+1. In 2015, we made the AM14U2 into a 4 wheel omni drive (attempted an H/slide drive but failed until post season). In 2016 we made the (pre-purchased) U2 into a 10 wheel tank like thing, and in 2017 we chopped the front third off the U3 and made it into a 4 wheel relatively wide skid steer. In 2018 we went pretty straight up, using the U3 frame and wheels and belts and U2 reduction gears, but no other mods. In each case, we still spent much less effort with our drivetrain than we did in 2012 and 2013, and were more happy with the drivetrain results. How happy are we with our mods? Our 2018 off-season robot is being built on our 2017 practice chassis.
The big difference for us is confidence. Since 2015, we have spent almost all of our off-season and most of our build season effort on manipulator systems. The only time we spent significant time since then on the drive train was 2016, when the drive train WAS a manipulator when it came to our primary objective of damaging defenses. Even then, without the confidence of working with the same chassis as the previous year, we would likely not have been confident enough to build anything as good as we did.
Additionally, while I have seen many VF chassis with clamp-on gearboxes at competition, the idea of the drive train gears being THAT exposed gives me the heebie jeebies. If you go to the enclosed gearboxes, the costs appeared significantly higher than $400 when I did an analysis a few years ago.