For the time being, we've removed personalized hot ranking. In general, we don't have enough people zapping, and especially people zapping consciously/honestly, such that it makes that big of difference in practice. (Also, the way we were doing would've complicated some of the enhancements I made in this release, and the enhancements I made will give way to more efficient personalized ranking.)
Speaking of enhancements:
- Each territory has its own trust graph now. In effect, this means that even if a post doesn't rank high on the homepage, it can still rank high in a territory.
- e.g. if someone creates a ~bip300 territory it can have a life of its own even if the majority of stackers never want to participate in the territory
- Each stacker has an independent trust score for each territory they zap in.
- You might zap well in ~Stacker_Sports according to consensus, but zap garbage posts in ~bitcoin, and this allows for such granularity in scoring
- Similarly, post zapping and comment zapping each have independent trust scores.
- You might zap great posts in ~bitcoin, but horrible comments, or vice versa, and this allows for such granularity in scoring
- Comment zapping by trusted comment zappers, now slightly boosts the ranking of the comment's ancestors
In summary, for each territory, a stacker zapping in that territory gets four trust scores:
- post zapping trust according to global consensus
- comment zapping trust according to global consensus
- post zapping trust according to the consensus of zappers in the territory
- comment zapping trust according to the consensus of zappers in the territory
This will probably need refining after some live experiments, but this is the direction we're headed in.
The only other meaningful change here afaik was fixing reply costs for the automated monday ~memes post which led to the 10 sat reply cost set by the founder being ignored (due to a case sensitivity issue ~memes vs ~Memes)
View all commits here: https://github.com/stackernews/stacker.news/commits/master/
Since starting the Open Source Contributor Award Rules (OSCARs), we've paid a total of 20.5m sats to 41 contributors for 186 contributions! View a nearly up to date list of contribution awards here: https://github.com/stackernews/stacker.news/blob/master/awards.csv
Previous release: #904497
Man I was cooking the DB's CPU for the last 4 hours with an index typo in this release.
CREATE INDEX IF NOT EXISTS some_idx ON some_table(some_value DESC NULLS LAST)is NOT equivalent toCREATE INDEX IF NOT EXISTS some_idx ON some_table(some_value DESC).The site was feeling so slow for a
second4-hour there.👀
I wonder if you hate working with SQL as much as I do
Relative to the alternatives, I love it.
I wrote a relational database from scratch in college for one of my classes (its the only assignment we were given). I get them. They're like a crazy best friend - crazy on the outside, deeply sane on the inside, irreplaceable.
There's a half century of research invested into them. Normal data is so so so so underestimated. (At least in applications like this.)
Fair. I don't even really know what the alternatives are.
My background doesn't deal much with live data, so I'm mostly happy using
pandasinpython, but no one uses that for a production database.I'm sure someone uses a pandas franken-db in prod. I read a book about using ElasticSearch as a production DB.
I think my beef with SQL is that wherever I've seen it used, complex queries are often written out entirely in a single chunk of code. Not sure if the language doesn't lend itself well to modularity, but these chunky queries always seemed hard to decipher and thus error prone.
That's a great point. I took the lack of composability for granted. I'm sure there are awesome alternatives aimed at fixing that.
Why change the language if you can just use a query builder? I really liked Knex at my previous job.
I'm thinking of writing a post on the math behind the Web of Trust. I think there are enough stackers here who would appreciate something like that.
I suspect they would.
Time to parse the files 😵💫
Oh if you're going to cover our WoT, I'm looking forward to a review of my math! I have a feeling my statistical work is pretty sloppy/ignorant.
I'll help look it over also
Will do!
Me, me, me. I want that.
Awesome. Love it. But are you trying to imply that I zap garbage posts in Bitcoin? Haha
You zap garbage posts in AskSN, which is essentially all of them.
Bold of you to insinuate my posts are rubbish
Never!
Okay, I'm coming clean here. I am baffled as to what the hell this all means. Can someone dumb it down for me? Will I notice anything as a user? Will I notice anything as a territory owner? Should territory owners be doing something regarding these changes? Will it end world hunger?
It's pretty subtle. Aspiration wise, ranking should be more representative of what everyone wants to see.
Your territories should have ranking more representative of what you and people in your territory want to see.
No. Normal engagement and zapping should be enough. In your territories, you just have a little more tone-setting influence through your zaps.
Yes, once everyone is on stacker news earning sats.
Is about my old proposal, remember? #844649
It means we have to post more on your territory Bitcoin Beginners.
Yes, I remember that. I think @k00b agreed also. It was just a matter of straightening out the wallets first. Maybe I didn't understand the latest SN update. Does this curb stackers who just gain trust in one territory from gaining general trust sitewide?
Im confused what is considered a garbage post? For example I zap a lot of posts in photos that often get little to no zaps.
So now is my WoT score nerf’ed because I zap non popular content on SN?
Nope, this should improve your trust scores because you have different trust scores in each territory now.
Lots of us zap stuff in the territories other people mute. As the result of this change, that no longer affects your trust score in the more popular territories, e.g. ~bitcoin.
I'm really interested to see what this does.
Territories already look happier. Content is always a bit of shit show on the weekends so hard to tell what the effect on the front page will be.
I feel seen.
(Seriously, though, separate trust scores for territories is fantastic.)
How does this work with homepage? Does homepage has its own trust scores?
Everyone has a "homepage" trust score for each territory they zap in - in addition to a territory specific score.
We don't add user-trusts-territory scores yet. That'll be what personalization brings.
Nice!
Probably related to the personalization topic, but I think a big unexplored space in the homepage specifically is making it work well for people that are outside of bitcoin bubble. I see that as one of the biggest opportunities.
If you make the homepage work well for a person X that has following properties when they open SN for the first time, then that would be amazing!
Person X hates hearing about bitcoin, they don't want to see any news about bitcoin, they don't care about any of the technical details behind bitcoin. On the other hand person X is interested in multiple topics (e.g. privacy, open source development, game development, startups, new technology, new scientific breakthroughs...) and this person wouldn't mind getting sats as tokens of appreciation and zapping sats to others when they interact on these topics. They hate bitcoin, they don't have strong opinion about sats. "Oh these sats can be used on other websites too? That's cool I guess"...
I think we know where this would go, so you would also need to add a toggle to start showing bitcoin info for person X at some point 😉
(just couple of my thoughts, feel free to ignore...)
https://stacker.news/~bitcoin/top/stackers/forever
Wouldn't it make more sense if this filter showed the data based on trust scores by territory?
It would. I meant to do that but this already took longer than I wanted. When I give territories their own rewards pools, I'll add this.
So, these territory trust scores, are they from before, or did they just start fresh?
They are based on all history from the beginning of SN.
This is great, look forward to seeing how it will affect the quality of posts i the home page and on each territory! Thank you
More awesome work.
I look forward to these posts the most.
Wow, 41 contributors, that’s a lot!
This level of trust isn’t something we can see, right? It’s assigned by the platform, which gives each user a score. How can I avoid garbage posts? I kind of know what is it and what I don’t want to see, but what exactly characterizes that? Sometimes I interact and contribute by commenting and zapping posts I find interesting, even if they have few sats.
How is the trust score calculated for territories that were archived and then brought back? Do they take into account that the topics might be completely different?
Could we get a little crown icon for being a big baller zapper?