Random "Pippen Fuseki" Generator

[moyoaji]

So in the opening move preferences thread Pippen posted this idea.

Here's another one, I call it "Pippen-Fuseki":

1. Mark every intersection on the 3rd and 4th line with a number (=104 numbers).
2. Get a random generator.
3. Play the first move in the corner on x-x (my choice: 4-4).
4. Then let the random generator decide the next moves under the following conditions:
a) If a numver comes up that is already occupied by a stone then let a new number be created.
b) If a number comes up that is (vertically or horizonally) as near as y (my choice: 1) intersections on one of your own stones then let a new number be created.
c) Stop this algorithm after your move y (my choice: 7) or if you have to create a 5th number on a move.
5. Continue the game according to what was dealt to you.

I would tend to believe that the success-rate of a Fuseki like that is as good as any Fuseki as long as we do talk about players weaker than 1p. Obviously one can go into detail of the algorithm even more to make the fuski more fitting. The benefit of a fuseki like that could be that you are unpredictable and to shock your opponent while you forcing yourself to let your creative juices flow instead of staying in some patterns/affections.

That sounded like fun. So I took the time to run with it and slapped together a Java program that actually does this!

It's buggy and easy to break, but hey, it pretty much works! Just don't try to break it and you can enjoy playing games the random way! As an executable JAR file you should be able to run it on any computer or device that has Java (if you can run the KGS client then you can probably run this).

I put it up on SourceForge if you want to give it a whirl: https://sourceforge.net/projects/randomfuseki/

If people like this and want me to try to work out the bugs/polish this up, I will. Otherwise you guys can just enjoy messing around with my crappy program. Or you can modify it yourself if you want. The source is in a zip in the src folder because I decided to make it nice and FLOSSy!

Here are some screenshots:

The Pippen Fuseki Generator!

FirstWindow.png (28.08 KiB) Viewed 9693 times

The game in progress!

GameWindow.png (29.32 KiB) Viewed 9693 times

And Pippen, if you want I can remove your username from the program. That was the working title, but I sort of liked it. Plus it was your idea.

[RBerenguel]

I wrote a small javascript snippet to do something similar: I just restrict the first line and obviously playing on a previous move. If there's an enemy stone in place skip the move. I have not tried it yet with a human (I beat CrazyStone for iPad1, don't remember the level.) I want to try it, though, but I feel it will be insulting to my opponent, specially since it is unexpected

[Dusk Eagle]

I think this fuseki has a rather large weakness if your opponent knows you're playing it, as they can just play contact moves against your stones which you'll be unable to respond to. It'd also be bad if the RNG gives you two stones close together and your opponent plays a move that threatens to cut them, but you can't connect because you have to play elsewhere.

I don't agree that this fuseki will have a good winrate all the way up to 1p, at least in non-blitz games, but I do think it's an interesting idea. If you allow yourself to respond to contact moves and peeps it loses the main weakness I wrote about. Though you may find it more fun to let the randomness screw you and then try to fight back afterward.

[Pippen]

I think this fuseki has a rather large weakness if your opponent knows you're playing it, as they can just play contact moves against your stones which you'll be unable to respond to.

You gotta realize that we talk just about the moves 2-7. There is no way someone can actually make your fuseki real bad, because you could always treat stones as light and aji-makers.

The point of the fuseki is that it is more or less plausible that almost any moves not too tight together and on the 3rd and/or 4th line are playable in fuseki. The further point is to break personal patterns. E.g. I am a guy that usually falls in love with certain patterns and tends to overestimate them. With the Pippen-Fuseki you are forced to play with patterns you'd never have played voluntarily and you might benefit from that - be it that you realize what else is possible to play, be it to be forced to play creatively instead of repeating your database-fuseki/joseki-knowledge. Also I would not underestimate the shock-potential. Imagine you play an important tournament game and you know you are the better "fighter", then be playing randomly you'll be a headscratcher for your opponent and you might bring him to think longer or to lose his style. I know some guys that like to play japanese-like - clear, simple and aestethic. I'd drive them crazy to have to play against a Fuseki like that^^.

One thing I learned through my 10 year Go-journey is that basically there is anything playable in early fuseki...look at Robert Jasiek and his playing style with black. What really differs the weak from the strong is how they do in middle fuseki and middlegame...in the first 10 moves you are hardly worse than Lee Sedol, it's the next 10-20 moves that begins to seperate you from him^^.

Basically my "Pippen-Fuseki" is just a prototype idea: What if you play your first moves semi-randomly, i.e. randomly with some conditions? Is it really worse longterm for your Go or would you benefit? I think programmers could sophisticate it quite more. How to imply not only moves on 3rd and 4th line, but also 5th line? How about to play the first 20 moves randomly for a computer program that will not need and have any strategical, but only tactical knowledge.

Maybe one day I'll play like 500 games with this style and see where it brings me...maybe someone else will...maybe someone will implement this idea into a bot (if it's not already)...maybe some new josekis will be found...maybe it's just another dead-end-idea^^.

[RBerenguel]

When I got the idea (it was 4 months ago or something like this... I'm not exactly sure but the source is in my old computer, and the new one is 4 months old, so..) my idea was to play at random ~25 moves to improve my middle game fighting. The plan was to just throw stones and then fight for the useful, use the rest as aji stones and somewhat find what my "middle game only" level is like (is it far from normal fuseki game? How many stones? And what happens is I do it only for 10 moves? etc)

[Pippen]

I also wanted to thank moyoaji for his effort. I'd be really curious how a program like crazystone would do with my setting against its original and if there would be significant difference. I think I will also check the sourcecode to learn more about java...I always wanted to be able to program...is there any catchy course at youtube that one can recommend? It's not so much the language but the tools I'd love to have.

[Boidhre]

I also wanted to thank moyoaji for his effort. I'd be really curious how a program like crazystone would do with my setting against its original and if there would be significant difference. I think I will also check the sourcecode to learn more about java...I always wanted to be able to program...is there any catchy course at youtube that one can recommend? It's not so much the language but the tools I'd love to have.

If you like, poke me on KGS some evening. I can run CrazyStone 2012 on my machine and relay the moves on the board. (I don't know a way to make the program autoplay itself, though is probably possible with Fuego?)

Actually, my idea doesn't work at all unless you've the same processor, memory etc.

[moyoaji]

I know it's been a while, but I wanted to let people know that, after over a year, I finally took the time to update this program. It was in pretty bad shape performance-wise (it would update the entire graphics of the board every time the mouse moved around) and had some glaring bugs (like playing illegal moves). It still is far from perfect, but I felt like it deserves at least a bit of an upgrade. Or a huge upgrade in this case. I would estimate its about 90% more resource efficient. That's more to do with how badly I wrote it before, but, to be fair, I wrote this generator in a day. And again, just don't try to break it and it should do fine. (It's still at the same place on https://sourceforge.net/projects/randomfuseki/).

Why fix it now after all this time? Because I updated the graphics engine for a different program that I'm writing and decided to integrate those improvements into this. I'm pretty happy about this new program and will be posting about it as soon as I've finished preliminary testing. As with this fuseki generator, it will be open source, involve some randomness, and, of course, made for people who love go.

[Pippen]

Cool stuff. I still think such an random opening-style will not harm you. Would be cool to implement it on a strong engine for the first 5-10 moves to see if such an opening can hold with its standard colleagues.

p.s. I don't know if its an error, but your program plays random first two moves even if I set it 4-4 for both colours to begin with.

[moyoaji]

Cool stuff. I still think such an random opening-style will not harm you. Would be cool to implement it on a strong engine for the first 5-10 moves to see if such an opening can hold with its standard colleagues.

p.s. I don't know if its an error, but your program plays random first two moves even if I set it 4-4 for both colours to begin with.

I'm trying to reproduce the bug, but the setting of the first moves is working for me. Are you putting in "4-4" for both players? The way the program parses that would put both moves on the upper left star point. I suppose I should fix that so it doesn't allow white to play its move on top of black's stone, but it does play 4-4 for white as expected.

[Pippen]

Here are two actual games, I played black in both cases and had good games going in my opinion. It looks promising, even if I messed up in the first one.





@moyoaji: How about being able to actually put in the coordinates, so that one can fix the first two moves? How about having the option to add a 5th line? (This is basically for me as a reminder) I might really begin to torture KGS with that fuseki^^. It might be more effective than my split fuseki and might change go like V-Style changed ski-jumping .

[moyoaji]

Wow. I can't imagine what your opponents were thinking with those opening moves. I wonder if either of them thought: "Am I playing a random number generator?" More likely, though, they probably thought you were drunk.

@moyoaji: How about being able to actually put in the coordinates, so that one can fix the first two moves? I might really begin to torture KGS with that fuseki^^.

You mean being able to declare multiple initial moves for each player? Like black played 16-4, 16-16 (nirensei) and white played 4-4, 4-17 (hoshi and komoku) now begin to make random moves for black?

[Pippen]

@moyoaji: Let's take the interface of your actual program to explain my ideas:

1. First line you could put in black's move, e.g. d4, or leave it blank to let the program choose a random move for black.
2. Second line you could put in white's move, e.g. q4, or leave it blank to let the programm choose a random move for white.
2a) In an extra line you can choose which color you want to go random from there. You can choose Black and White and e.g. after 1. & 2. both Black and White get random moves or just Black so that just Black gets random moves while one inputs the actual White moves to which Black reacts randomly.
3. Third and Fourth line are the same as now.
4. In a fifth line you could choose the line parameter from 3 to 5 (maybe even 6 or 7), meaning: if you put in e.g. 4 then the program will just consider 3 and 4th lines moves, if 5 then it consider 3,4 and 5th line.

With such a program I could play on KGS, because the first move(s) I could put in precisely and then choose the randomness for one color.

@all: What do you think of the games, I think my random-fuseki worked OK, I didn't blunder the fuseki, did I?

[moyoaji]

@moyoaji: Let's take the interface of your actual program to explain my ideas:

1. First line you could put in black's move, e.g. d4, or leave it blank to let the program choose a random move for black.
2. Second line you could put in white's move, e.g. q4, or leave it blank to let the programm choose a random move for white.
2a) In an extra line you can choose which color you want to go random from there. You can choose Black and White and e.g. after 1. & 2. both Black and White get random moves or just Black so that just Black gets random moves while one inputs the actual White moves to which Black reacts randomly.
3. Third and Fourth line are the same as now.
4. In a fifth line you could choose the line parameter from 3 to 5 (maybe even 6 or 7), meaning: if you put in e.g. 4 then the program will just consider 3 and 4th lines moves, if 5 then it consider 3,4 and 5th line.

With such a program I could play on KGS, because the first move(s) I could put in precisely and then choose the randomness for one color.

So you wish it, so shall it be...

https://sourceforge.net/projects/randomfuseki/

[Pippen]

Hopefully your little tool makes my breakthrough to a new level (maybe I'll get some easy resigns because people feel bothered^^). Also I'd be interested in the code since I think it's not too difficult so I could learn a little bit (but only if its convenient with you). Did you also have pseudo-code before programming or did you write the whole thing out of the cold hand? In any way: Big thank you. You invested time to write a code, just to help someone and make him happier. That's special, thx bro.