On May 10, 2023 at 1:59:02 AM UTC-6,
peps...@gmail.com wrote:
On May 9, 2023 at 8:57:11 PM UTC+1, MK wrote:
On May 9, 2023 at 6:52:38 AM UTC-6, Timothy Chow wrote:
XGID=-c--B-C-C--BbCa-ac-e----AA:0:0:1:55:0:0:0:0:10
X to play 55
bar/3 8/3 13/20
I really like this play which makes a valuable inner
board point and prepares to anchor on the 20 point.
My only concern ..... is that this play is utterly
pedestrian and obvious, and therefore anti-QF.
But Tim does post anti-QF problems sometimes.
I'll endorse your play -- bar/3 8/3 13/20
I'm glad that you endorsed my play and by the end
of this post, you will be glad that you did so too.
BTW, did you notice that I played all the 5's in three
clean moves, without experimenting with various
moves and shuffling pieces around endlessly..? ;)
Both Ex-Gee and Noo-Bg prefer bar/15 8/3(2) to
bar/20 13/3 8/3 (my bar/3 8/3 13/20 in legalese:).
So, since bots don't make the "obvious" play, first
I wondered if Tim sometimes posts positions with
bots being wrong, (like the one I had commented
on extensively not to long ago), but after seing his
post with rollouts, I realized that it's not the case.
He is just bott-kissing as usual, as everyone else. :(
Before moving on to demonstrate that the bots
are wrong, let me add my comments to yours. It
took me about 8-10 seconds to decide my play.
Bar/3 8/3 13/20 leaves a blot on 20 that O can hit
and also make a point at once, but X would only
lose 5 pips and then still have a very good chance
of entering right back and also build an anchor at
once as well. In contrast, bar/15 8/3(2) leaves an
"unprotected" blot on 15 which would go back 10
pips if hit and creates an additional blot on 8 which
would go back a whopping 17 pips and leave X's
outer board deserted. If O hits loose X's blot on 20,
I would double and then raccoon if O beavers.
This is an excellent early position to show that the
bots are wrong by doing "clinical rollouts" as I had
explained at length in my thread: "BG rollouts should
be done like placebo-controlled clinical trials":
https://groups.google.com/g/rec.games.backgammon/c/Z1QsqqGbTsw/m/x-wV5kisBAAJ
and in my reply in Tim's thread: "Slot or not?":
https://groups.google.com/g/rec.games.backgammon/c/tH5CbO_-8c4/m/G6EeVYu0BAAJ
This time I'll keep short by skipping the intermediary
steps and doing only 4 rollouts, again using Noo-Bg
since Ex-Gee doesn't have the necessary functions.
GNUbg ID: 4HPJADiYc8wBUA:cIkWAAAAAAAE
Rollout with both players "world class":
bar/15 8/3(2) cubeful= +0.891, cubeless= +0.570
bar/20 13/3 8/3 cubeful= +0.760 (-0.131), cubeless= +0.524 (-0.046)
Notice how the equity difference becomes already
smaller just by eliminating the jackoffski bias.
Rollout with both players "expert" (0-ply) and
with maximum noise (i.e. almost random play):
bar/20 13/3 8/3 cubeful= +0.756, cubeless= +0.906
bar/15 8/3(2) cubeful= +0.589 (-0.167), cubeless= +0.719 (-0.187)
Wow! Unlike the close numbers in rollouts in my
above linked thread, this time the top two plays
are reversed decisively, by a very similar equity
difference but in the opposite direction.
Just to double check, if we rollout the resulting
positions from the top two plays, (with the same
rollout settings, before the next dice roll), we get
these similar results showing that bar/20 13/3 8/3
is indeed the better/correct play:
GNUbg ID: 4HPJADhszpgBIQ:cAkAAAAAAAAE
bar/20 13/3 8/3 cubeful= +0.726, cubeless= +0.714
GNUbg ID: 4HPJADhsTswJIA:cAkAAAAAAAAE
bar/15 8/3(2) cubeful= +0.673, cubeless= +0.698
I don't remember if I ever said anything about why
I bother to do cubeless rollouts for money game
positions. So, let me explain here while at it.
The equities of Tim's position, (before the 55 is
rolled), are these:
GNUbg ID: 4HPJADiYc8wBUA:cAkAAAAAAAAE
Cubeful = +0.439, cubeless= +0.430
If you wanted to cube like the "Murat mutant" in
Axel's experiment, i.e. turn a cubeful game into
cubeless by forcing it to be played out to the end,
you would pick your moves accordingly and doing
"cubeless clinical rollouts" may retroactively tell
you how you would have done.
Well, Paul, there it is. Congratulations for picking
the right play in a position where the bots pick
the wrong play. I wonder if you may end up being
the first one to cross the line over to my side, (as
opposed to my previous wonderings wheather Tim
would be the one)...?
MK
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)