Suggestion for improvement of Temperature Map for Cube decisions.

[Ian Shaw]

Hi all,

In the Facebook group Backgammon News, Wayne Joseph raised some questions about 
the values displayed in the Temperature map. I've posted extracts from the 
discussion below for context, but the gist of it is.

The user was confused by the equities in the Temperature map being normalised 
to a 1-cube for both cube actions. This obscures the effect of doubling the 
stakes. 


1.      I think the values in the doubled part of the map should be normalised 
to a 2-cube rather than a 1 cube. 
This would make the value in the Top left to f the Doubled Box (Item 2) on the 
screen shot equal to the Double, Take Equity in the Analysis window.
It would also make it easier to compare the value of each roll, with and 
without a double. 

2.      The scale bar display does seem to anticipate that the right-hand side 
will show the doubled equity. In the example below, the max scale of +1.385 is 
about twice the value of the best roll 66 +0.659 for 66 and the minimum scale 
-0.077 is about twice the value of the worst roll -0.036.


I also think the map could benefit from showing market losers in a different 
colour. I don't find the shades of red all that easy to  follow. 


https://www.facebook.com/groups/1017340842031433/posts/1897126800719495/?comment_id=1899875730444602&notif_id=1707695820400391&notif_t=feedback_reaction_generic&ref=notif

Transcript follows:

Wayne Joseph
Sho Sengoku's Temperature Map is very useful to visualize the dangers and 
jokers of a move, as well as the volatility of cube decisions. 
Can anyone give clear definitions for what the items labelled 1-4 represent? I 
couldn't find a full explanation in the manual. [Ian notes - my previous 
attempt to post his screenshot of the temperature map failed].

HdvGAAiw3w0YAA:UQmgAAAAAAAE 

    GNU Backgammon  Position ID: HdvGAAiw3w0YAA
                    Match ID   : UQmgAAAAAAAE
    +24-23-22-21-20-19------18-17-16-15-14-13-+  O: gnubg
    | O  O        O  O |   | O  O     X  O    |  0 points
    |    O        O  O |   | O  O     X  O    |  
    |    O             |   |                  |  
    |                  |   |                  |  
    |                  |   |                  |  
    |                  |BAR|                  |v 5 points match
    |                6 |   |                  |  
    |                X |   |                  |  
    |                X |   | X                |  
    |             X  X |   | X  X             |  On roll
    |       O     X  X |   | X  X             |  0 points
    +-1--2--3--4--5--6-------7--8--9-10-11-12-+  X: ianshaw (Cube: 2)
                    Pip counts: O 103, X 113


Cube analysis
4-ply cubeless equity +0.273 (Money: +0.270)
  0.618 0.092 0.003 - 0.382 0.060 0.001
Cubeful equities:
1. No double           +0.513
2. Double, pass        +1.000  (+0.487)
3. Double, take        +0.384  (-0.129)
Proper cube action: No redouble, take (20.9%)

 


Ian Shaw
1 & 2 are the overall equities of each option (the sum of the 36 rolls).
3 & 4 should be the max and min range of the colour scale. But the outliers on 
the map in your picture are -0.036 to +0.659, so I'm not sure.

Wayne Joseph
For items 1+2, please could anyone explain in layman's terms why cubing to 4 
(ownership) results in around a noteworthy 60% decrease in the overall 
normalised equity (equivalent money game), given they are based on *exactly* 
the same 36 rolls? 60% less equity sounds like a lot to give up 'just' for 
redoubling.

Ian Shaw
If you look at the top-left corner of the 4-cube map, the value (+0.221 )is 
about half the value of the Double/Take Cube Analysis (+0.424).
That's because the values are in emg, so they have been re-normalized. This 
allows you to compare the value of the individual rolls on an equal basis, to 
see how much better they are.
For instance, the equity for 66 is +0.648 on a 4-cube, but on a 2-cube it is 
+0.965. This will be because after 15/3*(2) you can probably cash most 
sequences if you own the cube, but having redoubled you must still complete the 
bearoff to win.

Wayne Joseph
It's highlighted to me I still don't have my head around 'equity' (which 
internally, cognitively I kind of interpret as 'overall reward-overall risk'). 
Or the utility of those numbers.
>> +0.648 on a 4-cube,
>> but on a 2-cube it is +0.965.
Attached is a screenshot of the position after rolling 66 and hitting.
Looking at the screenshot, those numbers feel very counterintuitive (probably 
just me)
To me, OTB, at 83% favourite I would much prefer to be on a 4 cube (opponent 
held) in the screenshot, rather than holding a 2 cube. It feels like I would be 
fairly confident of bringing this home, for more reward (equity?) i.e. 4 points 
is better than 2
...but then I would have substantially *less* 'equity'. In a situation that 
feels 'better'???
>> +0.648 on a 4-cube,
>> but on a 2-cube it is +0.965.
May I sense check - OTB at 83% favourite would you (or others) prefer to be in 
the post 66 position on a 4 cube (opponent held). Or would you really prefer to 
be just holding the cube at 2?

Ian Shaw
66 was your best roll. So yes, you wish you had doubled if you roll it. Your 
doubled equity of 0.648*2 is better than the 0.965 from not doubling. If you 
still hold the cube, you are going to double him out unless he rolls 45 or 46, 
which is why the equity is almost 1.0.
On the other hand, if you roll that 62, you'll be very glad not to have doubled 
the stakes.

Wayne Joseph
Ok, from what you said, it sounds like there is a extra cognitive step required 
of manually multiplying all values in the 4 cube section by 2 to more easily 
compare like for like?
It feels like it might be a clearer and more intuitive UX to have the 4 cube 
values presented with the necessary subsequent multiplication step performed 
already? (i.e. 1.296 vs 0.965 after 66 is much more intuitive to me in showing 
66 has more value (and the difference of that increase) after cubing, rather 
than displaying 0.648 vs 0.965 which 'sounds' worse)


Ian Shaw
I think you have a good point. The numbers in the Double window should reflect 
the values in the analysis display, but in more detail. If I remember, I'll 
post some of this discussion on the gnubg mailing list. Perhaps one of the devs 
has a counter argument.


-----Original Message-----
From: bug-gnubg-bounces+ian.shaw=riverauto.co.uk@gnu.org 
<bug-gnubg-bounces+ian.shaw=riverauto.co.uk@gnu.org> On Behalf Of Philippe 
Michel
Sent: Tuesday, February 20, 2024 9:53 PM
To: bug-gnubg@gnu.org
Subject: Preview of forthcoming gnubg release

I have uploaded a new gnubg Windows build at 
http://philippe.michel7.free.fr/gnubg/

This should fix the bugs in 1.08.001 reported here or by other means:

- the gnubgautorc file corruption

- the "Move" quirk in the GUI menus

- "Analyse/Show Records" becoming inactive after using "Analyse/Show 
   Records/Plot History" (reported and fixed by Isaac Keslassy)

- a mistake in the description of the Isight racing count

and improve the pasting of XGIDs by accepting those copied from XG by a 
Ctrl+C (copy XGID+Board) instead of only those from Ctrl+Shift+C
(contributed by Takahashi Kaoru).

Plus complete Finnish and Japanese translations.


Regarding the gnubgautorc file corruption, I couldn't reproduce it at 
will and it happened only once for me in a week of use of the unfixed 
version. I'm confident it should be ok now but, if the problem hit you 
more frequently, please test this new version and report how it goes.


Barring another bad surprise on the gnubgautorc issue I will build and 
upload a 1.08.002 version this week-end.