Back to Community
Campbell, Hilscher, Szilagyi (CHS) Model - Probability of corporate failure - Update Version

Here is the current version, where I corrected the error in the computing of Sigma.

Loading notebook preview...
Notebook previews are currently unavailable.
7 responses

This is the list of the companies, that according to the CMS model could go bankrupt (pfd > 50%) in one-year horizon (as of July, 2015):

LOCM  LOCAL CORP                               100.00%  
LOOK  LOOKSMART LTD                            99.91%  
USEG  U S ENERGY CORP                          99.91%  
DXM   DEX MEDIA INC                            99.50%  
ASTI  ASCENT SOLAR TECHNOLOGIES INC            99.47%  
RLJE  RLJ ENTERTAINMENT INC                    98.43%  
VPCO  VAPOR CORP                               97.99%  
XGTI  XG TECHNOLOGY INC                        94.93%  
VGGL  VIGGLE INC                               91.37%  
VRNG  VRINGO INC                               89.47%  
INPH  INTERPHASE CORP                          88.03%  
HERO  HERCULES OFFSHORE INC                    86.01%  
WRES  WARREN RESOURCES INC                     82.82%  
PHMD  PHOTOMEDEX INC                           82.47%  
PRSN  PERSEON CORP                             78.96%  
JOEZ  JOE'S JEANS INC                          76.87%  
SWSH  SWISHER HYGIENE INC                      74.22%  
CLRX  COLLABRX INC                             62.89%  
SPEX  SPHERIX INC                              58.80%  
ADAT  AUTHENTIDATE HOLDING CORP                57.40%  
ESCR  ESCALERA RESOURCES CO                    54.40%  
NETE  NET ELEMENT INC                          53.01%  

I'm trying to embedd the CHS model into my trading strategy, but I've some difficulties in computing the three-months standard deviation of the daily returns: there is not get_pricing(.) function in Quantopian Algorithms and history(.) doesn't work in before_trading_start(.).

Why don't add the function get_pricing(.) also in Quantopian Algorithms?
Alternatively why don't add a measure of stock volatility in the Fundamentals Data (for example Beta)?

Errors...


NameError Traceback (most recent call last)
in ()
1 symbols = fundamental_data.minor_axis
----> 2 price_history = get_pricing(symbols, fields='close_price', start_date=t1, end_date=t0)
3 price_history

NameError: name 't1' is not defined

Errors...

NameError Traceback (most recent call last)
in ()
2 zero_centered_returns = returns_daily.sub(returns_daily.mean())
3 n = len(zero_centered_returns)
----> 4 sigma = np.sqrt(252.0 * (1.0/(n-1.0)) * zero_centered_retuns.pow(2).sum())
5 sigma

NameError: name 'zero_centered_retuns' is not defined

I notice in the code that you use the (SPY # of shares outstanding) *( the price of SPY * 10) to derive the market value of the index. This would not survive an audit. It would be better to reference the S&P 500 index of the constituent stocks, then make a new calculation of all of those listed securities' market caps summed to arrive at the market value of the index. As of this writing, there are 504 listed securities which comprise the S&P 500 index. SPX is the ticker for the index, but i do not know what data about SPX is available to us.

Your method still comes very very close to the actual, and your code mostly works. It could use a lot more descriptions so that other people can use it (it took me almost an hour to make it work up to the point of the summary stats).

I was able to use your code to calculate the parts of the LPFD for a single security (by filtering heavily in the filter section of your code to isolate 10 securities) in the quantopian notebook then compare that against my own data in excel. Very good work.

If I was a much stronger coder (I am a Master of Science in Finance student, studying for CFA Level 1 currently), then I would modify this to create a list of the "rising losers" ranked by fastest approach (over some measure of time, perhaps 9-12 months?) to 100% and "stable winners" those stocks near 0% failure with the least amount of variance. I would be able to reference my long securities against the stable winners to see if I should stay long (this could be useful for mid to small cap stocks with little analyst coverage) and if I was hunting for new short ideas I could reference the "rising losers" for more advanced fundamental analysis. By itself, I do not think the CHS model could be traded upon, but it is a fantastic tool for due diligence.

My poor coding skills (I have been studying python for 2 months now) could not fix the notebook to go all the way to the finish line, but I really appreciate you doing the work to get it as far as you have. Perhaps another fundamentalist like myself but with coding skills could fix this awesome tool?

I am having trouble running through the notebook without errors. The time variables t1, t0, t4 etc. don't seem to setting correctly. Does anyone have an updated version?

This is impressive, thank you for sharing!