As I said, I ran the stats from all my gold mods:
The theoretical outcome for stats, assuming uniform probabilities:
5 upgrades to a single stat: 0.25⁴ = 0.0039
4 upgrades: 4x0.25³x0.75 = 0.0469
3 upgrades: 6x0.25²x0.75² = 0.2109
2 upgrades: 4x0.25x0.75³ = 0.4219
1 upgrade: 0.75⁴ = 0.3164
I have 153 gold mods (the vast majority from pre-slicing era), with the following stats distribution:
5 upgrades: 1 (0.0016339869)
4 upgrades 27 (0.0441176471)
3 upgrades 140 (0.2287581699)
2 upgrades 248 (0.4052287582)
1 upgrade 196 (0.3202614379)
This is very close to the theoretical outcome. And furthermore, the under representation of 4, 5 and 1 upgrades can even be explained by the fact that I tend to sell mods with bad speed. So if speed was showing but didn't upgrade, that mod might have been sold instead of leveled to full. But that explanation is not even necessary: this sample could easily have been drawn from the theoretical distribution.
However, I am having quite particular bad luck with slicing. It could just be bad luck, or it could be a different system. Hence why I want data.