0.375 is a good scale, I wasn't too far off with my "Divide by 2.5" then

.
Thank you again Leusch and Mosquill, this allows me to get some good data on the vehicles with relatively minor hassle.
EDIT: Another weird thing I noticed. Whereas road and water speed are - as expected - always the same, the offroad speed varies greatly from test to test. Test setup is: A terrain ramp of 25? (grass material), vehicle starts from zero at the bottom and climbs up. Speed is recorded after half the ramp is passed (speed usually is level by then). The thing is, after I load the vehicle and hop in, the first run gives insane speeds. This is most apparent for tracked vehicles. E.g. the MTLB didn't lose ANY speed when climbinb (160 road, 150/160 hill), the BMP dropped from 180 (road) to 150(hill).
After some more driving around, stopping, reversing etc another test run shows dramatically changed results: The BMP only reaches 20-30 uphill, the MTLB dropped to 100.
Weird. What causes this?
EDIT2: Further testing of the hillclimb-speeds shows this:
First run: Speed uphill is almost unchanged from level speed, the vehicle loses 10/20 speed max. MTLB Road 160 - Hill 160.
Second run: MTLB Uphill 120
Third run: MTLB Uphill 100
Fourth run: MTLB Uphill 90
Fifth run: MTLB Uphill 70
Sixth run (waited 3min with idling engine in editor): MTLB Uphill 50/60
Stable after that.
This is pretty weird. Along with losing speed, the vehicle gets far more "slippery" and the grip worsens. On the sixth run the vehicle no longer goes straight up but starts to slip sideways, forcing me to make corrections by steering. This would point to the ground material behaviour but I absolutely don't understand why the friction should change during playtesting in the editor...