I'm not sure if this has been covered yet, but the error for tie breakers should be calculated like so:
Predicted outcome between T1 and T2:
P_T1 to P_T2
Actual outcome:
A_T1 to A_T2
Error
|P_T1-A_T1|+|P_T2-P_T2|

That way, in the example the error from the SAS-SAC round would be 1, not 0.