Always treat others with respect and kindness, even if you disagree with them. Avoid making personal attacks or insulting others, and try to maintain a civil and constructive tone in your discussions.
- Manlobbi
Investment Strategies / Mechanical Investing
No. of Recommendations: 9
Comparison between similar screens.
Two similar screens, using the Nasdaq 100.
First is ranking by "price/(52wkHIGH + 52wkLOW)", top 10 HTD 15
Second is "52 week relative strength", top 10 HTD 12
Recast monthly, equal balance of only the new picks.
The latest picks at end of November, 7 of the 10 stocks were in both screens. Probably not a huge surprise, but much more of an overlap than I expected.
The backtests show the first has about 21.7% CAGR and the second about 20.0%.
The first has a bit lower GSD, and a bit higher turnover.
The relative performance is the same with starting dates of 1990, 2002, 2015 with end date of 10/10/2025.
Using the S&P 500 the relative performances are the same, just with lower CAGR.
No. of Recommendations: 11
I haven't compared those exact strategies, especially as I usually look at deeper screens.
But the reason for suggesting the (price/(high+low)) expression is that it works better than (tr1y) or (price/high) during those rare times that the market is rebounding strongly after a recent deep sell-off. So the returns relative to market are pretty good during those rare extraordinary times, meaning the odds of beating the market in any given rolling year rise even if the end-to-end CAGR measures the same as other momentum metrics. It often increases the fraction of positions that are profitable, as a different measure of the same effect.
That makes the screen easier to stick with, and you get earlier evidence on the degree to which the screen is behaving as intended: if you're expecting it to add value in (say) 70% of rolling years, and it doesn't do that in the first three years, you have reason to be suspicious sooner than if you're running a screen with the same overall CAGR that manages it with rarer very high returns.
In short, it's intended to make the screen add some value in a wider variety of market regimes. Whether it does or not is of course a separate question, but that was the intent : )
Here's a Nasdaq 100 screen I created in 2019:
Nasdaq 100 member
Price/(52-week high+52-week low) top 50
5-year sales growth rate > 0
ROE * (5-year sales growth rate) top N
The top 10 monthly before friction beat the S&P in ~82% of rolling years since then. Overall CAGR advantage relative to S&P 500 over 9%/year out of sample.
The time period covers the pandemic plunge and rebound, so it's a nice test for the idea. The backtest returns in 2019, 2020 and 2021 are remarkably similar. 2022 horrible, as with many things.
(actually I used the largest 105 stocks listed on Nasdaq rather than nas100. So it might have included some financial firms. The Nasdaq 100 is basically non-financial market cap top 105, hold-till-drop 100)
Jim
No. of Recommendations: 1
Calculating CAPE is pretty hard. Could you just use the median p/e?
No. of Recommendations: 1
And what is the holding period for this screen?
Thanks as always for your great ideas.
No. of Recommendations: 3
And what is the holding period for this screen?
If you mean the "pick good industries that aren't above their usual respective valuations" screen, I have no idea. I was talking out of my butt when proposing it, just a riff on the OP's direction. It doesn't sound like the sort of thing that would make you broke, though, especially if you have positions in several plausible industries.
If you meant the MI screen I proposed, it seems to work well at 1 or 2 month holds, virtually no difference. Might be better with changing to top 150 Nasdaq stocks rather than just Nas100, according a couple of tests I did. Maybe not.
Jim
No. of Recommendations: 5
Calculating CAPE is pretty hard. Could you just use the median p/e?
You generally can't do math on P/E ratios. (imagine one period of zero or negative earnings...) You can calculate the median P/E, but you can't average a bunch of those, which you probably need to do. Instead, do math on earnings yields. So you can calculate the average EY=E/P over time or among companies.
I think the average (through time) of the median (among the companies) EY would be the best, it would give a very stable figure you could average to get a sense of "normal" valuation. Then compare the current median EY among companies to that stable historical norm. This is pretty much the method I use to value the Nasdaq 100.
Jim
