Ovyron wrote: ↑Mon May 16, 2022 6:20 am So, let me parrot myself. I spent like... a whole decade watching engines play each other and analyzing with the strongest ones that showed some style that I loved. Being king attacks or leaving pieces en prise or sacrificing rooks for pawns like they were worthless balderdash.
Back then we enjoyed seeing the games of those engines and produced the games to watch them, unlike the people of the era that only did it to measure elo difference, obsessed with some elo number and to rank the engines, for some reason.
And unlike people of today that figured a way to "measure style" by playing hundreds of thousands of games and checking the shortest mates only, which is not my cup of tea, but what you gonna do?
But that decade was so long ago my favorite engines are doubtfully relevant. The problem was that I stopped doing such things, right at the time some people started producing engines with great styles without caring about elo, but I never go to try those ones.
So here goes for the nth time:
Thinker 5.3b Inert - Top engine style-wise for the most years. This engine would disregard everything you know about solid play and truly go bonkers on a king attack or material sacrifice for reasons nobody could ever understand (as it didn't show a PV, you never knew what it was thinking.) This version played roughly at the level of Rybka at the time, which was mindblowing, though nobody back could keep pace with Rybka. Later stronger versions lost the touch, I figure seeing that the attacks we'rent successful and avoiding them, so this was the peak.
Zappa Mexico Dissident Aggressor - Heh, this was some setting for the engine, I don't even remember if it was for this one or Zappa Mexico II. The thing with it is that it could produce winning lines that no other engine would see, and that when played others claimed Zappa was losing for 3.00 or such, but Zappa knew it could hold it or was winning. The greatest at squeezing the life out of quiet positions.
Toga Chekov - Another setting for some Toga version, or other. Back then the insanity of the Stockfish versions of today was held by the Fruit versions, there was the official branch and other unofficial subbranches, and then there were the Toga branches, which were a mess so huge that you'd upgrade from a version III of one to a version 2 of another! Well, I got my hands in all of them and Chekov was significantly more active than any other. I mean, I completely forgot how the others played, but after all these years I still remember how Chekov played.
Hiarcs Paderborn 2007 - On those years one of my main analysis engines was Hiarcs, even after years and years passed, there'd be some kind of positions where Hiarcs would have an unique idea that worked. Back then I'd say the idea was "human-like", today I know better but still, the ideas were unlike-engines. This Paderborn version had a style distinct from previous and later versions, perhaps by a fluke, but it was definitively the most courageous of the bunch.
Pro Deo 1.2 Q3 - Rebel Decade 3.0 was my initiation into chess engines, who know what hobby I'd have had if it didn't exist. I never stopped using Rebel versions until Ed started making versions that wouldn't run on my machine, oh well. I'd make hundreds of personalities and watch them play against each other. Q3 is the name of the personality that played the best, later even Ed would make a pseudo Multiprocessor version that would play normally unless Q3's suggested move hit some score. Why 1.2 version? Well, it had the best style against engines of that level of strength, 1.6 would have to face stronger engines and wouldn't get to shine as much.
The King (ChessMaster) Tribute Personality - Another great engine with personalities was Chessmaster. I got them in their own GUI, put the children animations on the pieces, and watched them play. Not only the ones for this, but the ones from CM9000, 10 and 11. Some other communities were obsessed with making personality. Tribute played the game you'd imagine a Chess Master who was a King should play, even if his plans didn't work against stronger chessmaster versions.
Komodo KingHunter - For a very short while they allowed you to mess around with Komodo settings. They didn't get the point that it was done for fun, they wanted us to find better settings than default or something, since we didn't the feature was taken away. but in the meantime we got some entity REALLY obsessed with mating the opposing king, to ridiculous levels. The thing was that while this greatly weakened Komodo, it could do that AND play stronger than all the above members of this list! That was an unprecedented style+style combo, but at this point in time engines were so much stronger than people barely cared.
Houdini 6 Contempt 10 - The first engine to dethrone Thinker 5.3B. Thinker had so much style that stronger engines couldn't break the ratio, but Houdini managed to do it. I never get to know how this Contempt worked (despite its source being exposed as a stockfish clone later), but it's unlike any other Contempt implementation I've ever seen. I guess it's similar to the recently implemented Optimism of Stockfish, except this one focused in complicating the position even to its own detriment. 10 was the maximum, I always wondered how some Contempt 100 would play like, imagine an engine that instead of wanting to mate your king it wanted to complicate the position so nobody had an idea of who was ahead. Many times this was the only defense in losing positions, while other engines would just give up.
Rybka 3 Dynamic - There's many versions that could be on the list, like Winfinder or Mindbreaker settings, but his one where because it's the only one with a style aimed for material imbalances. You know, you can have an engine that likes pawn+rook better than bishop+knight and would aim for it. Or another that likes the opposite and would aim for it. Well, Rybka 3 Dynamic would aim for BOTH, being happy if the imbalance was on the board, no matter what side she was on. Otherwise she would play normally, but Rybka 3 was the peak in style of all the versions, anyway.
Naum 3 - Really? Naum? What is it doing on the list?So Naum was the anti-style engine, like, the opposite of all the above. The most passive, good for nothing style I ever seen. it wasn't even solid like others, just...not there. But that's what made it memorable, at her strength it was incredible to see her draw games against engines of much better styles without breaking a sweat. It becomes worthless when the level gets higher and its passiveness just lead to loses, but still the most memorable engine on the other side.
Fizbo - The most original move picker of all. One day I created a system to pick moves by using several engines as move suggesters and others and judges to create original chess entities. Well, Fizbo played more originally than my system! I don't even know if it was a bug or something, but Fizbo would just focus on the most strange move on the board and play it if it was possible. If there was ever an engine that played like a clown, it was this.
Andscacs - On the time of Andscacs the top engines were very samey. Stockfish was suggesting moves that when played against everything else would convince them that it was best, so there was some kind of consensus of the best moves and things became really boring. I even hit 0.00 as a backsolved score of chess positions. Now, with NNUE, I know that lost of those 0.00 were losing, but back then I thought I had the best lines. Andscacs challenged those assumptions by playing differently, it felt like a breath of fresh air. Most of its ideas didn't work but when they did it led to new lines and exploration of things missed by Stockfish. The most unique engine of its time (I think it was even radically different to itself from version to version!).
Leela - And then Leela came and destroyed all the styles ever played by an engine before. Kingscrusher is a Youtuber that has dedicated a great deal of time examining Leela games, as Leela was becoming stronger and playing chess that was outside this world! Dethroning Houdini Contempt 10 as it wasn't just the best in style, at some point it was also the strongest engine! But note I'm not saying lc0, at some point in the process of improving the style was lost, so who knows when it peaked.
Honorable mentions: Strelka, Shredder 13, Vitruvius - They'd have rocked my world if they were released years earlier than they did, by their time their style was eclipsed by others.
More honorable mentions: Critter 7.0, Fritz 10, Junior 10 - Great styles but "machine-like", on the list I have great styles that feel organic.
And now, to answer the question... what is my favorite engine?
Well...
*Drumroll*
Stockfish 15 with nn-f31e5e1cd71f.nnue net!
Heh, I forgot to mention all the above things were pre-NNUE times. NNUE changed everything and people don't like to admit it but NNUE Stockfish features a great playing style! The way it wins against everything else is formidable, and if you don't like the style, you change the net.
This net I mention was released just yesterday, and I watched in awe how it was disagreeing with previous nets on positions and playing a completely different chess from that witnessed from all previous nets! Too bad it didn't make the cut, but, well, it was losing to the previous net so it makes sense to use another one.
What seems to be happening with the improved nets is that they've been more and more aiming for positions that previous nets can't play well, automatically reaching for positions so complex if you're not a NNUE you're going to be dead. This creates and amazing style which is about tension on the board which is like walking on a tightrope, and it's so much stronger than anything else that it totally breaks the style/strength ratio.
So there you have it, my opinion, I think NNUE eat the world of chess on both strength and style, at this point seeing a NNUE outplay another is the most spectacular thing, and it has nothing to do with fewest moves to mate a king.
The Zappa Dissident Aggressor settings were my creation. Glad you like them!
https://open-chess.org/viewtopic.php?t=227
' they're no better than playing Stockfish 17.1 at low depth (you can easily defeat Depth 1 and Depth 2, so there's a Depth you can find that plays at your level.)