Red to play down 1-0 in match to 7.

What is the benefit for red to delay making points in the inner board? When Gnubg plays 6/5(3), 3/2, shouldn't black try to disengage with a small lead in the race on a roll of 6-3 or even 6-2? I expected to see 6/5(2), 6/4 played which is apparently a virtual coin flip compared to the actual move.

1. Cubeful 3-ply 6/5(3) 3/2 MWC: 41.87%
40.9 0.4 0.0 - 59.1 0.5 0.0 3-ply cubeful prune [grandmaster]
2. Cubeful 3-ply 6/4 6/5(2) MWC: 41.86% ( -0.01%)
40.9 0.4 0.0 - 59.1 0.5 0.0 3-ply cubeful prune [grandmaster]
3. Cubeful 3-ply 6/4 6/5 3/2 MWC: 41.83% ( -0.04%)
40.8 0.4 0.0 - 59.2 0.7 0.0 3-ply cubeful prune [grandmaster]
4. Cubeful 3-ply 8/7 6/5(3) MWC: 41.82% ( -0.05%)
40.7 0.4 0.0 - 59.3 0.5 0.0 3-ply cubeful prune [grandmaster]

Perhaps it is statistically better to time the completion of the inner board when black may be forced to leave a blot, but I can't imagine why it is equivalent to just making the four point immediately. The timing of disengagement favors red after the excellent 1-1 roll. Black will often lose ground in the race the longer it takes to break and run, so I have a hard time believing it is okay to encourage black to try and run with a roll like 6-3. It might be worth remembering 1-1 can be played like this for the possible shock value in a live match against a human opponent though.

For some reason odd-ply and even-play choices on gnubg occasionally differ. I think I read that the play gets better as you add more depth (as it should!), but that the even-ply equity calculations are typically closer to actual equity (given by rollouts) than the odd ones, so even should be better for cube decisions. There is some discussion here about it.
The reason I point this out is because 2-ply and 4-ply both slightly prefer 6/5(2) 6/4. I don't really see any reason not to do this now either. If someone chose the other play, I agree with you--as black I'd run with 63 immediately.

Maybe it has to do with so many branches of the decision tree leading to no blot situations. With the end of the race nearer on each ply, I can see how evaluations after a black roll would have a bias toward initial roll board smoothness instead of pointedness for red.

I might tighten up the pruning and experiment with something higher than 3-ply, if it doesn't take several minutes to calculate each roll. It's good to know there's some discussion about the ply-effect since one of the developers might eventually be able to find a way to compensate.

The short answer: you aren't getting a shot here [or for several rolls] so it doesn't really matter what you play. If you slot a bunch of stuff it's virtually impossible to not rebuild your board.

The odd-ply and even-ply effect is well known but barely makes any difference in this position.

I would have played 6/5(2) 6/4 but GNU is right that the equity difference between that and some other play is infinitesimal.