It reduces by a limited amount (I think a max of 10%) based on how far the player's weight is from the position mean (which is shown on the player card)

I'm going to reiterate this question. The players listed before are 36 and 37 percent below mean respectively, but rate in the mid-60 potential (my weights).

However, this player who is 43 percent over mean for CBs, but has 80+ in every attribute I value in a CB (SP, M2M, and Punish) only rates a 53 potential: http://mfn19.myfootballnow.com/player/4500

If anything, I would assume the second player to be much higher than the first two in rating because of how weight works in terms of strength. A 200 pound WR playing G is going to get blown over. A 270 pound CB is eventually going to get beaten in a footrace, but will stay with his man for at least a short time.

Does this mean that each point over/under mean weight subtracts 2 overall points? Because I could see all three of these players being about equal in terms of weighted attributes, but Rivera is 12 overall points lower than the two WRs.

I'm not looking directly at it but I feel like it's 10% per standard deviation from the mean. Remember that your heavy CB might be able to stay with the WR for the first 5 yards but a light, fast WR will easily outrun him on any route where it covers much distance. That's why in the NFL it is so hard to cover a strong, fast TE - because you need a LB-size player to deal with him but he doesn't have the speed of a smaller CB to keep up with him.

I'm not denying this. I'm just pointing out a 200 pound guard against a 300 pound DT is toast in a pushing match.