As John points out, a good place to start is another reference position you should know. X amount of men off with one in the air against a closed board with perfect distribution. As he said, you win in the ballpark of 6% and since we are far from the closeout a better guess is 8%.
                     |
| ||||
| XGID=aBBBCCB--A-------------ff-:0:0:1:12:0:0:0:1:10 | |||||
| Blue to play 12 | |||||
| 1. | Rollout1 | 9/6 | eq: +0.873 | |||
| ||||||
| 1 1296 Games rolled with Variance Reduction. Dice Seed: 28150473 Moves: 3-ply, cube decisions: XG Roller | ||||||
eXtreme Gammon Version: 2.10, MET: Kazaross XG2
However, I don't agree with: "I think you're more likely to anchor than not so I think the gain outweighs the loss. Plus some opponents will hit with all eleven 3s from the bar, which I think would be a mistake."
Since we are forced to hit with [31 32 33] (as John originally stated) and the other threes, [43 53 63] can be used to enter and double hit so our chances of anchoring are greatly diminished, I would play the non fancy play of 23/22* 3/1 2/1. It's still good I think to hit him off the front of the prime since the chance of your opponent now picking up a second checker is virtually zero. Knock him off it and hope that somehow messes up him rolling the prime forward.
Perhaps another reference position worth taking a look at is something like what is posted below. This will give us a rough idea of how much we stand to gain if we were to anchor. I chose the 23pt because I think it's more likely we anchor there but the 22pt or 24pt are equally worth knowing but extrapolation should suffice.
            |
| ||||
| XGID=a-aABBBBBB---A----A----ib-:0:0:-1:62:0:0:0:1:10 | |||||
| Blue to play 62 | |||||
| 1. | Rollout1 | Bar/23 | eq: -0.707 | |||
| ||||||
| 1 1296 Games rolled with Variance Reduction. Dice Seed: 28150473 Moves: 3-ply, cube decisions: XG Roller | ||||||
eXtreme Gammon Version: 2.10, MET: Kazaross XG2