Questions That Don't Deserve Their Own Thread

immagine

immagine

:running_woman:

Yes, I know I’m not the norm. And I trust ranks much less than almost everyone else probably.

But surely you know your rank better than the OGS system does. And so do I. I just know you and me can play an even correspondence game and have fun. And even though we never met personally or even played a live game here on OGS I am pretty sure you’d have to give me about 5 stones in a real time game to make it interesting. So even if rank measurement fluctuates and rank may depend on circumstances, doesn’t mean it’s not a useful concept.

1 Like

As far as I understand, the current OGS rating algorithm has been designed specifically with the intent to be much more rapidly fluctuating than most go ratings from other go servers or national associations.

I’m almost positive I’ve seen code where x is 30. But now I’m curious, I’ll look around when I have time.

And of course, it’s always hard to figure out what’s actually used in the backend. But for the most part, the go logic exists in the open source goban repo.

Edit: hey looks like my memory serves me correctly!

@ajiKacey Doesn’t look like there are hashes or a hashset involved, although that would be a reasonable approach as well.

3 Likes

I currently have a casual rengo in the “dropped games” section. It is now in the dame phase, likely to end in the very near future.

What happens when it ends? Does it move to a “Recently completed dropped games” section, so that I can see the result? Or does it disappear completely without informing me of the result?

I believe rengo should land in your Game history just like normal games.

1 Like
Prelude

19x19 board, ‘k’ komi, no handicap game. What is Blacks expected win rate?
AI quickly proclaims “50%”. A human respondent somewhat agrees.

It’s Michael Redmond playing White. Same question…
AI quickly proclaims “50%”. The human respondent disagrees.

It’s a DDK playing Black.
AI quickly proclaims “50%”. The human respondent strongly disagrees.

It’s asymmetrical time settings; White 0.2 sec absolute, Black 10 days
AI quickly proclaims “50%”. The human respondent strongly disagrees.

Would it be technically possible to display a graph of ‘realistic expected win rate’, weighted for player strengths and remaining game times?

Bonus question: is the first question merely a thinly veiled feature request to software developers everywhere? Yes. Yes it is. This is what the fans want.

A pretty similar question was asked on chess.stackexchange two weeks ago:

1 Like