Books 
From Better Backgammon, by Tim Holland

Should White double to 2? 
White should not double. Once again I will give you a problem which illustrates a direct contrast to the preceding problem.
As I explained in Position 41, the determining factor in whether to double is the ratio of probable gammons to probable losses.
In this game, whether it’s played 36 times or 360 times, you would have to be very unfortunate to lose even 5 percent of the time. On the other hand, I would be extremely conservative in saying that you will win a gammon 25 percent of the time.
Let’s use our simple arithmetic once again. This time we’ll do it on the basis of 100 games (I choose 100 rather than 36 in order to avoid fractions).
Assume you do not double:
70% of the time you win 1 point  +70 
25% of the time you win 2 points (a gammon)  +50 
5% of the time you lose 1 points  −5 
Net result  +115 
Assume you do double:
You win 100 games at 1 point  +100 
There is no point in discussing what would happen if you were to double and your opponent were to accept; for Houdini at his best would have to decline. As you can see, here is a game where you must not double.
Rollout
Tom Keith 2013 

Money play Centered cube White on roll 1296 games with VR Checker play: 3ply Cube play: XG Roller 
Cube Action  Game  G  BG  Equity  
No double 
W L 
.9688 .0312 
.5203 .0002 
.0471 .0000  +1.5443  
Double  Take 
W L 
.9683 .0317 
.5214 .0002 
.0474 .0000  +2.9911  +1.0000  (0.5443)  
Drop  +1.0000 

