Post New

DrBob2 combined PFCF and High Price in a screen about 20 years ago. Based on this, I made a new screen with more emphasis on High Price, with good backtest results:

Define {HighPrice2024a}
step0: no ETFs or CEFs or LPs
step1: no IPOs
step2: [Average dollar-volume over 63 days] > 0
step3: [Average dollar-volume over 63 days] Top 50%
step4: no recent mergers
step5: issued financial report within the last year
step6: [PFCF] > 0
step7: not in Finance sector
step8: Price Top 10
Holding period = 21 mkt days, 0.4% friction

           Screen             CAGR  GSD  Sharpe    From       To
       HighPrice2024a          17   24    0.71   19870302  20240718
100ClubNoFin_20021214_DrBob2   13   28    0.52   19870302  20240718
      SP500MktCapWeight        11   18    0.54   19870302  20240718
 
       HighPrice2024a          21   20    1.12   20090717  20240718
100ClubNoFin_20021214_DrBob2   14   31    0.61   20090717  20240718
      SP500MktCapWeight        15   16    0.94   20090717  20240718

{HighPrice2024a} picks for 20240718 (Berkshire Hathaway is in the SIP Consumer Non-Cyclicals sector):

Ticker   aprc   [PFCF]  styp.a    adv:1,63               Company Name
BRK-A   662820    31      11    5,313,969,000  BERKSHIRE HATHAWAY INC CL A
  NVR     8408    22      11      154,309,100  NVR INC
 BKNG     3931    22      11      816,848,600  BOOKING HOLDINGS INC
  SEB     3134    35      11       11,713,640  SEABOARD CORP
  AZO     2926    28      11      411,073,200  AUTOZONE INC
 MELI     1623    16      11      566,720,400  MERCADOLIBRE INC
 FICO     1585    85      11      270,210,400  FAIR ISAAC CORPROATION
  MTD     1356    33      11      207,045,300  METTLER TOLEDO INTERNATIONAL INC
 COKE     1089    30      11       66,204,860  COCA COLA CONSOLIDATED INC
 REGN     1067    30      11      439,984,000  REGENERON PHARMACEUTICALS INC
 ORLY     1028    31      11      432,283,500  O REILLY AUTOMOTIVE INC
  GWW      957    35      11      245,622,900  W W GRAINGER INC
 LRCX      953    36      11      970,600,300  LAM RESEARCH CORP
 ASML      924   9258     12    1,049,331,000  ASML HOLDING N V
 DECK      887    24      11      327,284,000  DECKERS OUTDOOR CORP
 MPWR      843    102     11      376,286,800  MONOLITHIC POWER SYSTEMS INC
 KLAC      787    46      11      612,061,700  KLA CORP
  GHC      776    31      11       13,340,320  GRAHAM HOLDINGS COMPANY CL B
 CTAS      759    77      11      277,006,200  CINTAS CORP
  NOW      739    48      11    1,024,738,000  SERVICENOW INC
  URI      728    12      11      376,779,900  UNITED RENTALS INC
 NFLX      643    41      11    2,208,407,000  NETFLIX INC
 INTU      637    44      11      888,018,300  INTUIT INC
  MCK      578    21      11      408,023,200  MCKESSON CORP

The 100 Club - a value screen
http://www.datahelper.com/mi/search.phtml?thread=1...

Define {100ClubNoFin_20021214_DrBob2}
step0: no ETFs or CEFs or LPs
step1: no recent mergers
step2: no IPOs
step3: [Average dollar-volume over 63 days] > 1,000,000
step4: Price > 100
step5: [MktCap] > 100
step6: [PFCF] > 0
step7: no Finance sector
step8: [PFCF] Bottom 5
Holding period = 63 mkt days, 0.4% friction

https://gtr1.net/2013/?~HighPrice2024a:h21f0.4::st...

https://gtr1.net/2013/?~100ClubNoFin_20021214_DrBo...

All absolutely SOLID firms!

I think the $ price value skews this towards BRK-A and NVR ( 2 of the longest scrips without a Stock split).... How about allowing for a little more leeway - include a rough criteria ( looks like the lowest valued is ~$600) - where you reweight it ( ie reverse adjust) for the Split - where it split in the last 2 years or something. or even historical.

Alphabet, Apple and latest NVDA and even TSLA might have a shot to be included. Otherwise this immediately penalizes a stock for a split. - Although that could be even a performance factor

Best

add tr(1,200) top 5

cagr 21.7
sharpe .75
beta 1

You don't need such a high .4 transaction cost for big stocks.
A study showed post Nasdaq100 formation .1 is adequate


https://gtr1.net/2013/?~HighPrice2024a:h21f0.1::st...

add tr(1,200) top 5

cagr 21.7
sharpe .75
beta 1

Only 6 negative years in 37 years. One down year since 2008. SAWR 12+%. Pretty darn good.

Aussi

And yet… the rationale is buy the highest priced non-financial stocks with positive cash flow, because… they will keep goin g up? High price begets buying in this megacap driven world?

Seems a thin premise but the GTR1 doesn’t lie

the rationale is buy the highest priced non-financial stocks with positive cash flow, because… they will keep going up? High price begets buying in this megacap driven world?

Seems a thin premise but the GTR1 doesn’t lie


Stated this way, it seems to be a case of shooting arrows at a wall and then drawing the bullseyes afterwards.

Trade only every 3 months, and add BCC timing step to go to cash periodically:

CAGR 21.678%
sharpe .796
beta .791
mdd -59.6 (essentially the same as S&P)

https://gtr1.net/2013/?~HighPrice2024a:h63f0.1::st...

Seems a thin premise but the GTR1 doesn’t lie

Right, GTR1 doesn't lie. But it allows us to lie to ourselves. It's called data mining.

Elan

The price of a stock is well known as one of the better fundamental predictors of future gain. There are several papers mentioning this. When using AI for selecting stocks if you input price to a Random Forest you can ask for the relative value of the predicting factors and last close always comes back as a good indicator. But I never thought the very highest would have that kind of return.
Using GTR1 for stocks from 1987 to 2024:
Top 5% by price alone CAGR 8.0%
Top 10% by price CAGR 8.4%
Top 33% by price CAGR 9.7
Bottom 33% by price CAGR 3.82%
Bottom 10% by price CAGR 0.1%
Bottom 5% by price CAGR -2.8%
https://gtr1.net/2013/?~Price2020to2024:s20000103h...

Bob....

(1) The concept is intriguing but needs a deeper look and much better understanding. Someone like Jim needs to weigh in a bit.

Please refer to what I initially said and elann said - "it might just be leading down the garden path - based on what you want to believe". While Ray completely trashed it - saying it has look-forward bias ( technical interpretation of what he said) - NOT TRUE ..... but there's a catch.

(2) The screen has a hidden "survivorship bias" - this is because you start with a Pool of stocks and assume the price ( embedded in the appreciation/depreciation as indicative) and you will have them till the end of the study period : INCORRECT.

There's a EXIT/DROPOUT route - thru Stock-Splits. I dont have access - but if you look thru the picks you will find Alphabet/Google, Apple, NVDA , not sure about Tesla ( since Free Cash Flow is a criteria) etc amongst the picks and then being censored ie Non Surviving once the Split is done.

So the KEY QUESTION: Does picking a High price stock and holding it till a Split ( which may or may not happen) - provide great returns? IDK>

(3) and because of the above and based on whatever little knowledge I have of the market - this is basically (skewed) as a Berkshire (A class) , NVR ( They are a conservative Homebuilding/Real Estate Company) screen. Those 2 are the longest running stocks without a split ( Note BRK-B is not on this) . The latest entrant to the list would be Booking.com

I tried to run a test using those 2 to compare - but GTR1 keeps giving me errors

(4) One key robustness test of this concept could be to ensure that ALL stocks that are eligible at the start are adjusted upwards for Stock splits and downward for reverse ones ( this is rare - because most of the time this happens in the lower end of the spectrum ie rarely do you see a $150 stock do a reverse and become a $1500 stock) and then compare it to the screen to understand its fundamental edge

OT : Bob .... be very very careful about using direct price in these ML learners. They simply would end up "committing it to memory" associating it - and then going forward if price moves away - the rules will not have support.

Also Last Price being best bet = Random Walk theory.

Suggestions/Alternatives : Follow the percentile concept like you did for the screen - Both Between Groups - ie Close as a % tile of all the classes of stocks in the cohort and Within group - ie %tile of the stocks price relative to itself ala Stochastics ( its funny how much of TA is simply a renaming/repackaging of Basic Statistical measures!)

Best

Stated this way, it seems to be a case of shooting arrows at a wall and then drawing the bullseyes afterwards.

Except that it is not entirely afterwards. My High Price screen from 20 years ago was based on the work of Ben Graham some 60 years ago.

DB2

anchak….
I don’t have time for hypotheticals but I did several backtests in different universes. The backtester I used includes dividends, slippage, splits, name changes and lists stocks that it can’t find at sell. As for stocks missing at sell time you can assume you have a 100% loss on every one and it doesn’t change the conclusion because there are close to the same number in the high price selections and the low price selections. Another interesting factor is holding the higher priced stocks also results in a higher sharp.

First of all, I’ll acknowledge that price as a factor doesn’t work at all in the S&P. But it works both in the US and global stocks and especially smaller caps.

I’m not sure you can get to the paper but “Machine learning goes global: Cross-sectional return predictability in international stock markets” The abstract “We examine return predictability with machine learning in 46 stock markets around the world. We calculate 148 firm characteristics and use them to feed a repertoire of different models. The algorithms extract predictability mainly from simple yet popular factor types—such as mo mentum, reversal, value, and size. All individual models generate substantial economic gains; however, combining them proves particularly effective. Despite the overall robustness, the ma chine learning performance depends heavily on firm size and availability of recent information. Furthermore, it varies internationally along two critical dimensions: the number of listed firms in the market and the average idiosyncratic risk limiting arbitrage.”

Across all countries using 11 different Machine Learning models the found price was 11th most significant factor in predicting future price.

Might not seem logical and I don’t understand why except perhaps failing companies usually slowly decline in price before they disappear and very successful companies don’t have to lower their price to attract investors. But I’m just telling you what works using machine learning on historical data tells me and many others.

By the way how can ML commit the price to memory of an individual stock? ML training data has been stripped of all stock ID’s and Ticker information before training.

First .... We need to be aligned on the premise that High Price ( combined with other safeguarding factors) - should tend to lead you to better firms. And this is amply clear from the filtered list!

The questions are 2 critical items

(a) Is the premise to hold a Top X% tile price stock till it splits - which in essence seems to be the screen's performance. This outcome is of course not guaranteed - I would wager a majority of the past holds and exits are due to this option

If so, then its also incumbent to benchmark it to a simple 2 stock screen ie BRK-A and NVR ( Somehow 2 stocks which seem to have bucked the Split exit path)

(b) If not, ie the high price itself is a demarcation factor - then those stocks need to be eligible for picking forward irrespective of the Split adjusted price .... ie like Google should be ~$6800 stock. The Split can't be an exit condition.

My guess is that the way the screen is constructed now - it does (a) while completely obfuscating this KEY EXIT factor


Bob - on the ML ( specifically tree-class algos - but it also generally applies). No it wont learn the individual stocks price - but it would learn the differential boundary criteria of the underlying cohort - eg If we believe the current picks are OUTSTANDING firms - then the price demarcation is appx Price > $575. And if say all of this holds true( ie the set continues to dominate and doesnt alter at the top) and the stock market doubles in 2-3 years and you hold the model constant ...... that rule will throw any Stock with price > 575 into the mix. But if the cohort has remained constant and it also doubled ( with no Splits etc) - the threshold now has moved to $1150.

You just have to have the mechanism to embed/factor this indexing into the model

Best

QUOTE "Except that it is not entirely afterwards. My High Price screen from 20 years ago was based on the work of Ben Graham some 60 years ago. ENDQUOTE

This is the reason why I said Ray's statement was a bit off-base. Except - I dont think Ben Graham would sell off a stock just because it split its price and $1200 is now $120 ( I am specifically referring to his august disciple's holding of AAPL - which I dont think Buffet sold either )

The screen is doing that - at least that's the hypothesis - I just somehow cant get GTR1 to work effectively or understand as others do here!

Best

this is basically (skewed) as a Berkshire (A class) , NVR ( They are a conservative Homebuilding/Real Estate Company) screen. Those 2 are the longest running stocks without a split

The OP screen {HighPrice2024a} has a Berkshire bias, but this does not explain the recent outperformance. Berkshire outperformed the market until 2002, but only matched market performance after 2002. {HighPrice2024a} could be modified to drop the highest priced stock, but I prefer to keep BRK-A in the screen going forward.

      Screen       CAGR  GSD  MDD  Sharpe    From       To
  HighPrice2024a    16   27   -59   0.52   19870302  20020628
  HighestPrice      21   27   -49   0.71   19870302  20020628
      BRK-A         21   27   -49   0.71   19870302  20020628
SP500MktCapWeight   11   17   -35   0.44   19870302  20020628
 
  HighPrice2024a    18   22   -54   0.88   20020628  20240725
  HighestPrice      11   19   -51   0.59   20020628  20240725
      BRK-A         11   19   -51   0.59   20020628  20240725
SP500MktCapWeight   10   18   -54   0.60   20020628  20240725

https://gtr1.net/2013/?~HighestPrice:h253::aprc:tn...
https://gtr1.net/2013/?h1::trp%281,1%29ne-999%7bU:...

Some screens benefit from dropping the top pick. For example:
Mega-Cap Screen "Notice how the Top-1 pick appears to be a significant under-performer."
http://www.datahelper.com/mi/search.phtml?thread=3...

NVR's price went above $200 in 2001, and NVR greatly outperformed the market since, with a CAGR of 19% from 20010725 to 20240725. {HighPrice2024a} CAGR was 17% over these 23 years, and so NVR boosted the screen's performance but was not the only winner in the screen. Dropping the top 2 highest prices, {HighPrice2024a_drop2} CAGR was 16. SPY CAGR was 9%.

https://gtr1.net/2013/?~HighPrice2024a_drop2:h21f0...


The stock price needed to pass {HighPrice2024a}:

Limit    Date    NVR
 200   20000308   44
 100   20001114  102
 100   20010503  185
 100   20040402  451
 100   20050429  718
 100   20081007  539
 100   20090310  363
 200   20110201  772
 300   20111221  676
 350   20150122  1280
 400   20150427  1339
 500   20170515  2235
 600   20190212  2691
 700   20190611  3371
 800   20200727  3912
1000   20200803  3874
1200   20201228  4157
1400   20210706  5186

https://gtr1.net/2013/?~HighPrice2024a_limits:h21f...

Post New

US Policy FAQ

---

Contact Shrewd'm
Contact the developer of these message boards.