I have surveyed and tested many open source go AI over past couple of months. Tried face book dark forest go and was not impressed.
I would say the Rayon (rn-3.9) is the strongest open source go AI as of today.
Here is Rayon Bot that I set up at DGS:
http://www.dragongoserver.net/userinfo.php?uid=97868

Shoreline

Hi,

I've tried looking for any information on the Rayon bot. Do you have a web link to the source code or it's homepage.

Thanks

I placed the link of the source code in the above post, the link to DGS user info, which has the link to source

By my testing, Rayon Bot at DGS can get 5d rank comfortably. It could be higher rank than that, we will see.

Leela just hit 9d on KGS. Is that stronger or weaker?

_________________
Occupy Babel!

Short answer: it's stronger
But why compare between non-open source and open one?

Not stronger for sure. KGS 9 dan is only equivalent of DGS 6 dan. There is leela old version at DGS only ranked at 3 dan.
There is claim that new leela is 2 stone stronger than old one, which means that only at DGS 5dan level.

My Rayon Bot accept challenge by bot. The leela bot at DGS did not accept games with bot. If leela operator at DGS
is willing to do a even game against my Rayon Bot at DGS, we will know for sure.

Shoreline

I'm not Gian-Carlo, but I'm willing to hook up Leela to DGSlib and run it against Rayon. What kind of hardware and time settings are you thinking of using?

_________________
The road to wisdom? Well, it's plain, and simple to express: Err, and err, and err again; but less, and less, and less!

Doesn't it because DGS maximum is 6d? Are all DGS 6 dan users get 9 dan in KGS (if they play on both server)? If not, what's the lowest KGS rank of 6 dan DGF user?

DGS uses Japanese rules, and Leela uses Chinese ones exclusively, so Leela won't play correctly on that server.

In any case playing random machine A against random machine B for a single game is a pretty silly way to compare two programs which you can just download. There's several tools that can play matches between GTP engines.

My setup for RN3.9 at DGS use 1 Intel PC with dual core CPU and 1 GPU (geForce GTX 1050). I would say if you set up a new leela with single computer with 1 GPU at DGS, the competition is still fair game.

I think at 5d to 6d level at DGS, the sampling size is too small to make correct statement. I used the sanseis web page of go rank comparison to say KGS to DGS conversion

I agree, but it takes resource and time to do the testing. DGS does allow chinese rule as well.

My point is, Rayon RN-3.9 is well known on strength because it participated in CGOS with ranking second only to Zen. If new leela shows up in CGOS, then we have some basis to compare. Otherwise, it is difficult to say. Internal testing also has its limitation because of hardware, configuration etc set up differences.

Because I misread the reference to it being free as it being open.

_________________
Occupy Babel!

I think comparing strength between leela and RN3.9 is fine. But the result does not change my title here because leela is not open source go program.

Home seems to be here:

https://github.com/koban6/Ray

Edit: Ahh it's the zakki branch https://github.com/zakki/Ray, not koban6 branch, as pointed out later in this thread.

I just saw DGS 7d today. The maximum dan at DGS is not 6 dan.

Yes, there are nine 7d's (currently one of them is active) on DGS.

DGS has a nice way to display all top players and their ranks in real life/on other servers:
http://www.dragongoserver.net/users.php ... 1507100060

It seems that 6d is the maximum rank one can claim without playing a game on DGS.

How to run this bot? I want to test it out.