There's a few reasons.
For one DM, as a person, is apt to be predictable. If you fudge frequently, those patterns will become apparent, and the players will learn to use them, just as they can learn to use a biased die that more frequently rolls high. If you fudge infrequently, the sample size the players have is too small for them to effectively predict you. If you fudge infrequently and hidden, the players won't be able to sort out the "signal" of the fudge from the noise.
For all that I may say I can identify occasional problem points that might call for a fudge, that doesn't mean that I think I know best for every single (or even most single) rolls, or that the game is best played with the DM determining many things directly. I generally like the way the dice play things out. Fudging is for me more of a "in case of emergency, break glass" sort of thing.
There are probably other reasons that aren't popping to mind at the moment.
Okay, that makes sense, but in that case I'd suggest that instead of saying things like, "I almost never do this," say, "I don't do this to push things toward a larger outcome I've predetermined."
Yes. I think I did use words to that effect upthread. Sometimes, you need to use multiple statements to cover one base.
(I almost left the "larger" out of that sentence, but that wouldn't work. Because folks do fudge to push things toward an outcome, right? Even if that outcome is, for example, a fight that doesn't end prematurely due to luck that's built into the system we play.)
As I have noted earlier, it is my observation (anecdotal, I know) that the use is generally not towards something, but
away from something. If I drive play towards a point, that highly restricts player and random impact on how things unfolds. If I drive things away from one point, all the other points are still available, so randomness and player initiative can still strongly impact play.
That's kind of a non-answer, isn't it? Why don't you want the players to know what's happening behind the curtain? Would it make the game less fun for them?
No, it is an answer to provoke thought, with a suggestion of the direction in which thinking might be useful.
You don't know why I don't hand over all the scenario information. But I'll hazard a guess that you don't hand it over either. Why not? We probably share several reasons. Some of those probably apply.
Someone upthread used the example of the BBEG rolling either three naturals 1 or three crits to open a combat.
It's extremely important to understand that not only is this built into a system with a random mechanic, it's not even particularly rare in a system in which that mechanic is linear. And not only is it built in, and not only is it fairly common, but any DM with experience knows all this.
Dude, you're talking to members of a species that knowingly builds homes on floodplains and on fault lines and is still stunned when there's a disaster. Risk assessment isn't our strongest suit.
But, to be honest, I think the situation is rather more rare than you suggest. It is, of course, dependent on the details of the situation. But we can do some envelope-back estimates for sake of demonstration.
The chance of rolling three 20s in a row is 1 in 8000.
The chance of rolling three 3e critical threats in a row with a longsword is more like 1 in 1000.
In 3e, the chance of those being confirmed criticals is no greater than that, but can drop quickly - if the beastie hits on an 11 or better, the chances drop to 1 in 64000 (or 1 in 8000 for the longsword).
So, as a very rough characterization - we are talking about an event that we might see once or twice if we run 1000 combats.
Take your typical 3e campaign. 20 levels. 13.3 encounters per level. That's like 266 combats. So, let us be generous - we'll see that event once or twice in a campaign? And not every time will I consider that event an issue.
So, now we compare the work to the risk - I can do careful checking on every design to avoid these. I can change my system (which I otherwise like a lot) to avoid these. Or, I can just accept that every once in a while I'll have to consider handling an exception on the spot.
Handling the exception takes me seconds. The extra encounter design work over the course of the campaign probably adds up to hours. The system design changes (with double-checking that I'm not introducing some other unintended consequence) is also in the hours-or-more category.
Seems to me the choice is clear. It doesn't pay to do the work ahead of time to specifically avoid low-probability events. Looking back to the thread title ("Fudging for fun and profit") I think you've helped me identify another reason to consider fudging - Profit! Fudging may be more cost-effective than some of the alternatives to reaching similar ends.