They're even here. Lots of very suspicious comments from accounts created <90 days and many accounts created after 2024ish tend to also align similarly, but with farmed karma.
The educational and informational queries were always the least valuable from a monetization standpoint. Chegg Answers could rank for these low competition (also low commercial intent) terms-- think queries like phrases from textbooks students would be querying. There is virtually 0 way (for people besides Chegg) to monetize these types of queries. Now Google can answer these queries directly, albeit with the assumption it costs them slightly more to serve these AI responses than a search query.
AI overviews are breaking the implicit "contract" for informational sites-- "we will create content to rank on Google with the expectation of monetization via display ads, mailing list growth and/or sales commissions of some sort." If these sites now lose 90% of their traffic, they simply go extinct. We have already seen the destruction of the old web era sites and the walled gardens being built. How many new sites, at the same frequency as 15 years ago, 1) get built and 2) get visibility without relying on one of the fickle walled gardens for an audience?
Google will probably figure out a way to monetize these informational queries by building better profiles of users. Or most likely, they start slipping in commercially biased responses-- either natively or disclosed, but probably based on all user conversations instead of the current one.
Look at where Fanduel/Draftking/Caesars type sportsbooks make their most margin-- it is parlays. Probably 95% of people wagering on these sites don't have even a tenuous grasp on basic statistics, yet alone how to derive actual probabilities of their action for simple spread/moneyline/total wagers. When you're letting them combine 5 wagers each with an EV of 90 cents on the dollar, the books are loving it. Layer on that these books simply ban winning players, it is insanely predatory.
Prediction markets, as they currently stand, are at least better with regard to having a lower take and are less predatory in their wagering products and marketing (although these points can very easily change, but the complex wagering menus will be less liquid and harder to grow). If the house cut is 1-3%, that is still drastically better than the other parimutuel wagering in America, horse racing, which is typically 20-25%.
Good idea. A few years back I built https://originationdata.com that compares mortgage lenders (both FDIC & FCUA members) using HMDA data. I modeled rates by lender, product type as well as by facets like MSA (as well as STL FRED data, too). It grew for a few years and I was ecstatic-- getting backlinks organically from some impressive sites (e.g. larger banks themselves, consumer publications) as well as positive user feedback. Then Google pushed their "Helpful Content Update" and Google search traffic absolutely tanked, so I kind of abandoned it and moved onto other projects that won't be SEO oriented, since Google's view of quality is unbeknownst to me.
Users are broken into two separate buckets: industry and consumers. Industry users keep using the site, based on the number of visitors with 50+ visits coming in directly every weekday. The site also gets cited by organizations with regard to their fees and rankings within geographies. This kind of proves the utility for at least this demographic.
Consumers, for a product such as mortgage, will be fragmented and infrequent users, who will only be in-market for a mortgage for a ~3-6 month window every X years. For this audience, discoverability is what matters-- and they will simply go to a search engine and look for "cincinnati mortgages" for which Google will gladly show 8-12 ads with CPCs of $20. An objective ranking based on rates and fees is useful for the consumer, but not an ad network who would rather drive multiple clicks on paid ads. Being objective and useful isn't enough to play in the space, unfortunately.
Oh snap! I was just looking at originationdata.com this week! So awesome. I had originally hoped HMDA data was more than annual, but no luck. It's also a shame that the current admin turned off the data stream here: https://www.consumerfinance.gov/owning-a-home/explore-rates/
I thought maybe you'd been hit by that update, but even more bummed to hear Google enshittification struck again.
The Modified LAR product is what you may want to look at, then. Yes, it is annual, but if you aren't against modeling data, look at the rate spread value, segment then project vs current FRED data and you'll get pretty close to actuals. You can also extract fees and derive APR in addition to having APY data.
It’s almost same price as Halo11, but seems to be far better product and fit! During first look of it, it felt expensive, but after looking at Halo11, it seems affordable :-)
The Logic of Sports Betting by Ed Miller is probably a good start.
Parimutuel vs house odds is probably a good start in seeing how odds change in different types of markets. Monte Carlo simulations will be useful in coming up with your own tissue odds. Then it is the matter of backtesting and comparing your derived odds versus the books' by looking at things like Closing Line Value, Margin of Error and Return on Investment.
For data, check out Kaggle. Learn how to scrape and circumvent platforms like PerimeterX, Recaptcha and Cloudflare. There are dozens of sites that provide historical odds data, even more basic sports statistic data.
reply