Getting useable numbers from the HP (torque) estimates in DSMLink is all about running controlled tests. A flat, smooth stretch of road with a pull off at one end is what I use. In my case, there's a “round about” at the other end, which makes it even better. I start a log, cruise a few seconds to the flat stretch, do a solid 3rd gear pull, turn around at the round about and do a 3rd gear pull in the other direction in the same stretch. Then pull off at the pull off area (happens to be a park-n-ride) and review the data. Pop the hood, let the car cool down significantly, make a few changes, and go retest.

Whether you get accurate numbers that you can compare directly to dyno figures isn't all that important. As long as you're getting repeatable numbers, you can do reasonable before/after type comparisons. Although, in my particular case, I do see almost exactly the same values on the dyno as I see with the estimates on this flat stretch of road.