Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
profiling the bots
9 views
Skip to first unread message
Paul
May 25, 2012, 12:11:40 PM5/25/12
to
When the world's strongest money players (probably the top bots if you
ignore the possibility of reaching a bizarre position a bot doesn't
understand and jacking up the cube) play against each other, it would
be interesting to see a breakdown of the frequency of each type of
result.
For example: P1 % of the time, an initial double is dropped. P2% of
the time, the cube ends up on 2 with the taker losing a single game.
P3% of the time, the cube ends up on 2 with the taker losing a
gammon ... etc. etc.
I think that this type of knowledge would be a useful flag to inform
players when they or their opponents are doing something wrong. For
example, it's far more common for the cube owner to get a gammon (a
successful execution of too-good) among very strong players than it is
among intermediates or beginners.
Thank You,
Paul Epstein
ignore the possibility of reaching a bizarre position a bot doesn't
understand and jacking up the cube) play against each other, it would
be interesting to see a breakdown of the frequency of each type of
result.
For example: P1 % of the time, an initial double is dropped. P2% of
the time, the cube ends up on 2 with the taker losing a single game.
P3% of the time, the cube ends up on 2 with the taker losing a
gammon ... etc. etc.
I think that this type of knowledge would be a useful flag to inform
players when they or their opponents are doing something wrong. For
example, it's far more common for the cube owner to get a gammon (a
successful execution of too-good) among very strong players than it is
among intermediates or beginners.
Thank You,
Paul Epstein
Tim Chow
May 25, 2012, 8:34:08 PM5/25/12
to
On May 25, 12:11 pm, Paul <pepste...@gmail.com> wrote:
> When the world's strongest money players (probably the top bots if you
> ignore the possibility of reaching a bizarre position a bot doesn't
> understand and jacking up the cube) play against each other, it would
> be interesting to see a breakdown of the frequency of each type of
> result.
>
> For example: P1 % of the time, an initial double is dropped. P2% of
> the time, the cube ends up on 2 with the taker losing a single game.
> P3% of the time, the cube ends up on 2 with the taker losing a
> gammon ... etc. etc.
If you have GNU Backgammon, I think that you can get this information
> When the world's strongest money players (probably the top bots if you
> ignore the possibility of reaching a bizarre position a bot doesn't
> understand and jacking up the cube) play against each other, it would
> be interesting to see a breakdown of the frequency of each type of
> result.
>
> For example: P1 % of the time, an initial double is dropped. P2% of
> the time, the cube ends up on 2 with the taker losing a single game.
> P3% of the time, the cube ends up on 2 with the taker losing a
> gammon ... etc. etc.
by setting up the initial position, and doing a long rollout of the
cube decision. Then you can click on the "View Statistics" button to
see all the information you're asking for (under the tab corresponding
to not doubling).
There are a couple of subtle options that I think you need to pick to
get good results. If I recall correctly, there's a "roll out as
initial position" box, to make sure that you can't roll doublets on
the first roll. Also, you can turn off variance reduction. This
should speed up the computation and it won't affect the "View
Statistics."
---
Tim Chow
0 new messages