A short travel through OGS history

I’d like to know what prompted you to start OGS. What other Go servers and resources (like sensei’s) existed at the time. What were the difficulties you faced. What was the initial site growth like, etc.

Moreover, some background on you and Arathalion as a Go player, how you got introduced to the game, how you met, what experiences you had, etc. would be nice.

4 Likes

Well, that story is going to sound super unlikely, but sure. So, to begin with it was just me, but Andrew (arathalion) joined me after about the first 2-3 weeks of me working on it. Neither of us had ever played Go before. I was interested in AI, and wanting to develop a chess engine at the time, as I was a reasonable club player. However, thinking that was a bit of a steep initial dive into AI, I looked around for something easier to start with (but harder than something like tic-tac-toe). There was an interesting article on chessbase about one of the Laskers saying that the existence of Go would be the proof to an alien race that humans were sentient life forms, and I decided “right, I’ll do a Go AI then, that sounds like a good starting point” (to think I thought this would be easier than chess, oh the naivety!).

Andrew was my first wife’s younger brother, and I was sort of acting as a big-bro-mentor into the world of programming projects and what programming would be like as a job as a taster, and he enjoyed it enough we worked on it together for some time, with me mostly doing the backend work and more complicated systems for the tournaments, and him doing a lot of the javascript, client side, interface and graphics side of things.

So the first thing I started doing (as someone who had web application development as part of my professional responsibilities at the time), I thought I’d start there to create an easy mock up board with basic game rules. As soon as it was working (sort of … the scoring rules were … ahem, not very well developed at the time), I ended up playing Andrew a fair bit for debugging, and then realised that the game was a lot of fun and I wanted to play a bit more. So I found DGS, played a couple of games, and other than realising how bad I was, was really very disappointed in the lack of organised competition play (when I played correspondence chess, I tended to play tournaments, leagues, ladders, knockouts … all sorts of things, but always some form of competitions)

So, at that point I decided to build OGS around creating a correspondence platform for tournaments, minitournaments, and scoring and seeding systems for them. I designed the group McMahon seeding system that was used - as far as I know it’s the only real effort made to create a hybrid of the normal real life MacMahon system into a format that would work well for correspondence Go (and it mostly transferred on the merge, although the developers never correctly implemented the top group seedings, leading to quite a few “group of death” issues, but pretty minor issues).

Initial site growth was very slow, as you’d expect when there are other places to play (in general there was always a "why would we play somewhere else when DGS works fine?), but of course Andrew and I were still terrible (maybe 15-20 kyu?) so there was some interest in single digit kyu players being able to come in and dominate a new scene. It wasn’t until it hit around 50-100 players that it suddenly really started to self-sufficiently grow, and having titles that people could hold and defend seemed to be pretty popular (as well as the ladder format, which I didn’t come up with at all, but works very well for correspondence games as it “naturally” encourages a reasonably fast play speed).

Other correspondence servers didn’t really exist other than dragon. KGS and IGS were the primary “western” servers - or at least the ones I knew about and heard about most often from people in real life - and CyberOro and Tygem the better known predominantly Eastern servers.

I got to 10-12k after about 6 months, 6 kyu on the year mark, 2 kyu the following year and ended up around KGS 1-2 dan after about 5 years of playing and studying fairly actively. Andrew stopped at around 8k after a year or two and doesn’t play any more. I still play on and off, but much less regularly, much less seriously, and have for obvious reasons stopped improving really. I still enjoy playing from time to time though :slight_smile:

21 Likes

wow wow! This is a lot more than I hoped for. Thank you very much.

I now understand the origin of Ladders, the concept of which is rarely seen in Japan, and that many things came from the Chess world.

Did OGS have a rating system at the time? Was it Elo? I learned a bit about Mark Glickman, the US Chess Ratings Committee Chair until 2019, when OGS adopted Glicko-2 system. As it seems it is still being far from perfect for OGS, do you have any insight?

5 Likes

Yes, it started as ELO, with 1 rank = 1 stone, which I think acts as a pretty good basic system for rating generally. In my opinion Glicko2 is just a good evolution of the ELO system, which suffers from certain simplicities (particularly ignoring rapidly improving players, or at least players not playing very much therefore changing nominal strength quite quickly over the course of “games played”).

Ratings and ranks are something I can sort of rant and rave about for a long time as I have fairly strong opinions which I’ll try to curb somewhat, but I suspect there is no perfect system, and what is adopted I think depends on what you’re trying to achieve. It is a lot harder in Go where there is a lot of intrinsic psychological baggage attached to rank.

Professional dan titles make a lot of sense, where they are earned and not lost, and earned based on specific achievement milestones. Outside of that, the purpose of any form of rank and rating system is to provide sensible matches, an understanding of “approximate” strength, and relatively uniquely to Go a benchmark for appropriate handicapping. Despite this, people have great desire to be SDK, or reach the coveted dan ranks, despite the fact that amateur dan ranks are so variable.

Essentially all of the main numerically based rating systems create a bell distribution of players, and the more players you have, the further towards each end of an infinite scale the outlying players go as the curve increases in scale. Furthermore, if there is a steady improvement in overall playing strength of a playerbase without some form of controlling factor, you get rank drift. A crude example of this would be if 400 people are all 6k and only play each other yet improve at the same speed, once they improve by 6 stones, they will still be 6k as the only place to grab rating points from is each other. despite having all become considerably stronger. There was a big discussion when I was a British Go Council member on how to handle this (the EGF slowly injected points into the system by a small amount for each submitted game, I don’t know if this is the case still).

As a result, all you can really aim for with a rating system is “anything that achieves the following”:

  1. People who don’t play that often don’t end up finding their rating incorrectly assessing their comparative strength for too many games before it corrects itself (Glicko2 is actually pretty good at this)
  2. The grades that are assigned to the rating levels offer more or less correct handicap (interestingly, the faster the time control, the more of a handicap is generally required, so this may need different weighting for different game speeds - again, OGS does this well, and that’s a post merge change with the Nova crew!)

… and that’s pretty much it. Whether a specific place tells me I’m 13 kyu or 6 dan doesn’t overly bother me, what’s important is that it’s internally consistent rather than the specific rank it tells me I am. I also think this is why it’s so important that if someone tells you their strength, they also tell you where that strength was given - KGS 1 dan, OGS 12 kyu etc, is of much more value than the rank by itself.

I think the rating system for OGS seems pretty good, and serves it’s purpose well. I do wonder if it seems a bit top heavy, as there seem to be a large number of very strong dans and not very many players around the 20 kyu mark, but maybe that’s because I’ve been hanging with the wrong crowd. Any form of rank to rating adjustment would fix it if that’s the case, but it’s not particularly critical to change anyway for the reasons I’ve mentioned above.

12 Likes

CyberOro

Interesting, not sure I ever heard of this server. Perhaps a couple of times.

The only Go site that died in my time was GoGameGuru. Actually, is CyberOro even dead? idk

About OGS ranks being top-heavy: as I’m sure you know, the ranks inflated by about 2 – 3 stones for many people after the transition from Elo to Glicko, and the inflation was more pronounced the higher-ranked you already were. krnzmb (6d) even became “12d” for a short period until an artifical 9d display-cap was brought in.

5 Likes

We have much more low ranked players than dans [Outdated] Improved OGS rank histogram

10 Likes

Many thanks flovo, to my eye that distribution looks about ideal. Yet another plus point for how OGS is handling ranks and ratings :slight_smile:

@bugcat Fairly certain CyberOro is now wBaduk, but my memory of a decade ago is hazier than I’d like!

8 Likes

Yeah, I think you’re right.

1 Like

the one downside is that the main complaint I consistently see raised against OGS is that “there aren’t enough high dan players” which, ironically, keeps other high dan players away… so while mathematically it might be ideal to have the centre of the bell curve (and thus the bulk of the population) in the kyu ranks, the psychological effect on the community results in some strong players unfortunately avoiding our server.

6 Likes

I find that somewhat amusing, if a bit sad at the same time. The last tournament I entered here a couple of months ago put me up against 4 players ranked 7d+, one of whom an ex-insei. I didn’t even have to pay for the privilege! You certainly won’t see me complaining (other than “my group was totally harder than the other McMahon score = 0 group!”, but to be fair I’d rather get opportunities to play top players anyway, it’s not like I’m ever going to win the thing in that field).

I’ve never had the opportunity for a similar experience on DGS, and even though players at that level do exist in public games on KGS, getting to play them without paying for a teaching game almost doesn’t happen until you’re at least 5d or so yourself. Obviously the work the far eastern servers have done to make themselves more easily accessible to those with no grasp of Chinese / Korean etc (with more userfriendly clients and so on) has allowed access to servers with a wide range of top players and professionals, but you’re unlikely to compete with that unless there’s a similar cultural assimilation of Go and professional playerbase in Europe / USA - lots of strides made in that direction, but obviously there’s still a fairly giant gap there.

It would be interesting to know what demographic the average complainer comes from, but I wouldn’t want to derail the thread too far down that road, as I’m sure this is something that you’ve already discussed with the rest of the team at great length. Either way, I have never seen any server where playing games against top players is so accessible to the rest of us mortals.

12 Likes

Thank you very much for the insight.
Setting the goals for the system to achieve instead of seeking a perfect rating system seems like a good approach.

If you find a piece of record or whatever that looks fit for recording the history of OGS, please add them here as you find them.

Thanks again.

6 Likes

This is a reasonably amusing picture of me at a go tournament in the UK promoting the server with my fab merch - I would guess this is Feb/March 2007 :smiley:

15 Likes

Nice to see you again, @topazg, I had feared that you got lost :slight_smile:

And thanks a lot for telling us the story of the old OGS … would it be asking too much if I’d ask you to add a few dates (approximate year) to your history of OGS? I think that would be really helpful for understanding the Western Go scene back then.

3 Likes

Sorry, I seem to keep failing to notice new posts here!

So the site became “public” in November 2005. The first mini-tournaments existed in December 2005, with 13x13 boards being implemented and a sort of beta 19x19 board at the same time.

A lot of the early players were family and friends: fingolfin was Andrew’s brother, Ariadne was my sister etc etc. The goal was to get enough people in that actual go players would at least have more opposition than just myself, and it made things like mini-tournaments feasible. They supported fantastically really until we had enough players that they could go off and do things they’d rather do with their time instead :smiley:

Full site wide major tournaments started mid 2006 before becoming the different titles that are still maintained now. The handover to dmg happened around 2008 I believe, as I was getting divorced from my first wife and just didn’t have any time for it any more. dmg converted the site to PHP and then maintained it up until the merge with Nova. Although I still had access to stuff and helped out quite a bit, dmg definitely did the majority of everything from his takeover up to the merge.

Much of the big changes occurred post Nova, so the current administration would be far better equipped to answer :slight_smile:

I also found my first ever 19x19 game lol: https://online-go.com/game/13141

11 Likes

I joined OGS shortly before this Discourse-based forum was launched, replacing the custom forum integrated into the main site.

I barely remember what those older forums were like, and I was not nearly as active in them as I am now here. Does anyone remember those forums and could discuss a bit about them?

Were the archives of those old discussions saved anywhere?

5 Likes

https://online-go.com/player/519197/

(A screenshot of the page to have better chances of survival if something bad happens to OGS :P)

Edit: chat icon
image

10 Likes

I just found this picture.

5 Likes

@李建澔2, you may want to read this thread from the beginning, that is: from the first post :wink:

7 Likes

According to timestamp 2017 January 27. We were young and hopeful back then.

4k overall, 6k live and 5k correspondence. I played as 5-6k EGF at the time, if ratings weren’t retroactively changed or something. This corresponds to:

  • 3k (2.4k) first Glicko rating (rating-history);
  • 4k (3.2k) changed Glicko rating (glicko2-history);
  • 1k (0.2k) current Glicko rating (v5-rating-history).
5 Likes

image

3 Likes