussed.)
In some cases (endings, KI positions and so on) Rybka had problems, e.g. in endings Deep Shredder was better. Other programs were useful, too, to get new ideas, so I'll gladly check your interesting suggestions, including Critter 1.4.
Hi Stefan,
How are you?. Replying you after a long time. Read all the comments here.
Let me tell you few things. The topic is "Analysis programs" and not Analysis tools, let me first elaborate on the engines first.
The top 5 engines are good in analysing positions you want. All have their strength and weaknesses. Namely - Houdini 2.0c , Critter 1.4 , Stockfish 2.2.2 , Depp Rybka 4.1 and Komodo 4 MP(Yet to be released, now SP only)
Regarding feature rich it is Rybka 4.1. It supports Large pages, Preserve Analysis , Randomizer, Monte Carlo , Multipv_cp and 2048 cores in cpu.
Houdini too supports Large pages, Save Hash , Multipv_cp and 32 cores support with learning feature and Nalimov + Gaviota TBs support. But i find the learning feature a bit faulty in it.
The best learning feature is in Critter 1.4a which implements a session file. Very good for Deep Position Analysis and Backward Analysis in Fritz GUI if you know the right configuration settings.
Here is the link to Unofficial one -
http://www.mediafire.com/download.php?w4x1kctds4oa9ogUnofficial, unsupported, and only 64bit Sse4 build:
- tweaked SMP implementation
- sligthly decreased Queen value
- time management: try avoiding blunders in sudden death games when in time trouble
- time management: fixed losing on time whith "ponder on" games on opponent's instant reply
- improved pawn blockage detection (3B4/1r2p3/r2p1p2/bkp1P1p1/1p1P1PPp/p1P4P/PP1K4/3B4 w - - 0 1)
The Stockfish 2.2.2 is good for analysis in IA but not feature rich. So you need a feature rich stockfish which is good for analysis. Here is an Unofficial one.
Stockfish 2.1.1 PA GTB Gran2c - download
http://www.open-chess.org/download/file.php?id=356]The changes from Stockfish 2.1.1 include:
Preserve Analysis: Stockfish keeps in mind what you have analyzed and doesn't research it until you unload the engine. Perfect for backward propagation of scores and saving time for transpositions.
GTB: Allows the user to use Gaviota Tablebases in Stockfish.
Granularity=2: Allows Stockfish to have more precision on move scores, so that instead of 0.00 -> 0.04 -> 0.08 it goes 0.00 -> 0.01 -> 0.02.
Mac, Win32, Win64 & source included. Source maintained at:
https://github.com/jeremybernstein/Stockfish_PA_GTBThe best GUI for Rybka is Aquarium because it uses its feature to the fullest.
Randomizer is good feature which I always use to get a General Idea of a position. Shared Analysis is another feature which is great with Deep Rybka 4.1 . If you haven't read then read here -
http://www.chessbase.com/newsdetail.asp?newsid=5074]You can do it manualy
Created a txt file Deep Rybka 4 SSE42 x64.uci_b or Deep Rybka 4 x64.uci_b and type the following lines.
[ENGINE]
Parameter=%PID %CPUS
[OPTIONSA]
Preserve Analysis=true
And save the file with extension .uci_b in the same folder where the chessabase saves the uci file in appadata/roaming/chessbase/Engines UCI.
While analysing Just right click in engine window and Shared Analysis.
Another feature you can use in Aquarium is the Delta which uses MultiPV_cp for this (UCI3 support).
MultiPV_cp option is to limit multi-PV analysis to moves within a range of the best move.
I use 'delta' in infinite analysis. Like set delta to 25 and variations to 200. Then when you run the engine it shows the top move and all other moves that are up to 0.25 pawn worse. In some positions it will show only one move but in other positions it may show many. I spend a lot of time running analysis and setting variations to 5 (without delta) wastes a lot of time if there is only one good move. On the other hand in some positions there may be 10 moves that are all good and I would only see 5 of them. That's why you will find delta a great feature.
Note on using Critter in Backward Analysis and in DPA
In IA use Always Resolve Scores and Depth Strategy (say =10) and this seems to be the preferred settings.
With a fast machine and Critter's new session file, you can run with depth 25, with +1 at the root, with a branching factor of 4, 4, 3, 2, an analysis length of 7, and a window of 30, cutting bad moves.
Hope this helps you and all. After trying please answer me if its helpful to you.
Regards
Om