Google Groups no longer supports new Usenet posts or subscriptions. Historical content remains viewable.
Dismiss
GnuBg - the 'proper cube action' statistic?
31 views
Skip to first unread message
sebalotek
May 19, 2013, 12:28:57 PM5/19/13
to
Hallo all,
I have a couple of questions.. I tried to post on the gnubg old nabble message board, but it doesn't look like it's accepting new posts at the moment.
Please can anybody explain the 'proper cube action' statistic in GnuBg to me.
For example, on the bottom right hand corner of the Gnubg analysis section I saw a figure of 15.2% for a bad No double / Take. (Sorry I can't post the image on this Google group)
What is the figure meant to represent in laymans terms?
And where does this 15.2% come from i.e. how is it calculated?
Many thanks for any help explaining this.
<nabble_img src="proper+cube+action+stat.jpg" border="0" />
I have a couple of questions.. I tried to post on the gnubg old nabble message board, but it doesn't look like it's accepting new posts at the moment.
Please can anybody explain the 'proper cube action' statistic in GnuBg to me.
For example, on the bottom right hand corner of the Gnubg analysis section I saw a figure of 15.2% for a bad No double / Take. (Sorry I can't post the image on this Google group)
What is the figure meant to represent in laymans terms?
And where does this 15.2% come from i.e. how is it calculated?
Many thanks for any help explaining this.
<nabble_img src="proper+cube+action+stat.jpg" border="0" />
Tim Chow
May 19, 2013, 4:29:34 PM5/19/13
to
On May 19, 12:28 pm, sebalotek <djskope2...@yahoo.co.uk> wrote:
> I tried to post on the gnubg old nabble message board,
> but it doesn't look like it's accepting new posts at the moment.
The mailing list bug-...@gnu.org is still active.
> I tried to post on the gnubg old nabble message board,
> but it doesn't look like it's accepting new posts at the moment.
> Please can anybody explain the 'proper cube action' statistic in GnuBg to me.
http://www.gnu.org/software/gnubg/manual/html_node/Getting-cube-decision-hints-while-playing.html
> For example, on the bottom right hand corner of the
> Gnubg analysis section I saw a figure of 15.2% for
> a bad No double / Take.
> What is the figure meant to represent in laymans terms?
This means that technically you shouldn't double. However, if there's
at least a 15.2% chance that your opponent will wrongly pass, then
it's worth doubling.
> And where does this 15.2% come from i.e. how is it calculated?
equity is 0.750. Then doubling loses 0.050 in EMG equity if your
opponent correctly takes. However, if your opponent incorrectly
passes then you gain 0.200 equity compared to not doubling. In this
case, GNU will report a figure of 20%. If there's a 20% chance of
gaining 0.200 then that balances out the 80% chance of losing 0.050.
More precisely, the formula is (loss)/(loss + gain) = 0.050/(0.050 +
0.200) = 0.20.
---
Tim Chow
Michael Petch
May 19, 2013, 9:51:06 PM5/19/13
to
On 19/05/2013 10:28 AM, sebalotek wrote:
>
> I have a couple of questions.. I tried to post on the gnubg old nabble message board, but it doesn't look like it's accepting new posts at the moment.
>
> Please can anybody explain the 'proper cube action' statistic in GnuBg to me.
>
Tim answered your question very well. I wish to point out that I
>
> I have a couple of questions.. I tried to post on the gnubg old nabble message board, but it doesn't look like it's accepting new posts at the moment.
>
> Please can anybody explain the 'proper cube action' statistic in GnuBg to me.
>
answered your question on the GNUBG mailing list. We received many
duplicate posts (I don't trust Nabble except as an archive). The most up
to date posts can always be found here (you can go to all the threads in
a given month by clicking "thread"):
http://lists.gnu.org/archive/html/bug-gnubg/
To see the posts from May you can can go directly here:
http://lists.gnu.org/archive/html/bug-gnubg/2013-05/index.html
You will notice that I posted a response this morning:
http://lists.gnu.org/archive/html/bug-gnubg/2013-05/msg00019.html
If you want to get proper access to the GNUBG mailing list (for bugs or
questions) I recommend you subscribe to the mailing list directly. That
can be done via:
https://lists.gnu.org/mailman/listinfo/bug-gnubg
--
Michael Petch
GNU Backgammon Developer
OpenPGP FingerPrint=D81C 6A0D 987E 7DA5 3219 6715 466A 2ACE 5CAE 3304
sebalotek
May 19, 2013, 11:04:48 PM5/19/13
to
Firstly, apologies to people on the Old Nabble board for the multiple posts. Everytime I clicked submit, I was shown a cryptic Http 500 error page which I have never encountered before and which I thought meant that my message was not successfully posted - so I removed the attachment, edited text and tried again, ..and again. How embarrassing! :-/
Thanks for the replies -
TC - "However, if there's at least a 15.2% chance that your opponent will wrongly pass, then it's worth doubling."
MP - "In your case it would mean that if you think your opponent will drop the
cube at least 15.2% of the time then it would be a correct double."
I find this interesting - in real life games, how does one arrive at a figure of the percentage chance that the opponent will wrongly pass? How can I know what my opponent is thinking inside their head?
Is it based on what I perceive is their skill level? Or the complexity of the position? Or some other factors?
Thanks again!
Thanks for the replies -
TC - "However, if there's at least a 15.2% chance that your opponent will wrongly pass, then it's worth doubling."
MP - "In your case it would mean that if you think your opponent will drop the
cube at least 15.2% of the time then it would be a correct double."
I find this interesting - in real life games, how does one arrive at a figure of the percentage chance that the opponent will wrongly pass? How can I know what my opponent is thinking inside their head?
Is it based on what I perceive is their skill level? Or the complexity of the position? Or some other factors?
Thanks again!
Tim Chow
May 20, 2013, 11:38:02 AM5/20/13
to
On May 19, 11:04 pm, sebalotek <djskope2...@yahoo.co.uk> wrote:
> I find this interesting - in real life games, how does one
> arrive at a figure of the percentage chance that the opponent
> will wrongly pass? How can I know what my opponent is thinking
> inside their head?
There's no way to pin down this probability with certainty. You just
> I find this interesting - in real life games, how does one
> arrive at a figure of the percentage chance that the opponent
> will wrongly pass? How can I know what my opponent is thinking
> inside their head?
have to observe your opponent carefully and use your best judgment,
based on what you've seen your opponent do in the past and (in a live
match) what the opponent's body language seems to be telling you. In
the absence of specific knowledge about this particular opponent, you
can use your judgment about other opponents you've had in the past.
What percentage of them do you think would pass?
Depending on how comfortable you feel with "gamesmanship," you may be
able to influence the probability in a live match with your own body
language. If you double very quickly and confidently, for example,
that may increase the chances that your opponent will drop.
---
Tim Chow
sebalotek
May 21, 2013, 3:00:43 AM5/21/13
to
Thanks for your examples Tim - the subtleties of body language, banter, (double) bluff and "gamesmanship" are part of why backgammon is such a great game.
>> In the absence of specific knowledge about this particular opponent, you
>> can use your judgment about other opponents you've had in the past.
>> What percentage of them do you think would pass?
Good question - I guess against gnubg or other programs the probability of them incorrectly passing or incorrectly taking is always close to zero.
For humans, I'm not really sure - on average perhaps 5% of the time? With no prior knowledge of the opponents skill level, aggressiveness or mood does 5% sound like a good estimate, or is this guess too high?
Cheers,
sebalotek
>> In the absence of specific knowledge about this particular opponent, you
>> can use your judgment about other opponents you've had in the past.
>> What percentage of them do you think would pass?
For humans, I'm not really sure - on average perhaps 5% of the time? With no prior knowledge of the opponents skill level, aggressiveness or mood does 5% sound like a good estimate, or is this guess too high?
Cheers,
sebalotek
Tim Chow
May 21, 2013, 1:10:00 PM5/21/13
to
On May 21, 3:00 am, sebalotek <djskope2...@yahoo.co.uk> wrote:
> For humans, I'm not really sure - on average perhaps 5% of the time?
> With no prior knowledge of the opponents skill level, aggressiveness
> or mood does 5% sound like a good estimate, or is this guess too high?
It depends hugely on the position. For some positions, over half of
> For humans, I'm not really sure - on average perhaps 5% of the time?
> With no prior knowledge of the opponents skill level, aggressiveness
> or mood does 5% sound like a good estimate, or is this guess too high?
all players would make an incorrect decision. For other positions,
essentially zero would.
If your off-the-cuff guess is 5% then I strongly suspect you don't
double aggressively enough, and that if you stepped up your doubling
tendencies you'd be shocked at how many of your opponents drop takable
doubles.
---
Tim Chow
sebalotek
May 21, 2013, 4:56:33 PM5/21/13
to
Thanks again for the feedback Tim.. I said 5% initially because I didn't want to insult anybody by underestimating their abilities. For me myself, I'm sure it's a lot more! I think I double reasonably aggressively, but it's a very fine line to balance and sometimes I think the gnubg stats show I am mostly too aggressive (however, I'm not the sharpest tool in the box and find my figures on the 'Show Records' page difficult to interpret).
For tricky cube decisions, I sometimes force myself to delay my cube action by one extra move after I initially thought I should cube. It sometimes pays off, but I've observed that a noticeable amount of the time I lose my market on the next roll and the computer drops.
For tricky cube decisions, I sometimes force myself to delay my cube action by one extra move after I initially thought I should cube. It sometimes pays off, but I've observed that a noticeable amount of the time I lose my market on the next roll and the computer drops.
0 new messages