I backtested my own stock rankings. They lost to the index(holderdashboard.com) |
I backtested my own stock rankings. They lost to the index(holderdashboard.com) |
- long bull market lower the probability of achieving results above market, especially if the performance is concentrated in a select group of stocks/industries
- not knowing the factors in detail, but you might not have a period long to assess the performance. Factor performance tends to be attached to "performance regimes"
- the factors you considered in the past were underpriced, and in your evaluation window they are no longer in such state, so expected returns are lower
- market has catched up on the factors that you are using (your "free lunch" has been eaten). I guess there's a reason why nowadays quants freely join podcasts while a few years ago we had to be very careful in interviews ("do they want to hire me or do they just want to know what I'm doing/not doing")
When I run these tests, my database had about 3900 stocks. Every day I run a cron job to check delistings from SEC fillings. Then, this number decreases a few units every day.
So, replying your bullet points: 1. Full data is about 3900 common stocks listed in Nasdaq and NYSE.
2. I had outperformances for both 5y, 10y and 20y ago until Today (first chart).
3. I don't understood your point 3. Seems the opposite of what happened to me.
4. Agree with you. Market adjusts accordingly to the winners. Also a pull quote from the article. Nice!
Thank You for your comment!
also can you write up your own hn posts please
You're right: I blindly thought I could find a correlation between good fundamentals and historical returns. Since 2024 I've been working in that as a side project. Like, I pivoted the "features" of the web app a LOT of times.. Adjustable weights, added and removed Net Deb/ Ebtida from the score, etc.
The core feeling and reason for me is that I never found a good screener for common people to understand what is a good company to invest. People usually follow hype news and just bet in the stock market. I wanted to create a tool to fight against that.
And only recently, when I was about to launch my website I got AI to review it and noticed the point-in-time bias. :grimmacing:
What do you mean write your own HN posts? (This is my second post)
from the guidelines. Your blog and other comment here look very AI edited at least, although if I'm wrong I apologize.
as for the idea in general - past performance and therefore fundamentals are pretty well known to not be predictive for stocks, and public fundamentals usually aren't good enough information to trade on. You want some specific industry knowledge, right? if you could expose that maybe you could give some advice about trends. But I think you should probably just encourage people not to hold or trade non index funds.
Your top level comment is also AI which is against the HN guidelines. Also it is allowed to post your own works, but they should only be a small fraction of your submissions. Please do not use HackerNews primarily for self promotion.
I suggest you read the guidelines: https://news.ycombinator.com/newsguidelines.html
Yeah.. After I run the correlation over the full stocks of a window I noticed that maybe that could be a good path. I just didn't wanted to keep digging things and put in this article. It would be a never endless battle to "justify" my stock rankings. Right?
But Consider It done. I'll definitely check my data and see how horrible fundamentals, high debt and other bad things could help us.
One drawback I'm still trying to thing about is the problem with the Cyclical companies. If I take the data as I did before, cyclical usually would rank bad (and they do in my current ranking). So, I'm still thinking how that will distort the data.
Maybe this is naive?
However, this way you'll ended up more exposed to the index than your own choices and decisions. Right? If you had put 80% in the index in 2006 and only 4 stocks (20% of your portfolio) in 'expected good fundamental stocks', yet the majority if your whole portfolio returns would be the 80% of the invested (103% return in 10 years).
So, a conservative strategy definitely could mix a LOT of index/ETF, but I believe this would also drag the returns to be conservative. The exposition to individual stocks (if we could ever be able to have a investing strategy) will produce returns based in the risk you're buying.
I'm not writing using AI. I'm just focusing in this side project for the last 2 years. It's right now my main side-project to transform a hobby in, maybe, a revenue source... So, I've been polishing and reviewing that article a dozen times since the beginning of this month. It has a LOT of iterations with friends on missing Data Analysis arguments and bad english.
Thank you also for your topic comment: That's kind of my direction now. I want to improve the tool to focus in helping people see that a business is not a ticker and a chart. They're about knowing the forces of revenue, net income, controlling cash and debt.
I just wanted to start with a feature that could easily compare my screener to others, so I build the rankings on Fundamental Data (and also Historical return).
My idea is to bring more features in the quality side of analysis: Michael Porter's 5 forces, SWAT table, release earnings comments, etc.
I'm not a finance expert also, but I could learn over the years the (not quantitative truth) that fundamental analysis at least is more reliable than buy moment or day-trading. Right?
So, my conclusion is in line to what you mentioned: The ranking tool I build (and tried to sell as a predictive tool) also fails in that plan. It's actually just too enought to separete the gems from the rubble to help a little bit the investor who wants to consider fundamental analysis...