No. of Recommendations: 4
I'm retired, so I focus more on risk reduction than beating the market. I liked Jim's ROE_Cash idea and the concept of having my own pseudo-index-tracking portfolio; however, if possible, I wanted the Maxx DD, Std Dev, and Beta to be lower than the standard index while still approximating its annualized return.
After trial and error I arrived at the following concept:
i) Use the Russell1000 as the stock universe (1000 largest Cap)
ii) Take ROE_Cash
iii) add market timing
BenchmarkPrice > 200-day SMA
VIX < 200-day SMA OR VIX < 15
VIX < 30
iv) stay out of industries that collapsed during previous recessions
Sectors: BIZSVCE, CONSUMERSVCE, CYCLICALS, FINANCIAL, INDUSTRIAL
Subsectors: STOREDEPT, MERCHRETAIL, PRODHOUSE, INSURANCE, TECHCONSULTSVCE
Industrys: FOODPRODUCTION, FOODDRINKMFR, PRODHOUSEIND, TEXTILE
v) eliminate companies that have wildly divergent analysts estimates
Between(ProjPECurFY , 0.5 * FMedian(" ProjPECurFY") , 2*FMedian(" ProjPECurFY") )
vi) keep companies whoes ROC > WACC + 10
vii) add a final sort on the 20 companies with the highest Timeliness Rating.
I use P123 so I don't have access to VL's timeliness rating. However, back in the mid '90's I had a friend who was a whiz with SPSS so one month we entered all the data on VL T1 stocks to see if SPSS could find a definite correlation of variables leading to T1 stocks. The following did pretty well in 1995:
Timeliness = Beta * TotalReturn26W * (CurFYEPSMean/EPSInclXorA)
It wasn't perfect, but it did identify around 3/4 of the T1 stocks in 1995.
Anyway, over a 23-year P123 backtest ROE_Cash_Defensive has the following numbers:
Annualized Return: S&P500 8.42%, Screen 9.62%
Max DD: S&P -55%, Screen -18%
Std Dev: S&P 15%, Screen 11%
Beta: Screen 0.33
It stayed out of the market during the 2001 Internet Crash (10-01 thru 3-03), the 2007 recession (07-07 thru 06-09), and the bulk of the COVID crash (12-21 thru 01-23).
I was wondering if this would hold up if done using VL and Jamie's screener or is this just a result of too much data mining on my part?
Taz
No. of Recommendations: 6
or is this just a result of too much data mining on my part?
Probably.
(I don't have P123 anymore, but I seem to remember that you could upload your own universe.)
This is 7 rules. That seems like too many. The more rules the more likely it is datamined.
Actually it is more like 9 rules, because #3 (market timing) is 3 rules.
I don't remember the details of ROE_Cash, but are there multiple rules & sorting in it?
Your #4 "stay out of industries that collapsed during previous recessions" is absolutely data mining.
This MaxDD of -18% vs. S&P's -55% is curious. I wonder if that isn't due to the datamined #4.
Ah, just before posting I saw this: After trial and error I arrived at the following concept:
That's the brightline hallmark of a data mined screen. "Try a bunch of things until we find one that gives the answer we like."
Every once in a while when we are checking out at the grocery store, the bill will come to an exact dollar amount with 0 cents, like $97.00. I will say in astonishment "Wow, this must be our lucky day, Zero cents. What are the odds of that happening?" Usually the clerk pauses a second then gets the joke and chuckles. My wife punches me in the arm and says "Don't laugh at him, he thinks he is funny. That just encourages him."
