Putting Piotroski to the Test
May 26, 2022
Show all

Two bad days in a row

Two bad days in a row seem to be bad news… at least when rebalanced quarterly, or bi-monthly. Here we also limit it to stocks with a market cap over $1 billion, so that the large number of smaller stocks don’t bias the results.

Backtest Report for Jul 16, 2000 to Jul 16, 2021 for Two Bad Days In a Row

This normal screen rebalances quarterly. It (vividly) explores what happens to stocks that lose more than 4% on two consecutive days.

Model


There are several ways to create the model:

  1. Type it all in – “change close 1 < -4 and (change close 1) as of 1 < -4 and marketcap > 1b”. The first two terms get grouped together, while the market cap term is alone, but the meaning is the same.
  2. Build it up – Type the “and” first, then add a term, so that the and term has three empty boxes. After that fill each one in, typing change for the first two, and “mc > 1b” for the third. Fill in the two boxes with close and 1 day each, and stare at the identical first lines. Realize you forgot to offset the second one by one day, and type “as of” in the second change, replacing the text, and hit return. Fill in the new box with a 1 and put the -4 on each of the two boxes at the end.
  3. Try something completely different – use “(maximum_within (change close 1) 2) < -4” as a replacement for the first two lines. Same meaning, but less text. And testing 5 days is much easier with this formulation.

Why this model? Stocks rarely fall on their face multiple days in a row, and we’re capturing those that do, and figuring out what, if anything, such a faceplant means.

Visualizations



This does not look like a great chart! This made negative money over the 20 year span, with few deviations.


Eleven of the past twelve years showed this screen going down. This is bad enough, but it didn’t do that well in most of its up years, outperforming the S&P 500 in only six years.


Two down days in a row gives us stocks that are not strongly correlated with the market – at least the average week is weekly correlated with the market. Of course, since that is because this screen spends more time going down than up, that is not a good thing.

Return Summary

Up to Jun 30

Returns S&P 500 Risk Free

Jun 2020 to Jun 2021

64% 40.9% 0.08%

Jun 2019 to Jun 2020

-30.39% 6.41% 1.21%

Jun 2018 to Jun 2019

-52.65% 10.92% 2.36%

Jun 2017 to Jun 2018

-20.03% 14.53% 1.58%

Jun 2016 to Jun 2017

-11.87% 17.78% 0.69%

Jun 2015 to Jun 2016

-34.49% 3.99% 0.33%

Jun 2014 to Jun 2015

48.71% 7.24% 0.08%

Jun 2013 to Jun 2014

40.52% 23.68% 0.07%

Jun 2012 to Jun 2013

59.09% 20.85% 0.12%

Jun 2011 to Jun 2012

-8.59% 5.63% 0.09%

Jun 2010 to Jun 2011

41.2% 30.41% 0.16%

Jun 2009 to Jun 2010

-16.87% 14.41% 0.21%

Jun 2008 to Jun 2009

-26.74% -26.22% 0.84%

Jun 2007 to Jun 2008

-31.27% -14% 3.06%

Jun 2006 to Jun 2007

-0.004% 21.44% 5.1%

Jun 2005 to Jun 2006

-34.24% 8.73% 4.39%

Jun 2004 to Jun 2005

-13.37% 6.14% 2.53%

Jun 2003 to Jun 2004

18.3% 19.17% 1.1%

Jun 2002 to Jun 2003

12.44% 2.33% 1.32%

Jun 2001 to Jun 2002

-21.36% -20.76% 2.21%

Close on Jul 17, 2000 to Jul 16, 2021

Returns S&P 500

Total Returns

-83.32% 324.2%

Drawdown

92.93% 55.2%

Annualized Returns

-8.18% 7.12%

Daily Beta

1.174

Annualized Alpha

-16.27%

Table 1

This table show us numerically how volatile and disappointing these results would be. The screen manages to bounce around, making it unshortable, and still produce an 83% loss. Its negative 16% alpha makes sense as it is as volatile as the stock market – just in the wrong direction.

Monthly Returns

Percentile

Returns (Monthly) S&P 500

1%

-28.74% -14.86%

3%

-22.58% -9.32%

5%

-20.66% -8.27%

10%

-14.23% -4.77%

25%

-5.12% -1.85%

50%

0% 1.29%

75%

4.89% 3.33%

90%

14.14% 5.61%

95%

17.42% 6.73%

97%

24.36% 9.32%

99%

33.98% 11.4%

Monthly, Up to Dec 31

Returns S&P 500

Average
Annualized, Geometric

-6.77% 7.43%

Median
Annualized

0% 16.59%

Standard Deviation

11.27% 4.64%

Monthly CVaR 5%

25.45% 11.32%

Monthly Beta

1.53

Monthly Sharpe Ratio

-0.05% 0.13%

Monthly Risk Free

0.001%

% Outperforms

40.24%

Table 2

This table shows how often it outperforms (not often), how much you’d lose if you had a 5% month (about a quarter of the portfolio), the standard deviation of the months vs the index (more than double), and more. This screen manages to make nothing on its median month!

Best Months Returns(Monthly) S&P 500
Oct 2001 37.22% 2.88%
Jan 2001 37.19% 2.16%
Sep 2013 31.03% 3%
Nov 2002 30.44% 4.88%

Worst Months Returns(Monthly) S&P 500
Sep 2001 -37.01% -10.07%
Oct 2008 -29.66% -15.82%
Aug 2019 -27.89% -4.17%
Jun 2002 -27.25% -8.13%

Positions Only includes positions that were active anytime between Jul 16, 2000 to Jul 16, 2021

Percentile

Returns per position Days held shortest to longest Market Cap low to high

1%

-64.07% 11 1.015b

2.5%

-53.21% 53 1.056b

5%

-43.25% 53 1.087b

10%

-31.19% 61 1.175b

25%

-13.95% 62 1.459b

50%

2.61% 62 2.411b

75%

23.29% 63 5.494b

90%

43.92% 65 11.71b

95%

60.59% 65 22.92b

97.5%

87.2% 65 45.23b

99%

110.2% 123 88.36b

Trade Summary

Results

Average across positions

6.12%

Median across positions

2.61%

Standard Deviation

33.48%

Average Holding Period

2 months, 29 days

Average Number Held

6

Above 0

53%

Above benchmark

50%

Worst Position

NBGRY(-97.89%)

Best Position

TCKRF(224%)

* Returns are yield adjusted

The positions table (above) tells us that the average position made money, despite the portfolio being a loss making machine. This is probably because when the market crashed, this screen triggered with more positions at a time. The market also does really well after market crashes, meaning that all of those positions did well. I would be less convinced of this as a diagnostic flag during down markets, as a result.

The Takeaway

Avoid stocks that have two bad days in a row, except if there are too many of them at the same time. In doing so, you’ll avoid a messy, volatile ride to a loss.

henry
henry
Henry Crutcher is an avid family guy, board gamer (think Settlers of Catan, Puerto Rico, etc), computer nut, and all around geek. Hailing from Louisville, KY, he has noticed that the weather in Louisville is remarkably similar to the weather in Atlanta, GA despite the 407 miles that separate them. He has two daughters, one cat, and lots of trees. He loves the Miles Vorkosigan series from Lois McMaster Bujold, for its mix of SF, comedy and insight into how people work. He also comsumes more than his fair share of cheesy business/economics books, such as The Ascent of Money by Niall Ferguson, or Farewell to Alms, by Gregory Clark.

Leave a Reply

Your email address will not be published.

This site uses Akismet to reduce spam. Learn how your comment data is processed.