http://wikiwiki.jp/splatoon2ch/?%A5%AE%A5%A2%A5%D1%A5%EF%A1%BC%B8%A1%BE%DA
This seems a lot more convincing than the squidboards data, but I'm not sure if any data were fudged. The problem I have with the squidboards conclusion is that they make the claim that some sub abilities have diminishing returns while some dont. The thing is, once people had a greater number of rare gear necessary to test 3 mains + 5 subs etc, there were no "hard caps" in sight, plus the values were much lower than expected according to the diminishing returns rules they set (for example, every main or sub after the first is worth 1/2). It is more likely that due to the uncertainty in the measurement and the fact that some abilities have a more delayed diminishing returns effect (for some abilites 0-2 mains or less is approximate to a linear distribution), they assumed that each ability has different rules depending on the number of mains and subs equiped. The japanese wiki makes the claim that Mains are worth 10 and Subs are worth 3, and it is the total power of all the mains and subs that is applied to a asymptotic curve. The diminishing returns effect depends on the steepness of this curve, all abilities have diminishing returns, and more compelling, they can be defined by a simple, single variable function.
The japanese wiki illustrates basic high school science. You run multiple trials and plot the damn thing and fit a curve.