Cred Improvement Proposal: Proper Props

I’d like to offer up a fairly uncontroversial proposal for improving our Cred scores: let’s start properly accounting for the Props channel.

At present, when someone makes a post in #props, it’s treated like any other message, and flows its Cred evenly between the author and any mentioned users. This isn’t really appropriate for Props, since most of the Cred should flow to the person receiving the Props.

I’ve written a pull request which allows us to add special logic for props channels which will make them flow Cred mostly to the people mentioned in the message. As a default, I suggest giving the props edge a 19x weight compared to 1x weight for authorship. That means if a single person is getting props, the author will get 5% of the Cred. (If 2 people, author gets about 2.5%, and so forth.)

I propose that we turn on this rule for SourceCred’s own props channel. At the bottom of this post, I’ve included a table that shows how the Cred changes as a result.

I think this is an uncontroversial change, but I’d like us to get in the habit of discussing changes to Cred on the forums before making them official, so please consider this change and share your thoughts and objections. :slight_smile:

Name Before After Change %
decentralion 24756 24102 −3%
wchargin 11403 11416 0%
beanow 5692 5649 −1%
s-ben 5101 5015 −2%
lbstrobbe 4856 4841 0%
hammad 3642 3702 2%
topocount 1790 1799 1%
burrrata 1735 1737 0%
KuraFire 1649 1731 5%
bex 1586 1616 2%
sourcecred 1396 1396 0%
mzargham 1261 1280 2%
sandpiper 1113 1262 13%
dependabot 1255 1257 0%
panchomiguel 904 970 7%
vsoch 910 913 0%
brianlitwin 889 895 1%
evan 642 703 10%
greenkeeper 667 666 0%
joiecousins 552 630 14%
yalor 481 561 16%
Thena 426 466 9%
youngkidwarrior 463 437 −6%
benoxmo 443 432 −2%
flyingzumwalt 362 398 10%
amico 259 273 5%
rpp 239 251 5%
doctorrobinson 290 233 −19%
github-actions 223 223 0%
owocki 210 208 −1%
deltafreq 205 204 −1%
ian 129 199 54%
cortanav 105 148 41%
protocol 140 140 0%
Louisgrx 139 139 0%
samkuhlmann 132 132 0%
sriche 129 128 −1%
peth 166 125 −25%
coopahtroopa 125 123 −1%
JavierCanovas 109 121 11%

(Here is the hacky observable notebook with data analysis, mostly included b.c. someone else proposing a Cred change might find it useful to fork off from.)

7 Likes

I’m in favor of this, a 19x / 1x weight split for the Props channel feels very fair and sensible to me; having the person who does the observation of giving props get 5% of the Cred encourages giving props to others more (great!) without making it a possible gaming vector (without it being super obvious).

Just so I’m clear, this is not just a 19/1 split for Cred weighting in Props messages themselves, but the 19x weight is also 19x compared to, say, a regular message of that person in a different channel, right?

Scenario: person A gives props in #props to person B. That message gets 1 emoji reaction. Does that mint 19 Cred for person B and 1 Cred for person A?

Scenario (continued): Now person B posts a message in a different, non-special channel. That message gets the same 1 emoji reaction, and it mints 1 Cred for person B..

(the scenario above is not correct, see below)

Is that correct? Is it more subtle than that? Am I getting the whole thing wrong? :slight_smile:

1 Like

This proposal only changes how Cred flows from the message, but doesn’t at all change how much Cred is minted. So in both scenarios, 1 Cred is minted at the reaction, and then flows to the message. If the message is in a regular channel, 1/2 of the Cred will flow to the author, and 1/2 of the Cred will flow to the mentioned person, so they both get around 0.5 Cred. With my update, 1/20 of that Cred will flow to the author, and 19/20 of the Cred will flow to the mentioned user, so the author gets about 0.05 Cred and the mentioned person gets about 0.95 Cred.

Once we have this change in, I would support increasing the channel weight on the props channel to (say) 5x. If we did that, then the 1 emoji would mint 5 Cred on the message, of which 0.25 would flow to the props author, and 4.75 would flow to the person getting the props.

2 Likes

Perfect, thanks DL for clarifying!

An observation: the before column of Cred scores you used does not match the current explorer table of our scores.

As a props whore, glad to see I’m only down 2% :sweat:

Change makes sense, and numbers look good. Was expecting more volatility actually.

2 Likes

I’ve been supporting this improvement for a while. I think this change makes a lot of sense: giving most of the Cred flowed to props to the person receiving the props. I’m in agreement with the original post and Faruk that 5% to the original poster makes intuitive sense.

It’s nice to see the percentage differences also.

I’m all for this.

Since everyone voiced in support of the change, we’ve now made it official: https://github.com/sourcecred/cred/pull/162

4 Likes