Wow that seems like an extremely small difference, considering 38.1 weight would be like black armor and stuff. Like.
Gafferjack is seeing a 1.7% difference in time per attack between being naked and wearing close to maximum possible armor weight. He appears to have run the test with a very fast weapon and with a relatively low weapon proficiency, both of which minimize the absolute difference in time per attack one would expect to see with different armor weights. The absolute difference in time per attack would be greater for a slower weapon, and the absolute difference and relative difference in time per attack would be greater for a character with higher weapon proficiency.
In fact, his results are entirely consistent with what my equations for time per attack verses proficiency would predict for a character with 98 two handed weapon proficiency wielding a ~102 speed weapon naked compared to wearing an armet, milanese plate, cased greaves, and heavy gauntlets. Gafferjack, please let me know if I am incorrect on any of that.
The following plot had been updated with the additional data points for weapon proficiency and armor weight Gafferjack has provided.
The stair-step patterns are likely the result of rounding. A small improvement in the data fitting can be made using these updated equations:
armor weight modified proficiency = base proficiency * (1 - 0.008 * effective armor weight)
and
effective armor weight = 3*head armor weight + body armor weight + leg armor weight + 2*hand armor weight - 7.5