pull down to refresh

... at the end of December.

As an experiment, we've entirely removed trust from ranking. In short, what this means is that sats zapped/boosted/downzapped to posts/comments are the only, and I mean only, thing that influences content ranking.

If you want more detail, I've implemented what was discussed here. To summarize:

  1. A boost of 100 sats (or any amount) behaves exactly like a zap of 100 sats. There's no difference. A boost is just a self-zap from the OP. All that matters is the number of sats
  2. A zap of 100 sats (or any amount) from any stacker behaves exactly like a zap from any other stacker (even @anon). Or, 100 sats of total zaps from any number of stackers. Or 100 sats boost. All that matters is the number of sats
  3. Boosts no longer have a minimum amount, can be any amount, and top boosts (pinned in territories and homepage) only last 1 week
  4. 50% of rewards go to top zappers of posts and the other 50% go to top zappers of comments
  5. Downzaps are ~3x as powerful for influencing ranking as an equivalent amount sats upzapped or boosted
  6. Posts are outlawed when 0.3*(zap_sats+boost_sats) - downzap_sats <= -1000 sats

As a hazard of being me, I confounded this experiment with a few others:

  1. The ranking algo for hot has changed subtly
  2. Trust, which is only used for rewards now, is a little easier to gain and trust is only based on the last year of zaps.

Our hypothesis is that we were overindexed on trust in ranking and still get pretty good results without it. Having backtested this, my personal suspicion is that that's true but we'd get better results with a tiny bit of trust added back in, some ability for historically great zappers to gain extra ranking influence. We'll see.

How will you tell if the results are "better" or not? Eyeball test? Anything else?

reply
102 sats \ 0 replies \ @k00b OP 14h

I'm mostly curious if they get worse. If things don't change, it's better because there are fewer hidden variables, improving it will be easier, and it's more fair.

My default metric is spending on relevant things. Comments are another which I take as folks finding content that's more engaging. Other than that, it's rough sentiment analysis and an eyeball test.

I'm also open to suggestions.

reply
11 sats \ 1 reply \ @Scoresby 10h

Do sats zapped on a post's comments count toward the ranking of the post itself?

reply
113 sats \ 0 replies \ @k00b OP 10h

Yes. It's been that way for awhile. Currently ranking is like this:

((sats + boost + (comment_sats * 0.25)) * 0.3) - downsats - (comment_downsats * 0.1)

reply
100 sats \ 1 reply \ @jasonb 18h
Downzaps are ~3x as powerful for influencing ranking as an equivalent amount sats upzapped or boosted

Am I interpreting this correctly? The wording is throwing me off. A downzap is three times more powerful than a zap or boost? If so, hell yeah! If not, can somebody help explain this to me?

As someone who’s never studied social media (and normally hates it), the idea of trust is kind of elusive to me in this context. That said, was trust also determined by someone’s downzapping skills? I kind of pride myself in my downzapping discernment. It’s one of my favorite parts of this platform.

reply
100 sats \ 0 replies \ @k00b OP 13h

You read it right.

That said, was trust also determined by someone’s downzapping skills?

Yes! Downzapping was a part of trust was calculated .

reply

Since one of the motivating factors was transparency, I wonder if the time decay factor could be made public as well. As an explainer for why some higher zapped content still shows lower than less zapped content

reply
0 sats \ 1 reply \ @k00b OP 14h

I'm not sure how to communicate it, but it's a pretty straightforward exponential decay with a half-life of 4 hours: 100 sats now counts for 50 sats of ranking 4 hours from now., 25 sats of ranking in 8 hours, and so on.

reply

Come to think of it, probably not worth the ui clutter.

reply

Hmmm I’m thinking top boost makes less sense now that boost is powerful on its own

reply

​My understanding was that one of the things top boost did was turn the top spot of the home screen (or of a specific t​erritory) into a monthly paid ad. I honestly thought this was a good idea. It was cool to see who wanted that space and what they put there.

This is the first time I've ever seen my own post in the very top slot (at least since territories launched):

I could see a case to be made for keeping monthly top boost as a sort of special slot.

reply
102 sats \ 2 replies \ @k00b OP 13h

At some point I started ignoring the top spot because it never changed.

Big boosts can dominate the front page for a day or two now.

I suspect that'll be better all things considered.

reply
163 sats \ 0 replies \ @Signal312 9h

Yep, for me the the top spot just became kind of visual noise, you don't see it anymore, until it changes.

reply

I'm very curious to see how this develops. It will be kinda tempting to mega zap the next big post I'm working on.

I also agree with comments others have made that the home page feels more lively.

reply

Thinking the same!

reply
21 sats \ 2 replies \ @Fenix 15h

What does this mean for a regular user like me? Is there anything to avoid or get involved in? It seems much more legitimate not to make a distinction between the zappers, I hope this test goes well for you and all SN members.

November?

reply
0 sats \ 1 reply \ @k00b OP 14h

It just means that you may have more power over ranking now if you didn't before.

November?

I was planning to ship this in November because I was teasing it in October. I didn't get around to it until now.

reply
21 sats \ 0 replies \ @Fenix 9h

Sounds great.

About November, I remember you saying that. Let’s test.

reply

Wow!! I think I like this change!! I always thought the boost minimum was a bit odd. Now I can self zap anything I want now and downzap anything I don’t want to see

reply
21 sats \ 0 replies \ @sox 16h

The name alone is so powerful that we don't care it's December.

This is gonna be a very interesting change :D

reply

1000 satoshis value is not even 2 $

reply

Means that one would just need to pay a small boost to get out of being outlawed though.

reply

I noticed right away when u boosted this post by only 1000 sats!

I think my trust was extremely low cuz of my many posts, so we’ll see if it helps me!

Conversely, @denlillaapan had high trust, but he’ll still do great cuz he got a lot of zaps anyways

I used to be rly motivated by rewards but ever since the daily donation dropped from 100k to 25j I haven’t thought about it at all!

reply

The first glance: more ~Stacker_Sports in hot, which could imply that between territories the playing field is more leveled out. I think that that's good from a de-algo perspective.

reply

Are we supposed to be able to boost posts like that (Forward)?

reply
22 sats \ 1 reply \ @k00b OP 15h

always has been

reply

News to me! I never even clocked that. haha

reply

Why november? We are already half december :)

reply

I assume it’s a play on No Nut November

reply

I was planning to ship this in November because I was teasing it in October.

It's a self-own. But the real self-own is that no one gets it.

reply
The ranking algo for hot has changed subtly

One of the benefits of this change is that there's no delay between zapping/boosting/downzapping and seeing rank change.

reply

You should add also option for multiple boosts from different stackers. Now this post was boosted with 1000 but about another stacker wantt to boost 10k ?

reply
113 sats \ 1 reply \ @k00b OP 21 Dec

They can zap it. It’s the same as a boost.

reply

Yeah but a post boosted with 1000 sats or boosted with 10k will stay the same time on top. Is not fair. Make them pay.

reply
As an experiment, we've entirely removed trust from ranking. In short, what this means is that sats zapped/boosted/downzapped to posts/comments are the only, and I mean only, thing that influences content ranking.

Not sure how, but maybe trust score could factor in based on active stakers. Like, more stakers means the score carries more weight and vice versa. Just spitballing here, haven't really crunched the numbers.

reply

This new algo looks sick, it's gonna kill all those old-school assumptions.

reply

@Coinsreporter let's see how your cricket posts do now

reply

1000 sats is approximately 1 usd

reply

1 sat is exactly 1 sat

reply

deleted by author

reply

It’s divided like it was before. No change other than it all goes to zaps

reply

deleted by author

reply
11 sats \ 1 reply \ @k00b OP 21 Dec

It works exactly like before. None of that code has changed. Trust is still used in rewards - but nowhere else.

reply

deleted by author