With the help of engineer (and, later, Offy designer) Leo Goossen, Miller designed a double-overhead-cam engine that looked lovely and ran even better. (In a mildly upgraded version of this car and engine, Milton went an astonishing 151 mph on the dry lake at Muroc in 1924.) The engine was lighter than the contemporary Duesenberg and stouter than the fragile Ballot, so after overcoming teething problems, Milton and his Miller 183 spanked the competition. But a AAA ruling prevented him from racing it at Indy, so Durant got behind the wheel and finished twelfth. A few months later, Roscoe Sarles burned to death in the car when he crashed at Kansas City.
Castle was convinced that the Baby Chevrolet was based on the Peugeot grand prix cars of the era. Many years ago, he'd taken measurements from the 1916 Indy winner and plugged them into computer-aided-design software. Castle also believed that the Baby Chevrolet was nearly identical to the next car built by Miller, the so-called T.N.T. This was owned at the time by Castle's friend, Dave Hedrick, another hard-core Miller-phile, out in Oregon. So Castle would print out full-size drawings of frame rails and axles and other components and mail them to Hedrick, and Hedrick would report back about what changes needed to be made.
"You should see the reams of correspondence between us," Hedrick says. "Some people just want their car to drive. But Bill wanted it to be historically accurate. Even I was surprised by his fanatical level of detail. I never would have gone to the trouble of making the front brakes work in the original manner. In fact, I probably wouldn't have duplicated them at all. You could fill an entire book with what he did with those front brakes."
Castle CAD'ed several brake components, made his own patterns, and had them cast at a local foundry. Then he figured out a way to make the brake pedal actuate all four brakes (via cables) while the handbrake outside the cockpit applied only the rears. The front spindle, meanwhile, confounded him so much that he built a wood model before he started cutting metal. "Normally," he says, "you have at least two degrees of kingpin inclination, but I kept having trouble getting the kingpin in that spindle. Finally, I pulled some pieces from the Speedway [restoration shop] -- they happened to be Duesenberg -- and lo and behold, they had zero kingpin inclination. Bright lights went on. With zero kingpin inclination, I could get the thing together."