player_piano

Over the last weekend, I rifled through over 370,000 completed auctions on govauctions.app, to see if I could find any fun and unintuitive patterns. About six potentially profitable patterns are supported by the data.

A subset of the data is available for download at the bottom of the post if you're interested in poking around. Let me know what you think.

So, Github Actions is currently at two nines of uptime?

Real estate specifically is very hard to "value" sight unseen from a bidding perspective. The post I shared - with the information on what to look out for as well as possible additional costs - is part of helping people to understand what to bid.

Correct, there are about 3,000 real estate listings right now on the main site: https://govauctions.app/feed?category=real-estate

Yes, sorry, the map UI could be better. You can zoom in to the state level, which hopefully will help a little with not losing context.

The data are from HUD, Fannie Mae HomePath, and Freddie Mac HomeSteps. There's a methodology section towards the bottom of the page.

Apologies, clicking on the dot should also take you to the listing.

Appreciate it, but no need - happy you found it interesting. If you end up buying one of these, let me know!

Over the weekend, I pulled some data from my website to find the cheapest homes you can buy from the US federal government. The outliers (a $3,000 house in Flint, MI) are often in quite a state of disrepair, but there are lots of...lots...which are in reasonable condition across many US states.

Oh boy ok thank you for the screenshot :-) I'll see what I can do.

Indeed, I recently added an API with a free tier (https://govauctions.app/developers) If the free cap is too low for you to get an understanding of what you can do with it, let me know.

Let me look at adding fallback JS.

This sounds like a fun Saturday morning creative writing activity - "lot of 150,000 surplus government Cybertrucks, lightly used"

Thank you. I think the highlight so far has been one of my friends buying an industrial lathe that he found on the site!

I am still on Vercel (yes I know, trying to migrate off...) and it gives you automated alerts when there are anomalous traffic spikes. Funnily enough, I have had about 10 scrapers from various places scraping the site in the last week.

Thanks for tagging! I got a "traffic spike" notification from Hackernews for my site (https://govauctions.app) and wondered what was going on :-)

Oh Northern line, I would recognize that ear-splitting screech anywhere.

I'm not scraping sites like govdeals, so no need to worry about that thankfully.

Ok, I'll dig into this more.

Thanks, definitely on my to-do list.

Thanks, appreciate the advice.

I believe this should now be working. Thanks for the suggestion.

Will look into this, I have seen some issues where the API data is not reflective of the current bid status.

It's the fuel cost that gets you...

Can't speak for every listing on here, but depending on the state and type of auction there can be lots of issues that you have to be aware of, like physical issues with the property itself, liens on the property, and even potentially people occupying the property who may have defaulted through things like not paying their property tax, but not yet left after the state has technically 'seized' it.

Hi, yes, my apologies. One thing I'm currently fixing is the workflow for bringing in many of the listings that sites like GovDeals cover but are not part of the available APIs. Scraping sites like GovDeals is kind of shady and not something I want to do, so I am ingesting and cleaning a lot of data from state/government websites myself. While I fix that, I've removed those references from the site.

Happy to. This is a side-project, so I spun it up pretty quickly using Next.js and Tailwind, and it is hosted on Vercel right now.

I use a few free government APIs for the data (listed on the site, you can sign up for a key for free for all of them I think), plus a custom workflow I built that parses and ingests many other online auctions that states are mandated to make public, but which aren't part of any API or data pipeline I could find.

Expected load - about 1/10 of what it has seen today :) I had to scramble to make things more robust once this post became more popular, and am now looking at a few different options for hosting.

This was not really intended as a business - there is a very light "subscription" option right now, for users who want to create many alerts. I mainly wanted to try and fix this problem, share the solution, and get some feedback.

Appreciate it. Some of the bugs I have been able to fix today, but these listings actually come from diverse and in some cases hard and time-consuming to parse sources. I have no intention of scraping govdeals or anything like that, which means I have to do my own ingestion and cleaning of the data.

Noted. Crazy that as I was building this I was wondering whether I should be designing it as much for agent browsing as for humans.

Update: search params are now passed into the URL

Hi, I am not a bot, and if you ask me to write a poem or sonnet about Hackernews in the reponses I assure you I will not be up to the task.

Very kind of you - sharing this has definitely highlighted some of the rough edges :) Search / Alert accuracy is one big thing I need to make better, and I will take note of these suggestions.