by rburgh » Tue Sep 27, 2011 11:27 pm
There are two different issues. Some cards have simply SI as a play result. I have always assumed these were equivalent to SI(cf) results. An example is Damaso Garcia 1985 (as noted in Petrosian's post).
Other cards have the play result SI-. One that I know of is the play result for 2-3 on both sides of the card for Jo-Jo Moore 1934. I strongly suspect that these are really SI* 1, lineout 2-20 results, but they are counted on DD as though they are full hits.
Strat stores split results in four (!) locations. The first result is stored in one array, there is another array that lists all the locations on the card with split results, and two more bytes are dedicated to the split number and the back-end result. I suppose this is actually more efficient than using two bytes for each possible roll.