Your play doesn't look that bad, but a rollout confirms that it is around a 0.05 error at DMP.
It does seem pretty clear to me that if you don't roll exactly 66 next roll, you are better off with the computer's play. I see three factors that make it better in that case:
- You don't like the stripped 4 point after your play
- If you clear the 7 safely next turn, you prefer having 3 checkers on your 6 as opposed to 4 checkers, as you'll want to strip that point next turn and will be able to do that more often while keeping good spares on your 4 and 5 points. Also if you roll a number like 6-1, you want to end up with 4 checkers on your 6 instead of 5.
- You don't want an extra checker on the 2 point since it's more flexible if you keep it above the Blue anchor.
Aside from the blotting 66, I don't think "odd on the top 2 points" matters here, because a sequence like 65, 66 plays OK after the computer's play.
So weighed against the blotting 66, I don't think it's surprising that these arguments add up to enough (I just wouldn't think 0.05). I mean these arguments get 35 to 1 odds basically as the non-66s occur 35 times more often as 66.