|
From: | Joseph Heled |
Subject: | Re: current development |
Date: | Sun, 26 Jan 2020 08:36:24 +1300 |
Lots of interesting comments by Sarah Payne and Chris Bray, most of which
make sense to me. But there's one part that I have trouble believing.
> In my opinion, a phone version would broaden appeal / access but the
> most critical issue is the neural nets. XG feels like a very different
> animal as an opponent - noticeably more opportunistic & aggressive, so
> some degree of congruence asap seems critical for gnubg to hold ground.
This sounds like psychological confirmation bias to me. I don't think
that XG and GNU play differently enough for any human to notice, unless
we're talking about the "crippled" modes where the bot intentionally makes
errors.
To be more specific, I would guess that over the course of 100 games,
there would be only a handful of decisions that GNU 2-ply and XG 3-ply
would make differently (excluding those non-contact positions where it
doesn't really matter what you play), and I very much doubt that XG would
be consistently "more opportunistic and aggressive." I would be happy to
be proved wrong, though.
Tim
[Prev in Thread] | Current Thread | [Next in Thread] |