Without knowing all the details, there are a few reasons that explain the outperformance:
- 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")
Ops... My bad! I forgot to put in the article the number of stocks.
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!
Everything old is new again. These new AI “funds” remind me the applied research that was done in the 2000s with strategy search through genetic programming, startups like Logical Information Machines, even Peter Thiel had a macro hedge fund.
Failed to grasp what collapse data this article applies to. There is for sure a certain amount on individuals that will be able to sense structural changes if they happen to be in the right place at the right time and they have access to the right data and a set of mental models to do so. However there is a random factor at play for all the things that need to be right. There are no seers, only lucky seers.
No need for evidence of net benefits to get mass adoption. We have mass adoption of digital touchpads in cars despite evidence they are not safe. We have widespread adoption of open spaces despite evidence of them not increasing productivity..
You must be 18 in the UK for getting a tattoo, buying alcohol, watching porn, purchasing cigarettes, using a sunbed or being tried as an adult. Why would they lower the voting age to 16?
There was some talk a few years ago in the country where I live to lower the voting age. That talk was mostly driven by the parties that would benefit the most from a younger electorate. It had nothing to do with “democracy”.
- 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")