Kung Fu Chess - kfchess.com • page 1/2 • General Chess Discussion • lichess.org

I built a chess variant that brings the real-time strategy aspect of games like StarCraft to chess. It's a game originally from the early 2000s that I'm excited to bring back: https://www.kfchess.com/. Let me know what you think!

TheBigDecline

#2

Just watching the game on the front page makes me nauseous, this variant is definitely not my cup of tea, and I only ever saw people talking about it here on Lichess ... nevertheless, I wish you all the luck in the world with your project!

PrinsHenrik

#3

I just played a couple of games and it is incredibly fun! Very well done.

SakisRouklas

#4

that was really cool, I liked it

NeverBeenTimid

#5

Yeah, it's a fun game.

MrCharles

#6

Interesting.

Not knowing how to play, I started a game. The "Novice" setting (AI) is weak. Too easy. I lost one game against the Intermediate AI because I wasn't paying attention. (So, the AI isn't totally dumb.) Haven't played Advanced yet, but, given I'm not that good a player and can beat both Novice and Intermediate easily, I'll guess the "Advanced" isn't too difficult, either. Both the Novice and Intermediate don't execute moving pieces it could move faster (not fast enough, easy to anticipate). I assume a threatened king moves faster in Advanced. To be fair, I'm moving something like 5 pieces at once. (Fists of Fury.) And, I'm a regular ZH player trying to anticipate drop-attacks based on exchanges, so ... the timing of being able to move a piece again is a small but easy adaptation.

Kept me entertained for something like 15 or 30 minutes, so ... not totally dumb, kinda cool.

Interesting. Not knowing how to play, I started a game. The "Novice" setting (AI) is weak. Too easy. I lost one game against the Intermediate AI because I wasn't paying attention. (So, the AI isn't totally dumb.) Haven't played Advanced yet, but, given I'm not that good a player and can beat both Novice and Intermediate easily, I'll guess the "Advanced" isn't too difficult, either. Both the Novice and Intermediate don't execute moving pieces it could move faster (not fast enough, easy to anticipate). I assume a threatened king moves faster in Advanced. To be fair, I'm moving something like 5 pieces at once. (Fists of Fury.) And, I'm a regular ZH player trying to anticipate drop-attacks based on exchanges, so ... the timing of being able to move a piece again is a small but easy adaptation. Kept me entertained for something like 15 or 30 minutes, so ... not totally dumb, kinda cool.

paladin8 edited

#7

Thanks for your feedback! The novice AI is supposed to very easy, so that someone who understands the basics of chess can win. I'd expect anyone who visits lichess to be able to beat the advanced AI with some practice, as it uses only very simple heuristics to decide what to do. I do want to flesh out the AI further, but making it smart without having it be computationally expensive is not that trivial!

AI is too easy.

The advanced AI is way too easy. I suck at this game and even I beat it.

MrCharles

#10

@paladin8

"I do want to flesh out the AI further, but making it smart without having it be computationally expensive is not that trivial!"

Exactly. Great observations. But, to promote strong play, you have to do this (or at least get to it eventually).

Aesop's Fables: Slow & Steady wins the race (little by little approach).

Consider a neural network approach to fine-tuning a 1-to-10 "AI" -- whatever you mean by "AI" -- could be anything, which I assume, at this point, is likely static programming with some timed thresholds or the likes. By "1-to-10," I mean ...

Novice Levels: 1, 2, 3
Intermediate Levels: 4, 5, 6
Advanced Levels: 7, 8, 9
Impossible: 10

You'll need a small stats package, and some dedicated time to review games. And that'll take a bit of time. While working on implementing that, think about how to improve existing ... back and forth between the tasks, and in a year you could probably have 10-Impossible implemented and have the AI system fine-tuning itself based on some sort of curve (maybe bell curve, or maybe base it on some external measure such as gathering info about strength of new players from other standard chess websites, etc).

Food for thought. It's your thing. Just willing to try and help you improve.

@paladin8 >> "I do want to flesh out the AI further, but making it smart without having it be computationally expensive is not that trivial!" Exactly. Great observations. But, to promote strong play, you have to do this (or at least get to it eventually). Aesop's Fables: Slow & Steady wins the race (little by little approach). Consider a neural network approach to fine-tuning a 1-to-10 "AI" -- whatever you mean by "AI" -- could be anything, which I assume, at this point, is likely static programming with some timed thresholds or the likes. By "1-to-10," I mean ... Novice Levels: 1, 2, 3 Intermediate Levels: 4, 5, 6 Advanced Levels: 7, 8, 9 Impossible: 10 You'll need a small stats package, and some dedicated time to review games. And that'll take a bit of time. While working on implementing that, think about how to improve existing ... back and forth between the tasks, and in a year you could probably have 10-Impossible implemented and have the AI system fine-tuning itself based on some sort of curve (maybe bell curve, or maybe base it on some external measure such as gathering info about strength of new players from other standard chess websites, etc). Food for thought. It's your thing. Just willing to try and help you improve.