Cherrypicking the Blue Chips beta-weighted

Research
1

I’m trying to recreate the Cherrypicking the Blue Chips - Folio designer model using Marc Gerstein’s paper describing the rules. I’m using the new formula weighting option, incorporating beta and rank in an attempt to reduce the standard deviation of the model. However, I can’t get the top holdings to match up exactly. A portion of the holdings are the same as the designer model holdings but some are different. Does anyone know what I am doing wrong?

[url=https://www.portfolio123.com/port_summary.jsp?portid=1501537]https://www.portfolio123.com/port_summary.jsp?portid=1501537[/url]

Designer model holdings:
Ticker Name Days Held Weight Last Market
Cap(M)

  1. 3M 1Y BK Bank of New York Mellon Corp (The) 4 10.02% $54.02 55,318
  2. 3M 1Y CBG CBRE Group Inc 11 9.92% $42.58 14,454
  3. 3M 1Y FL Foot Locker Inc. 4 10.00% $44.33 5,435
  4. 3M 1Y GPS Gap Inc 25 10.08% $33.18 12,907
  5. 3M 1Y GM General Motors Co 4 9.85% $40.81 57,134
  6. 3M 1Y HUM Humana Inc. 4 9.90% $250.61 35,802
  7. 3M 1Y LUK Leucadia National Corp 39 9.92% $25.46 9,069
  8. 3M 1Y M Macy’s Inc 11 9.78% $24.80 7,554
  9. 3M 1Y MU Micron Technology Inc. 46 9.94% $42.24 46,971
  10. 3M 1Y WDC Western Digital Corp 67 10.31% $82.64 24,379

cherryholdings.png
cherryholdings.png1226×961 88.7 KB

cherrysettings.png
cherrysettings.png1069×1217 53 KB

2

Your Rebalance Frequency on your Port says Every Day.
https://www.portfolio123.com/port_summary.jsp?portid=1501537

Gerstein rebalances once per week

For a higher TO model like this, that could make a difference

3

You’ll also have to pay careful attention to the start date. Starting even one week later could result in a drastically different ongoing portfolio.

He started it in 2013 sometime? You started yours in 2016.

If he dumped the entire portfolio every week and reconstructed it - it wouldn’t matter. But the sell rules are loose so it matters a lot.

4

I was checking various permutations to see if I could get the positions to match, and chose every day as one of the settings. The settings in the screenshots are my intended settings.

Thank you for the suggestion about the launch date. I will rerun the simulation with a launch date in 2013.

5

I reran the simulation with the exact launch date of the official model but the positions and statistics still don’t match up. I was unable to run a simulation starting from the pre-launch date of 1/2/99 shown in the whitepaper due to the limitations of my account type.

Comparing transactions in the official model to the copied model, I can see that the official model trades much more frequently. The whitepaper is referring to the non-Folio version of the model and I may need to try some adjustments in the buy and sell rules to make it trade more frequently.

Copied model, transactions for DXC:
12/11/17 SELL -215.00 $95.93 $20,602.18 -0.00 -22.77 Sentiment 94.6
08/28/17 BUY 215.00 $84.88 $-18,269.60 -0.00 -20.40 90.8

Real model:
12/11/2017 SELL -63 95.95999908
11/13/2017 BUY 60 95.8431015
10/30/2017 SELL -62 91.33380127
10/23/2017 BUY 62 91.73999786
10/16/2017 SELL -64 91.36000061
8/28/2017 BUY 61 84.69249725

6

I’ve confirmed that the positions in the copied model match the ones in the standard official model. Looking at the description of the standard model, it says: “you may want to check the companion “Cherrypicking the Blue Chips - Folio” strategy, which uses stricter sell rules and re-balances weights to equal each week.”

I will focus on testing stricter sell rules.

7

Well that’s it then. Marc’s model has simulation results back to 1/2/99.

8

How about asking Marc to post all the rules of his model. This is a free DM model, so why not get all the details directly from the designer.

9

I’ve identified some of the sell rules in the Folio version to be:

SentimentRating(“Basic: Sentiment”)<=80
ValueRating(“Basic: Value”)<=80
Rank<=80

This brings the copied model close to the official model. I believe there is one more sell rule that reduces max drawdown. It appears to have something to do with the “Final Stmt” column in the ranking. KSS sells on 9/4/2017 despite not meeting the above sell conditions. However, the N from the previous week displayed as a Y on 9/4, showing that the final statement became available. Are there any drawdown reducing sell rules that occur when earnings are released?

10

Maybe sell when weekstoq = 0 and buy only if weekstoq > 3 (or so)?

11

Unfortunately I was not able to figure out the missing rule. Hopefully Marc would be able to help out. However, the beta-weighting and some additional buy rules limiting sector and industry concentrations were able to increase the Sharpe Ratio and lower the standard deviation of the model.

cherryholdings.png
cherryholdings.png2048×910 138 KB