Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Aquarium / New Video: Best software for analysis
- - By Carl Bicknell (*****) Date 2021-01-30 09:11 Upvotes 1
Sorry to double-post but didn't want Aquarium users to miss out, given that your favourite software hit the number one spot!!

https://youtu.be/1Yg9DOlJKFI
Parent - By CumnorChessClub (***) Date 2021-01-30 10:30
Thanks for the post otherwise I would have missed it.

Great to see the person behind the video's on Aquarium. :cool:
Parent - By Dadi Jonsson (Silver) Date 2021-01-30 12:15
Another great video!

I don't think there can be any doubt that Aquarium is the most powerful chess analysis software available.
Parent - By dickie (**) Date 2021-02-02 12:01
Thanks Carl, I enjoyed watching your video from start to finish. Nibbler has a bit more analysis functionality than you displayed, but nevertheless I wholly agree with your conclusions. I have been using Aquarium since its inception and the few problems I have encountered have been mostly of my own making! For me, Aquarium remains the outstanding program for serious chess analysis.
Parent - By mattchess (**) Date 2021-02-03 02:25 Edited 2021-02-03 05:46
Thanks!  I was actually subbed to your channel already but for some reason this had not popped up in my queue - will enjoy this evening.

EDIT:  Excellent review - agree with your comments!
Parent - - By cma6 (****) Date 2021-02-11 03:54 Edited 2021-02-11 05:17
Excellent and witty video as always from Carl. However, with regard to Aqr's supposedly great back solving (assuming I understand the term), I have lost an occasional ICCF game due to what seems to me the failure of back solving. These were games where SF (on a LAN with 56 cores), was looking very deeply:  400 seconds/ply-maximum 600 seconds; wait for next move; AND 54 ply. Both of these failures were heavy piece endgames where at least one side had two heavy pieces and plenty of open files. So SF should do well because this is pure calculation.

In the first game, deep into the endgame, IDeA showed me at move 52 as White at 0.00 or perhaps -.20 but no serious problem. Only after getting to (or more likely a couple of ply past) the critical move, say move 56, does IDeA begin to show White falling to -1.90, although in that game, White miraculously held the draw.

By "only after getting to" I mean after session 4 on move 56 (30 days after session 1 at move 52, etc.) e.g., only by session 4 (30 days later) does IDeA show that my 52nd move (originally evaluated as 0.00) is now evaluated at -1.90. At first I thought this was simply horizon effect, e.g., with heavy piece endings one may have to look ahead 68 ply, but now I'm not so sure. Perhaps Carl will answer: 'What you experienced is the definition of back solving. Unfortunately for you, 'in light of future analysis' turned out to be 30 days later."

  The second heavy piece ending will be a loss for me. Again, IDeA was fine with my Q vs. 2 Rs, giving 0.00 till, say move 46. Then everything fell apart on move 48. Again very deep analysis. Doing i.a. to about 66 ply probably would have caught the problem, but it is impractical to do such deep i.a. unless one suspects there may be an issue with IDeA evals. It could be that in certain types of endings, one's radar should be on high alert.

  It is likely true that the best solution was already given by Carl in his video: IDeA with SF with most cores, while feeding tasks from Sandbox/lco into the project; but where can one buy Nvidia’s RTX 3080? Does anyone have an impression how lc0 performs on AMD RX 6800 XT vs. on  Nvidia's RTX 3080?

Also, in IDeA project properties/IDeA dialog, you mentioned in the video the default False Alarm check. Should that option be checked? And what about some of the other more obscure options in that dialog? What say you, Dadi?
Parent - - By Hamster (**) Date 2021-02-12 13:55
Would you mind sharing the game? Those concrete examples are great to test engines or settings.
Parent - - By cma6 (****) Date 2021-02-12 14:28 Edited 2021-02-13 01:02 Upvotes 1
Hi Hamster:
It is a game in process so cannot be shared at present. I will illustrate the issue with changed moved numbers and made up moves.
After very deep analysis, IDeA shows that Black has two good moves. 52...Qd8  IDeA = 0.00  N= 382
                                                                                                      52...Qa8  IDeA = 0.00  N= 287

I choose 52...Qa8 after deep i.a. Eight days later, I get my opponent's move 53 Ra2 and again do very deep IDeA analysis resulting in two top choices with IDeA = 0.00.  Ten days later I get my opponent's move 54 R1a1. Very deep IDeA analysis continues to show 0.00 but now only for one move.

However, i.a. going to 62 ply now shows White's score climbing to .88; the deeper i.a. goes, the higher White's score climbs. IDeA begins to show a problem for Black only after analysis on move 57. This kind of failure by IDeA happens rarely but regularly if one plays a large number of games. So IDeA seems very useful but not so reliable as very, very deep i.a.
Parent - By Hamster (**) Date 2021-02-23 06:47
Thanks for the explanation. Hope you can win the game!
Up Topic Rybka Support & Discussion / Aquarium / New Video: Best software for analysis

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill