Withdrawn, appreciate the consideration.
I’d like this if it wasn’t yahoo price data only. There’s minimal alpha and the data source is suspect at best. Richard stated in the last FSC that those models only using yahoo data are likely submitting momentum factors in the end, and empirically, I would also say that pricing data without fundamentals isn’t going to move the needle. I am a no on this one without fundamentals.
edit: this proposal would be INCREDIBLY STRONG if you had:
1 - 85% or more of the full universe covered with price and fundamental data
2 - some framework to mitigate p-hacking
3 - use of opensignals
4 - no use of yahoo (at all, ever)
5 - by sharing your data source referral code(s) you could potentially reduce the cost to the CoE over time
6 - a plan to provide explanation (write-up, video, etc.) of your modeling framework (what I’m trying to say is just updating github with the latest version isn’t sufficient)
7 - ?
1 Like
Despite your assertation in chat, the proposal is sincere and I think the value add is on par with other proposals including yours. I’m certain though, as you go, so goes the CoE and I’m okay with that. I can keep chugging along on my own work. No harm in putting it out there in case it filled a need and a niche.
With regards to the data, you know the majority of new entrants will use Yahoo or the like to get their feet wet. And, the data you suggest would cost more then the proposed fee. Which doesn’t make sense to me and doesn’t feel like a sensible request.
No worries.
one thing I am incredibly proud of with the CoE is how independent all of us are in how we vote and evaluate projects. I find myself in the minority more often than not, and I oppose things others strongly support, and vice versa. You couldn’t be more off base on that one.
We have several example notebooks that rely on yahoo data already, and you provided one for a while before pulling it. Why not propose the CoE purchase your now-private notebook? I recall you had some gems in there.
I am ONE of SEVEN. This is not the Council of Arbitrage.
2 Likes
The idea sounds interesting @objectscience, but unfortunately one of the best bits for me personally in Signals is the individual nature of the work. So while I certainly won’t object to your proposal, I won’t support it either.
@arbitrage — whachagotaginst Yahoo? You get what you pay for, that’s true, but getting it is great for learning web scraping, data cleaning, data organization, etc.I’m debating moving to a paid service, but yahoo is fine for initial model development.
What might be useful—for those who are really systems literate and not just dilettantes like myself—would be to set up a design for a relatively easy to maintain home ticker and info database that would be practical. I’ve got mine running now, but it really is a kluge (and idiosyncratic because I work in MatLab).
3 Likes
Appreciate the input and consideration.
@arbitrage I thought about the “just sell the notebook” idea, and I’ll revisit it down the road. I want to see how well the primary model holds up over time, if it will transition to target_20d, and what the Marketplace looks like before I make a decision on that. So for now I’m just going to table the proposal and let things play out.
3 Likes