Top GM opening preparation

[GrandKoalas]

Hello everyone,

I am new here, please excuse me if I am posting in a wrong place.

How would you use a 2000 dollars a month budget for a strong player opening preparation?
Is it better to buy a strong machine for 50.000 dollars and use it for 2 years, or rent for 2000 dollars per month?
A team of a few persons, we need to use both Stockfish and Lc0 (other suggestions are also welcome) for 10-15 hours per day.

Thanks!

[jefk]

interesting question, apparently you want to sponsor
a Serbian top GM team (?).

opening preparation imo depends on the style/methods of the specific Gm's themselves (*).
First of all there's Chessbase, and they can prepare against the (repertoire of) the opponents
they are probably are going to face. Also depending on time controls; for blitz/rapid
they can use more tricks (sharper openings) than with slow time controls, where it imo
boils down more to the fundamentals If they then want to use fast engine analysis,
you can use Chessify, and use the expensive option (fastest engines; which probably
will remain within your budget. On Chessbase you can also use with 'ducats' engines
in the cloud, and often there also are fast engines available. Probably (imo most likely)
a better option than buying a fast computer; also it's more mobile, with their
laptop/notebooks the superGM's can access such stuff from whatever location.

As for opening analysis from scratch, there is the Chinese database and
my software (decades old) 'Bookbuilder' with some advantages over the Chessbase
systems (eg backsolving, much more names of variations and last but not least,
analysis of -most- possible variations rather than relying only -or mostly- on GM
or correspondence games played in past). latest book uploaded just now
https://sourceforge.net/projects/bookbu ... s/download
and the whole program here
https://sourceforge.net/projects/bookbuilder/

The Bbuilder software has been recently picked up again by a Latvian amateur
programmer with my guidance, and (also for others here(**) ) we're currently looking for new
a tester/ programmer(s) the latter being interested in Borland C++ Builder version 6
(unfortunately the program is not very compatible anymore with recent Embarcadero
version 10 although a whiz programmer could start almost from scratch with the
older source code(s)). You can sponsor us if you would be interested have a look
eg at my blog (signature below). Good luck anyway with your team

(*) for some super GM's the opening preparation doesn't even seem to be
so important anymore , eg a certain Carlsen (who avoids slow time controls
nowadays) relied -besides on a general solid opening knowledge- on his
excellent middle game analysis intuition and additionally his superb
endgame (sometimes 'grinding') technique(s); Gm's tend to be stubborn,
- mostly justified i guess because every person and their methods are different,
and unless you 're a very strong player yourself with tournament experience, it would
be hard to advice those Gm's. In opening analysis some will be mostly interested in
statistics (which doesn't work for rare openings or gambits), or have no problem in
combining various lines based on the end evaluations they found online; some other
may still be interested in (transpositional) 'back- solving', just to be sure.
But the text(s) above regarding engines in the cloud as imo preferable solution
maybe of use for you; in addition you could buy a reasonably fast desktop comp
eg a threadripper 48 thrds and a fast/big Gpu for Lc0 or so at a decent price (eg 9 k (**))
at a central location for the team; throwing away more bucks for deeper analysis
(than eg. ply 40 with SF) isn't imo worth the money. The differences are then
not more than eg. 0.01 pawn value and thus meaningless; in chess it's about
finding the right (sometimes sharp) plan which the opponent doesn't see
(or sees to late) and thus increasing the likelyhood of one more slight or bigger
(positional or tactical) mistakes for the opponent. my 2 cnts

(**) i was intending to post this in the programming section, maybe i'll do
that later as well

(***) so i just save you 40k (for a comp); the renting of engines in the cloud will
in total also not exceed 10 k i suspect, so then you saved about 30 k

[Peter Berger]

Hello everyone,

I am new here, please excuse me if I am posting in a wrong place.

How would you use a 2000 dollars a month budget for a strong player opening preparation?
Is it better to buy a strong machine for 50.000 dollars and use it for 2 years, or rent for 2000 dollars per month?
A team of a few persons, we need to use both Stockfish and Lc0 (other suggestions are also welcome) for 10-15 hours per day.

Thanks!

Use almost every part of your budget for the help of strong humans who can come up with some general, good approach - and fresh ideas that might come in handy.

Hire one or two computer programming experts, preferably with some project management experience, to get the work organized and use the help of computers in the best way (they should be cheap).

You can probably use a little money on hardware to make your human ressources a bit more productive, but my personal guess is that this is actually not really too relevant.

If you are serious about investing 50.000 dollars, the sky is probably the limit. I personally think Ding spent way less than that on his recent match, and even Gukesh couldn*t really take advantage of this.

[jefk]

Use almost every part of your budget for the help of strong humans who can come up with some general, good approach - and fresh ideas that might come in handy.

yep, good idea.
Note eg. that a lot of innovations in opening theory have come recent decade(s)
from correspondence chess (as was also acknowledged in the last NIC yearbooks,
with a special column about that subject by Dutch GM Erwin 'Ami)
https://en.chessbase.com/post/new-corre ... abase-2020
https://shop.chessbase.com/en/products/corr_2024
https://www.modern-chess.com/the-corres ... -today-299

Regarding 'strong humans', there can be 1) strong human GM 'seconds'(assistants),
and 2) computer chess experts, for example (just one idea), in some lines you could also
use Patricia 5 to explore some attacking opening lines, rather than only SF and Lc0.

As for programming new stuff, well there i may be a bit skeptical, doing is in the right
way takes time, and should imo be well thought out. There's plenty of software
already available, and not only from Chessbase. Some improvements in open
source (eg Bookbuilder) might be useful, though

[Werewolf]

Hello everyone,

I am new here, please excuse me if I am posting in a wrong place.

How would you use a 2000 dollars a month budget for a strong player opening preparation?
Is it better to buy a strong machine for 50.000 dollars and use it for 2 years, or rent for 2000 dollars per month?
A team of a few persons, we need to use both Stockfish and Lc0 (other suggestions are also welcome) for 10-15 hours per day.

Thanks!

Sticking purely to hardware: buying gives far better bang for buck than renting, if you want to use it beyond 2 years because you keep the machine and a dual EPYC will still be fast in 2 years time.

However, if you are treating this as a one-shot two year contract, then renting will give better performance because the Stockfish Cluster is faster than even a dual EPYC, and that translates into finding the best move quicker, or having more lines in the PV while retaining a decent search.

Personally, because I like to prepare by creating a mini maxed tree in Aquarium, I would buy the workstation because Aquarium runs locally.

[GrandKoalas]

Thanks everyone for great replies!

[jp]

...
As for opening analysis from scratch, there is the Chinese database and
my software (decades old) 'Bookbuilder' with some advantages over the Chessbase
systems (eg backsolving, much more names of variations and last but not least,
analysis of -most- possible variations rather than relying only -or mostly- on GM
or correspondence games played in past). latest book uploaded just now
https://sourceforge.net/projects/bookbu ... s/download
and the whole program here
https://sourceforge.net/projects/bookbuilder/
...

But how does one open .pbs files?!

And what exactly do they contain?

[jefk]

But how does one open .pbs files?!

pbs stands for position base and it's opened with the program Bookbuilder
This program was like the program Chess Opening Wizard (formerly Bookup).
But in Bb the backsolving (minimax) is much faster and thus is can also
handle much bigger files.
Some know/knew the program, others not; i made it open source, and while it has been
picked up last yeara (compiling sourcecodes and some tentative improvements) by someone in Latvia,
some functionalities in the latest version still have to be fixed (eg. the backsolving); so for
complete backsolving i still use the older version bb36/37 (and for opening names/repertoire
the version bb44a beta version, it contains much more names than in Chessbase).

For the rest i'm not going to explain all (other) functionalities of the program,
there's enough documentation, the program itself also has a help file; i don't think
it's for GM but more/possibly for 'seconds' (assistants) who are knowledgeable about
computer chess, and thus can use an advanced tool as this to check/improve the opening
repertoire; in the past for example i had worked with IM Erik Kislik to improve/update
the pbs database, and we found many interesting lines (eg in the Najdorf, for Black).
https://www.chess.com/blog/EKislik/what ... r-openings

A lot easier than using the program would be to have a look at the book i wrote
based on the 'best' repertoire (based on nnue engines)
https://www.lulu.com/shop/jef-kaan/bett ... 8r8qz.html
imo suitable up to GM level, although it doesn't contain Gruenfeld for Black
(nor Ragozin(*) against 1.d4), Reti/Zukertort for White (to avoid Gruenfeld);
maybe my second (or third) chess book (depending on which one comes first,
the repertoire 'update' or the 'beauty contest' (most amazing computer games),
if i live that long (i'm now 70).

(*) working on that currently, as last, and final chapter. before, possibly approaching
a publisher ; GM's can simply use SemiSlav instead of QGD (or better, Ragozin),
to equalize, all up to them of course