Let's say an OMCG gives you a 100% chance of success (or, like, close to it) if it comes with an M18 and a 0% chance of success if it comes with no M18. You need to blow up the immobilized KT that is protecting the VP you need to cap to win, and an AT gun will never get there in time or penetrate the front armor of the KT fast enough to kill it. Aside from the KT your opponent doesn't have much AT, so your M18 will be safe. You've at Tier 3 and there's no time to build a tank depot or anything and you can't get Rangers close because they have MGs and snipers and stuff.
In that situation, if the random number generator gives you an M18, you win, otherwise you lose. You could try something crazier, like spending the manpower on Rangers and hoping you can fire up past the MGs and snipers to throw some nades and get some bazooka shots, but that's much less likely to work. So the best choice is calling an OMCG. But if the game doesn't give you an M18, you lose right there.
You could say "but you should just play better so that you never get in a position where you have to rely on an OMCG to win the game." But the OMCG is not the only thing in CoH that is based on luck. Vehicle criticals, artillery deviation (especially off-map shells), whether your infantry do the cover dance tango or sit there and shoot at the enemy, hits/misses for shots fired from AT guns and tank guns and other heavy guns, etc. all rely on the random number generator, and if you think the outcome of a match never turns on aggregate lucky breaks for one player or even on one lucky break (when an offmap artillery shell decides to land straight on a squad and wipe it out for instance) then you're ignoring a big part of CoH.
Now obviously the game isn't all luck, but sometimes important aspects of the game do come down to luck, and that can be the difference between one player winning or losing if they are at about equal skill.
If your only chance to win is getting an M18 then odds and rolls are irrelevant, because you have no alternatives.
If you believe a ranger has a chance to do the job, then it comes down to whether that is more probable than drawing an OMCG with an M18.
And yes, in fact i do say "but you should just play better so that you never get in a position where you have to rely on an OMCG to win the game."
There are thousands of events in a single game and letting its outcome depend on any one of them is a mistake to be avoided, and that is exactly what the small handful of people who repeatedly turn up in finals and semifinals are doing.
You cannot look at a single event without taking into account every event, decision and account that lead up to that event.