This is an analogy based on the "street race scene" from American culture that peaked in a span about from the 1950's through the 1970's. It refers to using the actual results of a race to judge the "better engine" instead of merely talking about the alleged elegance and/or "nice theory" of your engine design. It's a variation of "show instead of tell".
Of course, it might be the case that the Car and the Engine are very good... but the the driver is not...
MythBusters? often build a robot or wire-track to control repeated experiments to remove the variability of humans. As far as the "street", the builder either drives their own car or get a buddy to do it. If they F it up, it's their own problem: it's their job to package and present their evidence properly. Ideally it would be nice to try multiple drivers to remove driver-specific variability, but that's not the way they did it back then. It was essentially a mating dance, not a scientific experiment. Smarter drivers practice with a buddy first, unless that means getting busted. One can also swap drivers and run the race under both or multiple combos of drivers.
Interesting. Personally I have found something of the opposite. Most race results I have found completely lacking in understanding what engines are better. Once the engine goes into the car, there are all sorts of things that take away the "objective science" part of evaluation of the engine design, at least. Things like transmission, drive train design, wheels, the driver, suspension (if a race course), etc. I would like to just see torque and horsepower measurments measured right at the engine crankshaft. In fact, I find the whole thing so strange and curious that I can only explain it by concluding that GM has strong-armed the whole market into never questioning the standard piston engine and non-standard engine designs simply can't get a foothold. (See also ContinuousFuelCombustion.)
A "scope issue" may be whether the performance of the car in total was the issue at hand or the engine. Ideally you do both "bench" performance testing and in-car road testing, trying to keep all the non-engine variables the same. If bench testing and road testing give different results, then you've got some digging to do. In practice, the parties involved may not want to test so thoroughly. If thorough tests are not possible or available, I'll take an actual road race over talk.
Misread as RaceTheDamnedCat and some confusion.
Lisp programmer by chance?
See Also: IfFooIsSoGreatHowComeYouAreNotRichCategoryEvidence