david on Nostr: I start out calculating a weighted average, using follows and mutes. A follow is ...
I start out calculating a weighted average, using follows and mutes. A follow is interpreted as a “vote” for a score or 1, and a mute is a “vote” for a score of zero.
It’s a weighted average, and the weight equals the influence score of the follower (or the one who mutes).
So most average scores are going to be 1.
Next, I add up the sum of weights of all the “votes”. Kinda like you go to Amazon and see an average score, but you also look at the number of raters. This sum of weights, which I call the “input” or “total input”, is translated into a number between 0 and 100 percent which I call the “certainty” or “confidence”. Then the influence score equals the average score multiplied by the confidence in that average score.
The confidence score maxes out at 100 percent, which is why everyone on that page has practically the same influence score of 1, doesn’t matter if their WoT score is 20 or 50 or 100 or whatever.
It’s a weighted average, and the weight equals the influence score of the follower (or the one who mutes).
So most average scores are going to be 1.
Next, I add up the sum of weights of all the “votes”. Kinda like you go to Amazon and see an average score, but you also look at the number of raters. This sum of weights, which I call the “input” or “total input”, is translated into a number between 0 and 100 percent which I call the “certainty” or “confidence”. Then the influence score equals the average score multiplied by the confidence in that average score.
The confidence score maxes out at 100 percent, which is why everyone on that page has practically the same influence score of 1, doesn’t matter if their WoT score is 20 or 50 or 100 or whatever.