Vote for Fritz 17 Replacement

[Master Om]

Hoping for Ginkgo, that's one I never got to use.

What I am hearing is Ethereal. So Andrew Grant and Laldol author of Fritz 17.

[Ovyron]

I'll be passing then, I'll not be paying for an open source engine, and the highest thing I can run is CPU Leela because my graphics card doesn't have CUDA and OpenCL doesn't work, so I bet Fat Fritz would be more like Skinny Fritz around here

Looks like this year the only offering will be something I can get for free or something that needs a GPU I don't have, what a shame.

[supersharp77]

NO !

http://www.talkchess.com/forum3/viewtop ... 05#p817405

Free add-on isn't quite accurate, since it really is the showcase piece of Fritz 17, but yes, Fritz 17 will actually come with several engines. There is a new Fritz 17 engine, a conventional AB engine that has no relation to Rybka or Vasik Rajlich, so no idea why his name is dragged into this at all. There is also Fat Fritz, my neural network that runs in a slightly modified lc0. There will also be a smaller Fat Fritz Jr. network that should run well on weak GPUs and even half decently on a CPU at longer time controls. Finally, it also comes with Stockfish (just as in Fritz 16) and Leela itself. Leela will come with the 42850 network, automatically installed and configured. You would not believe how many people have requested this last point.

Really? That Sounds Interesting....Of course people will buy for the "New Fritz 17" and for the 'Fat Fritz' and 'Fat Fritz Lite' Engines....

[hammerklavier]

I miss Frans Morsch... the original Fritz programmer! Come back Frans!!! One of my favorite engines was Fritz 3! good 90's days! ...

[Ovyron]

Eh, I really loved Deep Fritz 10.1, it had a great and unique playing style, I even liked it better than Deep Junior. All that was gone with Deep Fritz 11. It lost the spark and if I showed you a bunch of games I doubt you'd be able to identify the ones by Deep Fritz 11, 12 or 13. It joined the crowd. Like Hiarcs, Shredder, and Junior where you'd be better using an old version.

It would be great if engines like Fritz 3 or Fritz 10.1 had their source code opened, perhaps someone could improve the engine to today's standards without killing the playing style. Robert Houdart showed that was possible, by having the last iteration having the best style, though perhaps it was a fluke (considering Houdini 4 had the worst style of its generation.)

[mclane]

Which Fruit/stockfish clone will it be ?

Let’s hope it plays different. I don’t need 10 engines all playing the same moves in the same search depth window and with the same score or search technic.
I need DIFFERENT engines.

[Ovyron]

I think differently, I hadn't seen so many engines playing so different chess. When it doesn't happen is that there's some best line to play and they find it, so more depth doesn't help because the best line has been found. Something interesting is that most chess positions are like that, the ones that let style show are rare, and that style will just play weaker against best play.

[Albert Silver]

I think differently, I hadn't seen so many engines playing so different chess. When it doesn't happen is that there's some best line to play and they find it, so more depth doesn't help because the best line has been found. Something interesting is that most chess positions are like that, the ones that let style show are rare, and that style will just play weaker against best play.

I have a different opinion on what the best move means. For me, as long as a move will take you to the best result possible, assuming no mistakes, it qualifies as a 'best move'. Suppose that the position is balanced and no mistakes from either side will lead to a draw in the long run. Then any move that doesn't actually lose is a 'best move'. I don't care if some moves lengthen the game by 50 moves, since chess doesn't reward length of game, nor does it matter if every move played leads to situations where all moves but one will lose. Chess does not reward points for less or more risk per se. As such, and this is the key point: it means there is often a lot of room for style overall.

However, chess as a game between imperfect players is not so simple. We make mistakes and risk is rewarded and punished in practice. And in that sense neural networks are a fantastic different outlook on chess analysis by a machine. They may sometimes miss the so-called best play but they always seek out the optimal play which is a very different thing.

[chysiddh14]

provide all free chess engine ethereal ,xiphos n laser
most powerful than rykbba

[Ovyron]

For me, as long as a move will take you to the best result possible, assuming no mistakes, it qualifies as a 'best move'.

I used to think like you Albert, but that just led to drawing most of my correspondence chess opponents, because I played leading the game into positions that were easy to play for both sides, and thus didn't put their skills to the test. When I adopted a "the best move is the one that increases the chances of the opponent to blunder" paradigm, I started winning games against people I wasn't supposed to according to elo, and sometimes so easily that I didn't even know how I won those games, my opponents just blundered them (I guess the sad part of it is that it doesn't feel like I'm winning games, they're losing them.)

Suppose you have access to an oracle that tells you the 'best move' in any position according to your definition. Hurray, you can no longer lose any game! And if the opponent blunders it'll tell you what moves you can make to win! Doesn't that sound great?

However, suppose that you play the oracle's moves against, say, Stockfish at one minute per move, and the game ends in draw. You try another opening, and the game ends in draw. What is going on? If Stockfish at one minute per move is good enough to draw the oracle, what gives?

Now you try Stockfish with 10 minutes against Stockfish with 1 and it wins, it'd show that those moves were better than the oracle's. The only difference is that in 10 minutes Stockfish could play a game losing blunder and the oracle would not, but what's the use of an oracle anyway other than avoid losing?

You can't even use it to analyze your games, it'd just show you how every attempt you can make will fail because there's always some defense. All you can do with an oracle is checking what are the losing moves on a position to avoid them, but to find the real 'best move', the one that would beat your opponent, you need to know that your opponent is fallible, and it means there's some combination of moves you can play against them to make them blunder and win (if it wasn't then they'd be infallible) and try to find it.

Starting from the assumption that the opponent will not blunder will not work for you.