Not logged inRybka Chess Community Forum
Up Topic Rybka Support & Discussion / Aquarium / Aquarium 2020 Hoped for Feature
- - By pawnslinger (****) Date 2019-07-29 22:33
I believe it is a fruitless exercise (to hope for new features in Aquarium).  But, oh well, I am bored this afternoon.

Recently, I have been analyzing up to 4 games simultaneously.  For the most part, I have them in separate Aquarium runs, so they don't actually share resources within Aquarium (usually).  But, sometimes, it is more convenient to have a couple of projects running simultaneously within 1 Aquarium.  Right now, if 2 or more games are being analyzed within 1 Aquarium, I use the "Set CPU Usage" to something like 50% in both projects.  It works okay.  But it could be better.

It would be nice to be able to assign engines on an IDeA project level, thus dedicating certain engines to each project.  The benefits would be worth me upgrading.  If anyone developing Aquarium still cares about selling upgrades.
Parent - - By Ghengis-Kann (***) Date 2019-07-30 17:09
Hi Pawnslinger!

That's a good one.

I currently have 9 different installs of AQ2019 for the purpose of running simultaneous projects with specific engine assignments.
They all have their own dedicated database trees as well to avoid the lock out problem.
It is an absolute necessity to have multiple separate installs when analyzing one project with LC0 and another with Komodo or Stockfish.

My personal plea is and has been to return the record numbers to the database.
There is an empty column where these numbers used to be (in Aquarium 2011).
When you use "Mark Range" it refers to these numbers but you can't know what they are because they have been made invisible.
Parent - - By pawnslinger (****) Date 2019-07-30 18:22
Wow, Ghengis!!  9 installs... I won't catch that anytime soon.

I have 5 installs currently, 4 on my workstation, and 1 on my server.  I do not know why (perhaps my LAN has been hacked), but I had soooo much trouble with RTHomeServer, that I quit using it and just installed Aquarium directly on the server.  Runs okay, but much less flexible.  It is better than turning the server into a flower pot, I guess.

Anyhow, the server has become the bottleneck now.  I have been lazy... I need to install a 2nd or 3rd Aquarium on the server to give it more flexibility.
Parent - - By Ghengis-Kann (***) Date 2019-07-31 18:28
I've had some problems with RTHomeServer406 generating memory access errors that temporarily disable the engines and have tried everything to get rid of them, including disabling windows error reporting and data execution prevention with registry edits. Also assigned engines to specific processors in my dual processor server but nothing works.

I have also been getting problems with the program freezing on exit, or when re-starting Infinite Analysis, but years of Aquarium usage has enabled me to accept these things as inevitable.

Have you had any luck with Linux?
I am going to build a new toy for myself when the 64 core Threadripper chips come out and am torn between Windows 7 X64 Pro or Ubuntu Linux as the primary operating system.
Aquarium usage plays a major part in that decision since I think I can get by with Linux on everything else I do.
Parent - - By bifmeister Date 2019-07-31 23:11
I would really like to see a Linux version or at least something that would run reliably under Linux.

--Bryan
Parent - - By pawnslinger (****) Date 2019-08-01 00:33
I 2nd this idea.  I would love a Linux version.  But this would be a really big request, more than likely, I think.

I could finally leave Windows behind.  Aquarium is the only reason I run Windows... and the Windows update policy is driving me crazy.
Parent - - By Ghengis-Kann (***) Date 2019-08-01 17:33
Forced updates are one of the reasons I have drawn a line in the sand over Windows 10.
I have hoarded a pristine copy of Win 7 X64 Pro that I hope to use on my next build, but it will hinge on whether the motherboard manufacturers release Win7 compatible drivers.
If not then damn the torpedos I'm switching to Linux come Hell or high water.

If necessary I will run Aquarium on one of my existing computers and use the new one to supply remote engines.

Pawnslinger, did you ever succeed in getting remote engines to run on Linux?
Parent - - By pawnslinger (****) Date 2019-08-01 21:30
Hi Ghengis:

No, I gave up trying, because of time pressure (odd to say that about correspondence chess).  Yes, time pressure plays a big role in correspondence chess... just not the same intensity as OTB.  The day I cracked a tooth at the US Open was the day I gave up OTB.

Anyhow, due to time pressure, I also just plain gave up on remote engines.  My LAN seems to be too subject to disruption, the remote engines kept dropping offline.  The day I found my wireless printer using 30 Mbps of my LAN bandwidth (doing nothing I could figure out)... and choking out my remote engines (and my remote desktop), that was the day I stopped using (a) that printer... turned it off (permanently, I hope) and (b) stopped using RTHomeServer (too much time being lost).
Parent - - By Ghengis-Kann (***) Date 2019-08-02 16:48
Hi Pawnslinger.

I have had fewer problems with my network since I pulled all the wireless cards and went back to cables.
I was getting pings as long as several seconds with the wireless setup that were causing engines to drop, and they are down to a few milliseconds with ethernet cables.
Parent - - By pawnslinger (****) Date 2019-08-02 18:26
Hi Ghengis:

I wish I knew why the wifi printer was chewing up so much bandwidth... nothing in the house was queued for printing.  The traffic was all between the server and the printer.  My son said someone was trying to use the printer to hack into our LAN!?  Anyhow, that was just one of our problems.  Every week for about 2 months something different came up, one week it was the router - it dropped everyone off the LAN to do a software update.  So all users, wired or wifi, were affected.  I couldn't believe it was happening, I had just updated the router firmware the week before.  It is a Netgear router and apparently they follow the Windows policy, i.e. update automatically at the most inconvenient time.  At least it had a switch (in the Advanced setup) that I could use to disable automatic updates.  Finally I just had to get some work done, and said enough is enough.
Parent - By ANUAR71 (**) Date 2019-09-01 02:28
as a player in playchess and infinity chess, i hope in the future, aqt format can be convert to hsh format .
In infinity chess there is a ctg option, which can be played directly in hsh format. Now i have aq 2018 but had to use aq 2012 for analysis and to make the book play right there . . .

:cool::cool::cool:
Parent - By mattchess (**) Date 2019-11-23 05:11
Completely unrelated to any practical capability - but I would really love to be able to change it to a dark theme not just for the board but for the whole GUI.  Does anyone know if there is any way to do that?  I've not found it.
Parent - By bformless (**) Date 2019-11-23 22:51
I would like to have a convenient converter from .elm/.hsh-files or .aqt-files to the Chessbase .ctg format, so that i can use my trees in Chessbase too. (It would be ok if the converter is in the Aquarium Tree Utilities.)

Thank you in advance. ;)

cul8'er.
Parent - - By retep1 (***) Date 2019-12-05 16:44
https://shop.chessok.com/index.php?main_page=product_info&cPath=7_43&products_id=859

"What's new in ChessOK Aquarium 2020:

    3-month Chess King LEARN subscription, giving access to 60+ courses available both on the website and in mobile apps! The courses are available in 9 European languages, sorted by category and difficulty level, and include such gems as Chess Tactics for Beginners 2.0, Chess: From Beginner to Club Player, Chess Strategy, Chess Middlegame series, several courses from CT-ART and Chess Tactics in the Opening series and more.

    Updated databases now contain 7 324 000 million games in total!
    1-year access to Lomonosov Tablebases! The subscription will be active for 1 year starting the date of activation at the tablebases service."

No further development? :mad:
Parent - By braindied (**) Date 2019-12-07 03:54
I see Aq2020 is 35.96 euro, and Aq2019 is 35.95 euro. Must be getting something new for the extra 0.01? ;-)

I am very interested to know whether the 2020 version fixes the UCI display issue with engines that output WDL strings, and therefore nothing displays in IA pane.
- - By Goran Trajkoski 02 (*) Date 2019-10-31 15:21
Any news about 4k support for aquarium 2020??!
Parent - By Ghengis-Kann (***) Date 2019-10-31 16:55
Not sure what you are talking about.

My game computer has a 43" 4K monitor and has no problem running Aquarium.
The textures are inherently simple because the visual aspect of chess is inherently simple.

Do you want photorealistic photon mapping with subsurface reflections off of the finger grease somebody left on top of the pawn?
- - By pawnslinger (****) Date 2019-11-03 18:50 Edited 2019-11-03 23:14
I do not know if this feature has already been provided, since I am still using AQ 2017... if it has already been provided, please advise.  Thanks.

One of the biggest flaws in AQ 2017 is the way it begins adding candidate moves to new leaf nodes.  I have pointed this out before and suggested this new feature before... so if I bore you, I apologize.

When AQ 2017 establishes a new root node (or the first root node of a tree), a strategy script is provided that you can modify that will run automatically (or not, if you have disabled it).  This script provides a way to overcome the flaw described above, but ONLY for root nodes!!

I propose that new strategy scripts be provided, for simplicity sake, we can call them "node" and "first node".  "first node" might be executed when the first leaf node is not a root node.  This happens often in new trees that do not begin with move 1.

Such scripts could be easily modified and overcome the flaw mentioned above.  It would seem that flaws should be overcome, right?
Parent - - By pawnslinger (****) Date 2019-11-03 19:24
If implemented, such a "node" or "first node" script should allow the user control over the depth of analysis used during the script execution.... it could be that the user may desire to use a greater or lesser depth to seed new nodes.
Parent - - By Ghengis-Kann (***) Date 2019-11-04 17:10
Would't it be simpler and better if you could define the entire expansion based on a specified tree strategy?

Like "Prolongate up to 3 alternatives if the evaluations are all within 0.5 centipawns of the best evaluation"

This would enable you to replace vague terms like "wider" with exactly what you want it to do.
Parent - By pawnslinger (****) Date 2019-11-04 18:17
I do not know... but I think it would be simple to IMPLEMENT my suggestion, since the program is already setup to use scripts like this.

Personally, I seed new nodes (when I find them manually) with up to 30 candidates at a very low depth.  My theory is to populate the nodes with candidates (no matter what their evals) as quickly as possible, so that IDeA can start proper prolongation, which will throw out that initial eval pretty quickly.

I find it generally useful, but still not what I would call optimal.  Unfortunately, quick and dirty sometimes is the better way, as time pressure in all phases of life and chess come into play.

Oddly enough, after long observation of AQ's prolongation strategy, I find that AQ does what I do, but prior to populating the node.  By refusing to populate the node more comprehensively, AQ is leaving good moves untouched on the table (so to speak).  And normal expansion / prolongation can take a long time to overcome, if it ever does.
Parent - - By pawnslinger (****) Date 2019-11-04 19:59
After thinking about your suggestion more... I think we could just do away with terms like "wider", etc.  I set most of my trees to "wider" and use the tree bound values to control the actual width in centipawns relative to the root node.  Depending on the situation, I have trees as narrow as -+0.17 all the way up to -+6.00.  It just depends on how many alternatives a particular tree can accommodate.  If I set the values too wide, some trees get more than 50k alternatives in a single pass... and that is just too many.  I try to keep it around 1k to 5k in a single pass.
Parent - - By Ghengis-Kann (***) Date 2019-11-05 18:18 Edited 2019-11-05 18:50
It would be really nice if it would automate what I do, which is use an IA engine to seed IDEA.
Currently running Stockfish as the seeding engine and Komodo MCTS to build the tree.
Tried Leela but it is slow even on very expensive hardware, and tends to agree closely with Komodo MCTS.
It also has no problem with taking hundreds of moves to finish a decisive endgame.

You would chose an engine and how long it would spend generating alternatives at each new node.
Branching density  could be controlled by setting the number of simultaneous analysis threads in the same way as usual.

We figured out in a prior thread how to do this manually, by setting multiple IA tasks at different parts in the tree, so the machinery is mostly in place.
IDEA would just have to move those IA tasks forward as time went on.

If you are manually adding 30 alternatives then it would really make sense to create a 30X1 tree strategy.
Parent - By pawnslinger (****) Date 2019-11-05 19:16
I use Custom Task with Alternatives set to 30, auto-plies set to 1, score bounds set to -+ 500.  It is tedious due to the fact that AQ requires me to reset the score bounds each time, unless I get lucky and find an empty node (in which case AQ does not alter the score bounds).

Depending on the situation, and how I feel about a position, I do also a lot of IA seeding... mainly thru "back sliding" thru the main variations.  This can be quite useful, but I find that there are many positions that engines seem not to handle very well.... even using this technique.  Too often top opponents have surprised me with strong moves.  Moves that IA didn't have on its radar.  By strong players I am talking about ICCF IM's and above... usually 2450+ elo (which for a CC player is pretty strong, 2450 elo would put you easily into the top 50 USA players).

Engine wise, I use a mix.  But mainly an old version of Stockfish.  I like consistency, so once I find an engine that works pretty good, I stick with it.  Takes a lot to get me to change.  I subscribe to the old adage, "A good photographer can take great photos with a Brownie camera"... of course, applied to chess engines.
Parent - - By pawnslinger (****) Date 2019-11-05 19:37
I have just now tried my hand at creating a 30X1.xml and it seems to be working well.  But I was wondering, is there any way to control the analysis depth?  Right now it is just using whatever the tree is using, but what if I want to use a different depth?  Say something like this 30X1-25.xml, where you might have different scripts for different depths and a default script for using the tree depth.  Is it possible?  Here is my script right now:

<?xml version="1.0" encoding="UTF-8"?>
<strategy description="ply1: 30 alt">
<white ply1="30=400" maxply="1"/>
<black ply1="30=400" maxply="1"/>
</strategy>

It would great if we could add 'depth="25"'... is something like that possible??
Parent - - By Ghengis-Kann (***) Date 2019-11-05 23:16
I don't know of a way to include the analysis parameters into a tree strategy, but there may be a way to remove some of the tedium from your custom task usage.

You could add ply2 to your current tree strategy with 1 alternative, then ply3=ply2, ply4=ply2...etcetera out to 25 or whatever depth.

This would play out a line for each alternative to that depth with the current analysis settings.

I will call these "Spider Strategies" Where NXM is N alternatives with M depth of linear legs.
Could be useful for a quick survey of where each alternative is taking the position and save a lot of time over my current strategies, which can generate up to N^M tasks.
Parent - By pawnslinger (****) Date 2019-11-06 06:39
I had a little bit different of a plan.  What I would really like is a way to do ply1 for 30 alternatives, but use a fairly shallow depth... then do ply2 for 3-5 alternatives (never just 1) and use the default tree depth for them.  In this manner the 30 alternatives could be quickly generated and regular analysis done on ply2 (where it is really needed).  A reasonable plan, in my view, would be to gen the ply1 at a depth of 24 and then ply2 at a depth of 26-34 (whatever your choice for effective analysis).

This could gen a pretty good feel for almost any position... give a good basis for further analysis.

Too bad we cannot control analysis depth within the script.
Up Topic Rybka Support & Discussion / Aquarium / Aquarium 2020 Hoped for Feature

Powered by mwForum 2.27.4 © 1999-2012 Markus Wichitill