💯 to needing a better metric regarding npubs.
Maybe PageRank. Except … that won’t work. PageRank is an extrinsic metric. Follower count is another extrinsic metric. I want an intrinsic metric.
Example of an extrinsic metric: fame. It’s a relationship between the user and the community.
Examples of intrinsic metrics: height, weight, intelligence. Those things are independent of the community.
PageRank is an extrinsic metric, kinda like fame. We need to modify it so it’s an intrinsic metric, more like intelligence.
GrapeRank is a modification of PageRank that converts it into an intrinsic metric.
Suppose I want to know the height of user1. Users 2, 3, and 4 estimate 5’11, 6’1, and 6’0, and communicate those results to me. What do I do with those numbers?
One idea is to add them together. User1 is … 18 ‘ tall. Obviously that’s the wrong metric. This is the PageRank approach.
The right metric: weighted average, which comes to 6 feet, give or take, depending on which of users 2,3,4 I trust the most.
GrapeRank is a weighted average. Simple as that.
GrapeRank is a nonlinear equation, as opposed to PageRank, which is linear. But who cares? Use the easy algo and get the useless metric (18 feet), or the slightly harder algo and get a useful answer that estimates objective reality (~ 6 feet).
Trynta write this out in long form today. 🤓
This reminds me of the discussions about building a WoT algorithm, because who's to say which one is correct? Everyone has different opinions and thoughts.
“Who’s to say” — always hard to know where to go with that question. Who’s to say that coke is better than Pepsi, or bitcoin is better than shitcoin, or meat is healthier than seed oils, or that the earth is round and not flat?
Some of these statements communicate subjective taste. Some are statements of fact.
Some are statements of utility. Hammers are more useful than feathers for hammering a nail. But who’s to say? That’s just, like, your opinion, man! 😂
Some WoT algos are useful. Some are not. So do you think it is reasonable or worthwhile for me (or anyone) to promote algo1 as being more useful than algo2?
Is it a waste of time or a good use of time for us to ask the question: is this algo useful? Could there be a better one? https://i.nostr.build/EXhwoHEw0yQvGTRi.jpg
It's almost as if we need several WoT or ranking metrics to choose from because what's useful to one person may not be to another. For example what if I feel zaps are the signal and you feel that comments are a better metric? We all have our own rankings.
YES THIS 🎯
This is what I mean when I talk about “interpretation.” You gather follows, zaps, likes, number of responses to someone’s posts, whatever data you think is relevant. You interpret it meaning you translate each category of data into a “rating” in a common format selected by you bc you find it useful. I propose the rating format needs, at the least, a score. Even better to have a confidence and a context. I decide I’ll “interpret” a follow as a score of 1 (probably not a bot) and a mute as a score of 0 (probably a bot). A zap I also interpret as a score of 1. But maybe you don’t trust zaps bc they’re not verifiable (unless it’s ecash) so you don’t use zaps in your calculations. In any case, you now have a ton of “ratings” that you can use to calculate a WEIGHTED AVERAGE which feeds into the GrapeRank score which will ultimately be more useful than follow counts or PageRank in the same sense that hammers are more useful than feathers at banging nails into wood. (Well, I guess that’s just my opinion … utility is subjective and arbitrary … who’s to say?)
More useful bc you can use the above score as the weights to calculate the next round of scores, like: which relays, mints, etc are the most trustworthy? Actual pain points.
Oddbean new post about login | logout 