I'm not sure how exactly to determine that, but I used this system to compare real-world results to theoretical LSAC predictions. Obviously there are a few problems, because the real world is more subject to circumstance that theory, but I dunno, might be useful. This is all 2005 data.

Let's take UT for example. Within the 165-169 score range and the 3.25-3.49 GPA, 175 students applied and 51 got in. This means that among this group, about 29% got in.

For someone with a 165 LSAT and a 3.25 GPA, LSAC predicts about a 20-25% acceptance chance.

For someone with a 167 LSAT and a 3.37 GPA, LSAC predicts about a 25-30% acceptance chance.

For someone with a 169 LSAT and a 3.49 GPA, LSAC predicts about a 40-50% acceptance chance.

So the % chance given by LSAC is somewhat similar to the chance for someone around the middle of the number bracket being evaluated, at least in this case.

Let's look at a splitter with a 175-180 LSAT and a 3.0-3.25 GPA. In this bracket at UT, not a whole lot of people applied (6), and only one got in, at likelihood of 16.6%. Obviously, it's difficult to come up with a representative sample for this bracket---splitters are so weird---but we can work with what we've got. By LSAC:

175/3.0 --- 50-60%

177/3.13 --- 65-80%

180/3.25 --- 80-90%

Clearly, at least at UT, LSAC overestimates a splitter's chances. Granted, again, this isn't a representative sample, since splitters that big aren't exactly common and UT has its own idiosyncrasies.

If you like this system (and if you have any arguments against using it I'm all ears, statistics isn't really my thing), this you can toy around with the numbers at different schools. Likely, some schools are going to be more lenient about splitters than others. Or they'll hew more closely to their index formula.

I think LSAC probably looks at index numbers, and when you've got a high index number but one score is a little low, then many schools will be less likely to consider you despite your high index. Making a 3.0 and a 180 might give you roughly the same index as someone who makes a 4.0 and a 170, but I'd put my money on the 4.0.