Of all the mechanical screens here in recent years, what are a couple that are good mechanical screens or ideas?
Of all the mechanical screens here in recent years, what are a couple that are good mechanical screens or ideas?
Go read some of mungofitchâs screens he has discussed. A lot of the old-time MI screens crapped out a few years ago.
2 Likes
The answer to any question often depends more on who is asked than what the question was : )
For a classic MI screen, my suggestion is unsurprisingly one of my screens.
https://discussion.fool.com/a-spy-alternative-screen-34516863.asâŚ
"This is a retirement portfolio kind of screen: LargeCapCash.
The goal is a screen which is as safe as the S&P 500 but with the hope of somewhat higher returns over the long run.
This specific screen has low turnover, lower concentration risk than the S&P with largest position at 2.5% of portfolio,
probably significantly lower company specific risk because of requirement of very strong balance sheet and profitability.
And a very strong large cap bias."
A 40-stock portfolio from that thread has beat the S&P by a nice amount in the 1.84 years since the post.
That depends which particular variant in the thread you considerâit was a long discussion.
The variations requiring a dividend seem to do a hair better, in my research.
e.g, a 39 stock âdozensâ portfolio, 13 stocks each month, each pick held 3 months, has beat SPY by 9.0%/year since the post.
Not proof of anything, but it has definitely failed to blow up so far.
That particular test variant goes like this:
VL universe
Require any valid timeliness rank
Require any positive dividend
âReturn on Shareholdersâ Equityâ top 32% (this is the figure updated annually, not quarterly)
[cash - long term debt] top 40 (gives a very strong large cap bias)
Final sort within those 40 by [cash - long term debt] / market cap
What do you end up with?
A portfolio with no position weight over 2.6% of your portfolio, much lower company specific risk than the S&P 500
Every pick very large cap. As of a month ago average market cap $141bn, smallest $1.1bn.
Every pick has very high ROE
Every company has a high cash balance
They all pay dividends. As of a month ago the picks had an average yield of 1.87%.
33.2 year backtest advantage over SPY of 4.9%/year after trading costs.
It was largely market tracking during the 1989-1999 bull market.
Since Jan 2000 the backtest beats SPY by 6.9%/year.
Iâm sure there are lots of better screens, but itâs a possible start for your search.
It depends what you have as investment goals.
Jim
22 Likes
Of all the mechanical screens here in recent years, what are a couple that are good mechanical screens or ideas?
Two of the better SIPro screens are based on changes in the amount of stock. Shrinkage looks for companies that have been buying back a lot of their stock and then sorts by high ROE. Jamieâs backtester shows that, since discovery in 2003, even for an annual hold the CAGR is around 13%, couple of points better than the S&P. For the three-year period 2018-2020 a monthly Shrinkage approach had a CAGR of 23%.
https://discussion.fool.com/3-year-sip-backtests-to-2020-novembeâŚ
In addition, FWIW, the top short screen was Dilution.
DB2
8 Likes
For a classic MI screen, my suggestion is unsurprisingly one of my screens.
https://discussion.fool.com/a-spy-alternative-screen-34516863.asâŚâŚ
"This is a retirement portfolio kind of screen: LargeCapCash.
--------------------------------------------------------------------------------------------------------Jim,
What happens to this screen if you add momentum, like 6 month gain, or % of 52 week high?
A lot of the old-time MI screens crapped out a few years ago.
Unfortunately this seems to be true.
I just took a look at the performance of all of the standard screens in lohillâs GTR helper. Picked the top 10 screens based on their perpetual SAWR (safe annual withdrawal rate). Then compared their last 10 years performance compared to the SP500.
Screen Name CAGR TR Log2TR SAWRĂ20) GSD(20) DIGSDĂ20 LDD LDDD3 MDD Ul(20) Sharpe(20) Beta(20) Tl(20 AT Start Date Years CAGR10y SAWR10y
WK_Voom 31.0 1,595,608 13.5 12.7 36.7 39.2 19.3 12.2 -55.7 15.1 0.93 0.9 31.2 8.6 19870302 35.08 21.9 13.4
WK_Voom_Liquidity 32.1 2,004,279 14.0 12.4 35.5 38.0 18.6 11.9 -55.5 15.2 0.97 0.9 32.6 8.4 19870302 35.08 19.4 12.3
S&P500 10.6 3,343 5.1 5.5 17.6 20.7 11.8 10.7 -54.8 13.8 0.53 1.0 8.5 0.0 19870302 35.08 14.4 12.3
Up5X3 27.6 46,022 8.5 12.9 33.7 38.2 19.2 13.7 -62.9 15.1 0.94 1.0 27.2 12.2 19970902 24.58 15.0 10.2
RS-100 31.9 2,421,781 13.9 12.4 35.5 38.3 18.8 10.9 -55.7 14.9 0.96 0.9 33.6 11.1 19870302 35.08 13.3 8.8
LowPS+ 34.5 4,642,333 14.8 12.1 43.5 47.0 22.8 14.6 -71.3 18.0 0.90 1.1 31.1 11.8 19870302 35.08 13.8 8.5
YEY_SIPro 29.2 811,773 12.9 11.2 31.6 35.6 17.7 12.0 -61.1 13.6 0.98 0.9 29.2 5.9 19870302 35.08 13.3 8.1
Low_Mult 27.1 671,405 12.0 10.6 33.7 35.7 18.0 11.1 -51.8 14.9 0.87 0.8 30.4 9.5 19870302 35.08 12.8 8.0
VG-Horse 31.9 108,780 9.7 12.4 41.3 42.1 20.7 13.3 -59.5 18.1 0.90 0.9 35.5 9.9 19970902 24.58 12.5 7.7
Rabbitt 29.2 68,719 9.0 11.4 37.3 40.4 20.1 13.8 -67.6 17.5 0.91 0.9 30.7 11.2 19970902 24.58 8.0 4.8
BI 24.4 23,831 7.6 10.3 31.6 34.1 17.4 12.3 -53.3 14.8 0.87 0.8 31.5 10.7 19970902 24.58 2.7 1.7
Assuming a 0.35% friction on trades only 2 of the 10 screens outperformed the SP500 based on SAWR.
Following Mungoâs advice to use his screen doesnât seem to be a bad idea. Iâve been using it with only SIP data and a slight mod to filter out very high or low beta stocks that has outperformed the S&P for the last couple of years.
RAM
4 Likes
qqq has beaten the s&p by 30% over the last 5 years
2 Likes
qqq has beaten the s&p by 30% over the last 5 years
The question is, do you extrapolate that or expect mean reversion?
My personal opinions on thatâ
Theyâre both overvalued compared to any prior norms, but SPY more so.
And the value of the Nasdaq 100 companies rises, on average, very much more rapidly than those in the S&P 500.
So on a relative basis, over a decently long time frame, I would extrapolate the result to be continued outperformance of the Nasdaq side.
At a moderate rate.
However I MUCH prefer QQQE over QQQ. Much more of a sure thing, vastly lower company specific risk.
QQQ is very oddly weighted (not really cap weight), with huge concentration in just a few names.
If you know those few giant names and want to own them and have an opinion about their valuations and prospects, buy them individually.
If you donât have a solid opinion on them, you donât want a portfolio concentrated in them.
Either way, donât buy QQQ.
Jim
14 Likes
âŚbuy QQQEâŚ
never even occurred to me there was an equal weight ETF on the Naz 100. Thanks for the directional, Mungo - implementing tomorrow.
Do you happen to know any free resources to generate the âLargeCapCashâ screen? (or an variation of it)
Thank you
1 Like
Do you happen to know any free resources to generate the âLargeCapCashâ screen? (or an variation of it)
You can do it yourself with a little bit of work. Well, using Russell 1000 instead of VL. My local libraries cancelled their VL subscriptions, so I canât use VL anymore.
Download the R1000 data from barcharts (free login). Doesnât have cash & debt, though. You have to grab that data from other sites, like ycharts. You probably have to scrape the data, which is a PITA.
2 Likes
I should probably add, âLearn to code.â LOL
I had to do it all in scripts (bash). I doubt there is another way if you canât do it all with VL.
One option is to subscribe to Portfolio123 Screener which also gets you 5 years of backtesting for $300/yr.
Second option is to subscribe to a AAII Stock Investor Pro for $300/yr. With that you can either run the screen in GTR1 which allows you to find out the current picks if you have the MD5 has from your SIP subscription download. Or you can run the screen using the less sophisticated built in SIP screener.
Iâve tried both and either works, the P123 is the better screener but SIP enables longer backtests using GTR1. P123 has 20 year backtests but that costs $1K/year min (they also have the lowest price for Compustat data for only an additional $12,000/year for any big time rollers here)
I have backtested a simplified version using only SIPro data and it does almost but defiantly not as well as the GTR1 version.
SP = 500 OR ci.SP = 400)
IND_2_DIG <> 7
IND_2_DIG <> 12
COUNTRY = âUnited Statesâ
CASH_Q1 > 0
Equity_Q1 > 0
NETINC_12M >0
fcfps_12M > 0
RROE_12m > 85
SORT BY ((CASH_Q1 - LTDEBT_Q1)/MKTCAP) Desc
RAM
5 Likes
I just took a look at the performance of all of the standard screens in lohillâs GTR helper. Picked the top 10 screens based on their perpetual SAWR (safe annual withdrawal rate). Then compared their last 10 years performance compared to the SP500.
RAM
I donât totally follow your logic. If they are compounding that much more than the s&p, they sound pretty good.
RAM; I donât totally follow your logic. If they are compounding that much more than the s&p, they sound pretty good.
Perhaps my data formatting which had too much superfluous information with the results off the page.
Focus on the right two columns. . .the results for the last 10 years.
Screen Name CAGR SAWR Start_Date Years CAGR10y SAWR10y
WK_Voom 31 12.7 19870302 35.08 21.9 13.4
WK_Voom_Liquidity 32.1 12.4 19870302 35.08 19.4 12.3
**S&P500 10.6 5.5 19870302 35.08 14.4 12.3**
Up5X3 27.6 12.9 19970902 24.58 15.0 10.2
RS-100 31.9 12.4 19870302 35.08 13.3 8.8
LowPS+ 34.5 12.1 19870302 35.08 13.8 8.5
YEY_SIPro 29.2 11.2 19870302 35.08 13.3 8.1
Low_Mult 27.1 10.6 19870302 35.08 12.8 8.0
VG-Horse 31.9 12.4 19970902 24.58 12.5 7.7
Rabbitt 29.2 11.4 19970902 24.58 8.0 4.8
BI 24.4 10.3 19970902 24.58 2.7 1.7
The top 10 screens over the last 24 to 35 years which includes the last 10 poorly performing years:
Had a median CAGR of 30.1% compared to the SP500âs 10.6% over the same period.
Had a median SAWR of 12.5% compared to the SP500âs 5.5% over the same period.
But over the last 10 years.
Had a median CAGR of 13.3% compared to the SP500âs 14.4% and only 3 of 10 better CAGR.
Had a median SAWR of 8.3% compared to the SP500âs 12.3% and only 2 of 10 better SAWR.
My limited logic says the screens were more than likely a formula to lose money.
RAM . . . Still donât know how Iâm a little ahead of the game.
13 Likes
I created a Portfolio123 version:
Universe(Prussell1000)
DivPaidA > 0
FOrderOLD(â$Timelinessâ,#All,#DESC,#Previous,TRUE)<=600
FOrderOLD(âROE%Aâ,#All,#DESC,#Previous,TRUE)<=80
OPMgn%TTM > 25
FOrderOLD(âCashA - DbtLTAâ,#All,#DESC,#Previous,TRUE)<=40
FOrderOLD(â((CashA - DbtLTA) / MktCap)â,#All,#DESC,#Previous,TRUE)<=13
where the custom formula $Timeliness: BetaTotalReturn26W(CurFYEPSMean/EPSInclXorA)
I added the Operating Margin > 25% because it eliminated all of the low margin businesses (IT, Services, etc.) and pumped up the annual return by several percent. My current P123 subscription only allows 5 years of backtesting. This screen has an Alpha over the S&P500 of 11.22% over the past five years with 0.5% slippage on each trade and 1 year holds. 13 week holds had a 7.72% Alpha over the same period. I could not figure out how to do the alternating 13 week holds. If anyone has a longer P123 backtest subscription I would be curious to see how it fares.
Taz
1 Like
I got it now. I think I was looking at the last 24 - 35 years.
Taz2: I added the Operating Margin > 25% because it eliminated all of the low margin businesses
(IT, Services, etc.) and pumped up the annual return by several percent.
I tried using the SIP field Operating margin12m > 12% in a more elaborate GTR1 version which I like because it has the advantage of using daily start dates which are especially valuable in backtests over shorter time periods. Was only able to go back to 20000403 but:
For the last 22 years the Operating margin reduced CAGR by 5% and SAWR by 5.5%.
For the last 5 years the operating margin requirement reduce both CAGR and SAWR by 3.5%.
For a short backtest of only 5 years the daily start test results of CAGR varied by over an 8% range from min to max depending on the start day. Likewise, the SAWR varied 6.7% from min to max. The daily start variation of CAGR min to max was only 2.8% and less than 1.9% for the SAWR.
The more one uses Robbieâs GTR1 backtester the more you can appreciate the sophistication. Not easy to learn but worth the effort.
RAM
4 Likes
My local libraries cancelled their VL subscriptions, so I canât use VL anymore.
You could get a subscription to the Value Line service.
Just sayinâ.
Jim
8 Likes
You could get a subscription to the Value Line service.
Just sayinâ.
$598 for âThe Value Line Investment SurveyÂŽ - Smart Investorâ Is that the one? VL web site is very unclear on the details. I assume thatâs per year? Again, VL does not say.
For using it only one or two times a year, either a 1 1/2 hour drive to the physical library or paying $45 for an out-of-district library card (which would give me online access) is more like it.
'course, if we get into a bear market like many are predicting, wonât be holding stocks for a year or 18 months.