I have calculated the following:

If dealer has 16, then he has 1/13 chance of making either 17, 18, 19, 20 or 21 and he has a 8/13 chance of busting. I will call this p(17/16) ie. likelihood of drawing to 17 off a 16 = 7.69%

If dealer has 15, then again he has a 1/13 chance of making any of 17-21 plus he also has a 1/13 chance to draw to 16. At 16 he would then again have a 1/13 chance of getting 17-21 etc.

Thus for stiff hand of 15 I would say probability of getting 17 is 1/13 + 1/13x1/13 i.e. P(17/15) = 7.69% + 0.59% = 8.28%

Following on, likelihood of getting a 17 of a 14 stiff is:

1/13 (i.e. drawing a 3)

1/13 of getting a 2 and thus having a shot at p(17/16)

1/13 of getting a A and thus having a shot at p(17/15)

Thus this is equal to 7.69% + 1/13x7.69% + 1/13x8.28% = 8.92%

Using this approach I calculated P(17/13) to be 9.61% and P(17/12) to be 10.35%.

The numbers are all the same regardless whether we are looking for the dealer to draw to 17, 18, 19, 20 or 21.

This also helped me to determine that likelihood of the dealer to bust off 12, 13, 14, 15 and 16 are 48.27%, 51.96%, 55.39%, 58.58% and 61.54% respectively.

I'd like to know whether my calculations seem correct, or whetehr I have made an error somewhere?