Fritz 18 still without NNUE?

[Uri Blass]

The blurb: "The new Fritz 18 engine has become about 30-40 Elo points stronger, but that is of no practical relevance. The current engines are only tested against other engines. For normal analysis, all engines find the best move within a few milliseconds in 99.99% of positions. That is why we have based our analysis on practical utility: Fritz simulates a strong club player and LEADS you with subtle hints to a strong attack you can be proud of. You have never trained with so much fun."

Sounds like a promising thing for the majority of human players.

That is simoly not true, Engines that use handcrafted evaluation have a hard time finding the optimal move in the opening/first few moves, and that is no different of Fritz. For the longedt time, Stockfish and Komodo would choose to play the French Defense against 1.e4, and the Exchange against the French, despite the Najdorf and Petroff being far superior for black. and the Steinitz/Winewar being better for white. I don't see it being any different with Fritz 18/Gingko. It is why those players who want to use engines for opening prep and normal analysis have long since moved to engines with neural networks in them.

I do not think that we know that the french defense against 1.e4 is an inferior opening.
Strong GM's with rating above 2700 still played this opening in 2021.

Normal analysis by humans is also not for the opening position and there are chess database for the opening position to see what the top players play.
If you need analysis then it is after the opening position.

[mvanthoor]

Because, Alex, Fritz is a GUI first and foremost. FAR more simpler than Banksia...

That wasn't always the case. Fritz was an engine, running in the Chessbase GUI. Shredder ran in the same GUI, and the program was sold as Shredder. Only with Fritz 14, the GUI became "Fritz", which could then run any engine. The program sold as "Fritz" since then ran engines such as Pandix, a remake of Rybka, and now Gingko.

They did it like this so they could keep the Fritz name, and combine the GUI with any engine they want. IMHO, they should just sell the GUI and combine it with a few open source engines.

With regard to ease of use: that is in the eye of the beholder. Also, Chessbase products have traditionally been buggy. I'm using the Fritz 11 GUI, with the next to last update. The reason is that the last update tries to backport some features from Fritz 12, and in doing so, they broke half the Fritz 11 GUI which they never fixed.

[matejst]

I believe you probably meant it, but just to clarify I would add that since Fritz 7 at least the GUI could use UCI engines. Fritz was a chess playing program only until version 3. I know for sure that Fritz 4 could run CB engines, and with an adapter, I could until recently run winboard and even UCI engines too in Fritz 5.

In general, the Fritz GUI was a great GUI. I did not try the new ones -- new does not always mean better -- but 15 years ago it had good database capabilities, a lot of options for playing and analyzing, and was not too bloated. Compared to the Banksia GUI, it was cleaner, more ergonomic, and the focus was on chess, not on engines' tournaments. (Personally, I prefer the old Shredder GUI but it cannot read most annotated pgn files correctly.) I said it several times already: it all depends what you expect from a GUI. If it is to play and learn to play chess, Fritz is excellent. If it is to run endless engine tournaments, I guess there are better options.

Then, for most chess players (and today probably not even for GMs) SF is not a must. What is: from my experience it is an engine with Elo/strength/skill levels, multiPV, personalities, TBs support (preferable to Syzygy), and engines that prune less are preferable to engines that search faster and deeper. Eventually, CB's marketing is on point.

[jdart]

Fritz is aimed at a wide range of consumers, not just very strong players and engine enthusiasts.

For a long time Chessmaster was probably the top-selling PC programs, despite not being among the very top programs. Fritz is similar. It is well marketed, it is good enough for most consumers.

Personally I don't have any problems with the F17 interface. It has many features besides being a playing program. Among other things I use it for matches that involve Chessbase format books, which it supports. (Chessbase is another thing: I like it a lot, but every version I have used has been buggy).

[matejst]

Fritz is aimed at a wide range of consumers, not just very strong players and engine enthusiasts.

For a long time Chessmaster was probably the top-selling PC programs, despite not being among the very top programs. Fritz is similar. It is well marketed, it is good enough for most consumers.

Right now, there are no free GUIs aimed at the general population, with perhaps the exception of Lucas Chess. PyChess feels unfinished, Scid, ScidvsPC, ChessX are mainly for databases, Tarrash CG lacks advanced options. Projects like Arena and Banksia are highly specialized for engine testing, but then, BSG is still in beta, so there is still hope it will able to perform basic database operations -- I saw some positive steps toward ordinary users in its development lately.

[Madeleine Birchfield]

Then, for most chess players (and today probably not even for GMs) SF is not a must. What is: from my experience it is an engine with Elo/strength/skill levels, multiPV, personalities, TBs support (preferable to Syzygy), and engines that prune less are preferable to engines that search faster and deeper. Eventually, CB's marketing is on point.

For that, one could simply use Leela on the CPU with various different nets which correspond to different Elo/strength/skill levels/personalities.

[Madeleine Birchfield]

BSG is still in beta, so there is still hope it will able to perform basic database operations -- I saw some positive steps toward ordinary users in its development lately.

Such as?

[matejst]

Such as?

Analyzing on the fly has become simpler and ergonomic and a function for deep analysis is developed. For the rest, we still have to wait: setting engines and opening books is probably too complicated for most users, database functions are still rudimentary, Playing/analysing -- the first two things most users do are difficult.

[mvanthoor]

Right now, there are no free GUIs aimed at the general population, with perhaps the exception of Lucas Chess.

IMHO, LucasChess is also convoluted and goes against every conceivable GUI practice "normal" people are used to under Windows and Mac. (Under Linux... there are more than a few programs that have... eccentric GUI's.)

Banksia GUI is the only thing I've tested that comes somewhat close someday be a replacement for Fritz 11. Fritz 12+ became convoluted in the same way as every other Ribbon-using program did.

PyChess feels unfinished, Scid, ScidvsPC, ChessX are mainly for databases, Tarrash CG lacks advanced options. Projects like Arena and Banksia are highly specialized for engine testing, but then, BSG is still in beta, so there is still hope it will able to perform basic database operations -- I saw some positive steps toward ordinary users in its development lately.

Scid and forks are somewhat OK, but it should have moved to GTK or QT a long time ago. If Scid moved to QT or even GTK3 so it would support standard theming, it would be a great piece of software. Now, the GUI is out of place in every GUI you use it in. Arena seems not to have been developed for over 6 years now, except for native ports to Linux... and there are still plenty of bugs in it. Also, the GUI is also not the cleanest ever.

Assuming I ever get to the point where I write my own GUI, it'll probably be close enough to Fritz 11 that they'd practically be interchangeable. (Except for the database part, probably.)

[matejst]

Fritz 11 with just a few improvements -- would be great! I agree completely with the rest of the post.