Ask HN: Has GitHub stars become a gamed metric? | Hacker News Reader

Ask HN: Has GitHub stars become a gamed metric?

16 pointsdiceduckmonk19 comments

It seems that FOSS and “open core” has become a hard requirement to gain street cred, especially for dev tools.

Businesses and VC-backed projects are doing it, but they need to monetize, somehow. For established businesses, they have lots of resources and sharing a wide range of tools that are orthogonal to their core business, and low risk. For startups, they go with the open core model for their flagship product.

I’ve noticed some VC backed startups having thousands of GitHub stars but very little online presence, sometimes with zero references on HackerNews. Also, merge requests go abandoned or closed when you dig deeper. Has fraud detection proof bots for GitHub become more prevalent and accessible?

View on HN

I do not trust star count on Github at all, and I see no reason why I would? There are much better ways of seeing quality, you can look at the code, you can judge from what you hear at conferences and read online. When I see some org brag about Star count, I assume its just a weird marketing gimmick they are using. How many stars does Linux have (or the github clone rather)? React? I have no idea, probably many but I know they are solid projects regardless of looking at their star number.

Yep. Marketers and investors care about star count. Developers rarely care.

I don't think it matters all that much.

It's like buying followers on social media: sure, with higher metrics more people may give you the benefit of the doubt when they first discover you. But at the end of the day, you need actual traction if you want to make an impact.

I can't think of a time in the last few years where it was so hard to figure out what libraries people were using to solve some problem that my only hope was to open up the top few NPM search results and sort by stars. A bad library with a lot of stars and no users is still a library nobody talks about nor uses.

The number of stars has been not very reliable on its own since a long time.

For example, a repository from Google or a random Chinese university always had much more stars than other similar repositories.

I prefer to look at the number of stars with other metrics, such as the number of issues. 3000 stars and 50 issues? Something is very fishy. 12000 stars and 6000 issues? Perhaps the software is not very good. You have exceptions of course.

I also like to look at the contributors graphs. After some time, you recognize the patterns, and you can quickly notice when the number of stars doesn't match with the graphs you see.

Exactly my thoughts. There are so many great metrics available for the repo for anyone to easily evaluate. If anything seeing an obviously massive mismatch of stars is an even better indicator of the project - it's potentially untrustworthy.

May I ask why you find it fishy? I’m one of the maintainer of an Open source project with ~2000 stars and 50 issues [1] (some bugs, some questions and a lot of enhancement proposal). I don’t think there is anything strange in these kinds of numbers.

[1] https://github.com/Orange-OpenSource/hurl

Sorry, your project doesn't have a .io domain, it must be of poor quality.

Or would you rather I look at the what it does, its contributor guide, release history, roadmap, tests, etc instead of some arbitrary cosmetic value?

You have 250 issues, which is a normal number to me. I consider the total amount of issues, not only the open ones.

If it was 50 I would think many users did star your repository because the idea sounds useful but few use it.

Ok, If you count the total number of issues, I understand you point now.

What do stars even represent, really, apart from any gaming that may or may not be taking place?

I doubt that people star things on github for the same reasons. And certain people star _far_ more things than other people. Hence even if the metric means something, it's going ti be very biased.

Poll: roughly how many stars do you award per month on github, and roughly on what basis do you award stars?

I'll start: I award maybe 1 star per month if not less. Basis: projects of close friends or spectacular demonstrations of some sort.

A ton. Usually metric is ("is interesting" or "looks good and may want to check it in some later date") and "it works".

I use stars as bookmarks. Anything I may want to come back to in the future may get a star. Projects I'm deeply familiar with won't need one.

Probably similar, 1 a month. I'm using them for bookmarks, but they also get copied by a github-backup cronjob.

Certainly, plenty places also are pushy about reminding people to star. There is no convention for what a star even means, so they've always been a pretty useless signal. Personally, a project hyping how many stars they have is a negative signal to me, and I don't really consider star count when evaluating things.

Stars don't represent anything real. You don't star a project when you clone it, when you grab a packaged version from a package repository or the home page, when you use it as a transitive dependency, when you scale it out successfully to thousands of machines, when you pay for a support contract, when you contribute a pull request, etc. In turn, when you do star a project, it doesn't require the project to even work, let alone do things well, respond to feedback or contributions, etc.

And startups have long asked for people to star their projects for visibility.

I don't know at what point it hasn't been "gamed." Maybe there are now bots starring repos, but is that meaningfully different than masses of very real humans being excited by hype?

Stars are important but not the only metric I rely upon.

Time of latest commit, number of open issues, forks, and pull requests, and number of contributors are important as well. The level of documentation is also important.

I use a lot of open source packages because often times I don’t want to spend a few hours rebuilding the wheel, like building a client library for something. If I find some that meets my criteria, I can be solving actual problems sooner.

When something doesn’t exist or is supported, it’s important to contribute either by raising an issue or filing a PR.

Anybody looking at github stats for validation deserves to see a lot!

Thats my philosophy

This repo SD AUTOMATIC111 (https://github.com/AUTOMATIC1111/stable-diffusion-webui/) is not official/ community run, has a few weeks of history, and has 14k which I think are legit.

Lots of user on github maybe? Network effects, more feed recommendations, etc

Github is becoming FB

Yes, but that example trended on the frontpage of HN which only furthers my point around the ones with zero HN references.

Reply on news.ycombinator.com