Average dps is more important than potency. Crits can cause anomalous parses but in an average they won't. A random event ceases to be random over and infinite number of trials. Thus if someone has a 35% crit rate, you can guarantee on average 35 crits out of every 100 attacks. You can factor that into your damage. What kills a target is damage, not potency. Take for example if two players were playing the same class with the same ilvl but had focused their stats differently you could possibly see a sizeable gap in dps. If this was true of everyone across an entire raid you could be missing over 1000dps or more. "But we're doing the same potency per second" wouldn't cut it on an enrage.

Another example to chew on, buffs don't always align perfectly, perhaps it's worth delaying a skill for buff cooldowns. If you don't delay the skill you may get higher potency per second, but lower dps.

To sum it up crits arn't random with proper testing, dps is what kills targets, potency per second can be misleading, and the best measure of a properly built and played character is the one with the maximum average dps.