No, they have a die they *think* rolls high. Remember, you are talking about a species that thinks that their favorite sports team will perform badly if we wear the wrong socks on game day.
It is much more likely to be perception rather than reality - unless they actually rolled a bunch of times, recorded the rolls, and did a statistical analysis, their perception is subject to all the usual human cognitive biases. That die probably just happened to roll well in a couple of dramatic game moments, and got labelled "lucky". The times it rolls a 2 when it doesn't really matter are probably discarded from consideration.
For most dice, I agree. With d20s not so much. A few years ago, I was intrigued by some videos on the internet with people testing their dice by floating them in water (put Epsom Salt in the water if they sink). I found plenty of bias with the water test. (As in,
every die seemed to float a particular way.) When I checked them by rolling the dice a ridiculous number of times, I noticed that only the d20s seemed likely to bear out biases related to their floating results. "Speckled" d20's were the worst, as they would show biases without obvious physical defects. Most of the other dice types (by material) that showed bias had obvious physical defects like some numbers worn out, etc. Some of the speckled d20s could show pretty strong bias that didn't take very long to show.*
I suspect that the other dice sizes just really don't "roll" enough for it to matter much, but that d20s that are tumble-sanded are close enough to round for the biases to matter.
Now, for a player playing a caster, that d20 bias might not matter. But for a fighter-type, it
could have a huge impact long-term. (At least, if you play with that die as often as we play.) However, even then it can depend on exactly how the die is biased. Many of the d20's would have 2-4 results that were more likely to come up. But! Often the results were on opposite ends of the scale. (sure you roll more 18's, but you also roll more 4's...good luck with that.)
* Examples that I recall:
1) one green gem d20 had an obvious molding issue around some of the numbers. It rolled '17' with alarming regularity, but didn't seem to have any other disparities (the other results all seemed equally depressed by all the 17's).
2) a purple speckled d20 rolled '12' about 4 times more often that expected and '4' almost twice as often as expected.
3) a really old yellow d20 with obvious erosion/wear on some of the corners practically refused to roll a '1'.
4) a horrible speckled die with yellow, blue, and white speckles. I got it in the bargain bin, and it must have had a void or something to go with the ugly. You could watch it spin off-center in the air and it just loved rolling 16 or one of the faces adjacent to 16.
5) a rusty-red d12 speckled die. The only non-d20 die I recall with an issue, but its biases were right on the edge of statistical significance.