This is the RUSA fleche rule that I was thinking of: "Distance traveled is calculated on the basis of the shortest route between checkpoints that can be legally traveled by bicycle. Maps or mapping software with accurate mileages will be used to determine distances." See http://www.rusa.org/flecherules.html at Art. 8.

The corresponding brevet rule is: "Additional checkpoints should be located at each end of the brevet route, as well as, at any point along the route where a shortcut might be taken." (emphasis in the original).

I'm not sure if a difference is intended, but these does seem to be one as the rules are written and, at least around here, there's a difference in practice, too. The fleche rule is manadatory -- "is calculated" -- and specifies the method of calculating official mileage. The brevet rule is permissive -- "should" -- and includes no such method for distance calculation.

Around here, there are so many good roads that it'd be nearly impossible to comply with the fleche standard on a brevet without either using major roads that no one wants to ride or without having a silly number of controles. For example, between Grove City and Chillicothe, you have your pick of either SR 104 or US 23, both of which go from point-to-point on opposite banks of the Scioto River. Unless you're insane or like extremely busy (but bike legal) 4-lane highways and big trucks, you'd never ride on either. Getting from Grove to Chilli on brevets using the network of farm-to-market roads that got built here 200 years ago, though, is going to take a few miles longer (throwing in Circleville as a controle doesn't eliminate the problem). My understanding of the rules is that in a brevet, you get credit for your actual miles ridden (in the example, 49.7 miles using a route from this year); but a fleche route that used those two controles would credit you for only 45.4 miles (using mapblast.com to make the calculation, which puts you on US 23).

I'm guessing (but I don't know) that the brevet rule gets interpreted literally in the American West, and perhaps other places, where there just aren't many roads because the place is so new (California), unsettled (Kansas and Eastern Colorado), or because terrain severely limits the road opitions (interior New England). Here, our RBA has even joked in pre-ride briefings along the lines of, "go ahead and try a short-cut. I've got the number to call your next of kin." Most randonneurs would never want to ride a state highway if they could take back roads. And unless you really know the terrain, the last thing you want to do in "coal country" is to follow that cut-off into some random hollow. Maybe it goes through cleanly. Maybe it just climbs the monster hill and then turns to gravel at the top. Assuming some randonneur was an unethical rule-breaker, the price to pay in time lost and extra distance ridden if you guessed wrong (it would be a guess, the maps of SE Ohio are notoriously inaccurate about what constitutes a "paved" road) is just too high. Even the cheaters are better off staying on route.

Of course, obvious short-cuts and short-cuts to avoid some particular, planned unpleasantry (I'm thinking of the 1200 feet of climbing in little more than 5 miles in last-year's 400K) are taken care of with controles to eliminate the possibility. But the roughly 10% "error" that's built into most routes because there's always a shorter paved route available is just accepted as a necessary evil.