How to make KGS better?

[gennan]

My wish: Make ratings more volatile.

Discussion of "(optional) automatic scoring": "optional" (both players agree to allow it) is essential because correct scoring (with dead stones still on the board etc.) is in EXPTIME (extremely complex). An impression that it would be done on some other servers is no evidence of being done correctly there.

More rating volatility: seconded

Automatic scoring: I can understand that scoring can be extremely difficult in edge cases. But in practise it doesn't have to be mathematically proven to be correct in all cases.

It's like the travelling salesman problem. It's a known hard problem to find the best solution in general in reasonable time, but navigation software solutions are fast and good enough in most cases, even though it may not give the best solution in all cases.

On FlyOrDie for example, the automatic scoring of final positions is quite good (even when it has to solve some tsumego to determine the status of groups) and that server is not even dedicated to go. It uses territory scoring and IME miscounts happen quite rarely there (perhaps 1 in a thousand?).

[Marathon]

The discussion, so far, is mostly about software changes. What about suggestions for social, cultural improvements, such as admin policy?

[SoDesuNe]

The discussion, so far, is mostly about software changes. What about suggestions for social, cultural improvements, such as admin policy?

Never had to deal with an admin before, so no experience there and for me, social functions are not important. I'd still happily play on KGS without rooms and a general chat function - when the rest of my suggestions get implemented. PMs are a must-have though.

I'd still refer to lichess.org.

[gennan]

I don't know anything about the admin policies of KGS. My interactions with admins have been exceedingly rare.

What I've always found a nice social/cultural feature of KGS is kibitzing on higher profile games. But I don't have much to suggest to improve that without software changes. Perhaps more advertising and announcements on social media (like https://www.facebook.com/groups/go.igo.weiqi.baduk/, https://www.reddit.com/r/baduk/) when some interesting matches are played on KGS?

[shimari65]

Our list of priorities isn't a secret, but at a certain point, more input isn't that useful. We don't have enough resources to solve all of our existing problems as it is. We hope to make gradual improvements over time. We made the list in consultation with a number of users. For now, trying to get registration to work without Java is priority #1. After that is improving ShinKGS to become a viable alternative. You can see our list here:

https://docs.google.com/document/d/17cm ... sp=sharing

[xela]

Never had to deal with an admin before, so no experience there and for me, social functions are not important. I'd still happily play on KGS without rooms and a general chat function - when the rest of my suggestions get implemented. PMs are a must-have though.

Interesting. For me, social functions are the most important thing. They are the reason why I switched from IGS to KGS all those years ago. I don't see KGS competing with the Asian servers for number of players any time soon, so in order to survive KGS needs to hang on to the unique selling point -- and that's the social side plus the excellent game reviewing facilities.

But that's just my opinion. See my post in the other thread on getting some unbiased data :-)

[Marathon]

https://docs.google.com/document/d/17cmLVaao1l1RAPktLyybJK1EAxQRnF4UQBCMchst75I/edit?usp=sharing

Uh, could you use a color other than yellow on white for one of the categories? It's hard to read.

[bernds]

If you could open-source the KGS protocol, then you would no longer need to fix client-side bugs. People will happily write their own clients (they are already trying to!)

I'd certainly give some thought to adding a KGS module to q5go. It would depend a little on whether the protocol requires anything that doesn't fit in the existing structure of the program, but on the whole I think it would probably be doable.
But maybe this isn't such a high priority for KGS - it appears that these days, a lot of people actually want software to run in their browser, rather than as a dedicated application. I find that terrible, but what can you do

[Jujube]

In my opinion, you should open source the project, including client, server, protocol, assets, website, and so on. Stick it all on Github, and choose the person who is (a) available and (b) understands the code to be the Benevolent Dictator. Get rid of your back-of-a-menu issues log and instead have the community contribute to the project. Get a new website, the current one looks like it should run on Netscape.

Look at Lichess for inspiration. They get enough money in donations to run a dozen servers, over 100 cores and pay programmers too. But to get the donations, you need inspirational, motivated programmers and influencers (yes, I said it) to inspire people again. People have already moved to Foxy, they need a reason to come back. Meanwhile, if you're going to focus on KGS Plus content, look at all the Go streamers who have Twitch and YouTube channels. How are you going to offer something more than what they can? Why should I listen to KGS Plus when Pro Yeonwoo is going over 3-3 invasion AI joseki, where I can actually see her face and have a "better than utter crap" aural experience?

Also, I don't think it's out of the question to start again. I mean, programming has moved on further than Java 2 or however old KGS is. One excellent person is all it takes. Lichess was largely written by one person.

[Marathon]

When WMS owned KGS, it was requested several times to make it open source, and he declined. Here is what I remember from conversations in KGS, and web pages, as best as my somewhat faulty memory tells me:

WMS created a client for IGS, and continued to improve it. His was one of many IGS clients. At some point, he realized a consequence of having many independent clients was that IGS could not add new features without breaking clients. That's what motivated him to create igoweb, later renamed as KGS. As he managed KGS, he continued to have the concern that opening the source could lead to the situation in which adding features could break clients.

I'm not taking sides here. I'm just trying to add a historical note.

[shimari65]

I am not a programmer, but my understanding is that the way the code is written is pretty complicated. It isn't something that can be made opensource easily as there is a very steep learning curve to how it all functions, and even minor changes can cause the whole thing to crash. This is what the programmers tell me at least, although I am summarizing it in my words (which may be subject to my lack of comprehension).

Totally agree on the website, it needs a complete overhaul. Of course, no one goes there once they have the client. When we get the web registration module up, we will redo the home page. Everybody talks about how all the users have left, but in reality there are up to 800 on at once over the weekends, and 350-600 on at any given time on other days. So clearly people do still use the server, and value the social aspect and the community building that our structure makes possible. And that is precisely why we are continuing to fund it.

[Jujube]

I work, penance for my sins, as an IT consultant. More often than not, I work with clients that have these set phrases that drive me crazy:- "isn't it as simple as...", "couldn't you just...", "what would be good is if we...". And, I have to turn around and say: no, dammit, it's not just as simple as taking this spreadsheet and "sticking it in the database".

So I truly understand and appreciate that things are never as simple as they appear. We're outside the tent pissing in, not inside the tent pissing out. That being said, the level of technical debt and rigidity in KGS, I guess, is pretty high. It must feel like you're pushing the tyre up the hill when trying to realise your ambitions.

Are you absolutely sure that it's a case of "better the devil you know"? If you find your guy / gal, you can make "KGS 2" in 18 months. Browser-based, sick UI, tight gameplay, livestreams with decent video and audio and interactive boards, tournaments, massive increase in player base?

Chop the head and cut the fat I dare you.

[xela]

It isn't something that can be made opensource easily as there is a very steep learning curve to how it all functions, and even minor changes can cause the whole thing to crash.

There's the server code and then there's the client code. If you make the server code open source in the sense of letting people see but not change it, then it becomes easier for third parties to write their own client. People are already having a go at writing third-party clients. Sharing some information surely makes crashes less likely compared with the current situation of people shooting in the dark?

[Marathon]

In addition to what I wrote above about the reason WMS gave for not having KGS go open source, I'm reminded of this:

WMS, or someone, was concerned that the effect of bugs in poorly written clients could extend outside that user/client environment. They could, conceivably, cause a problem with the server.

IIRC, WMS said he liked open source software. But, he felt it wasn't right for KGS at the time.

[Domie]

Hi everyone,

This is quite hard to know exactly what it is done and what will be done to improve KGS.

I heard about a new client with a web interface. I heard about Go Universe ...

While awaiting developments, would it be possible to work on light changes on cgoban to get them in a very near future ?

Though I am not a programmer at all, I guess that put a nice sound for chat alarm is not one of the twelve labours of Hercules.

Well you see what I mean ?