Is it more difficult to maintain a value-weighted or an equal-weighted portfolio?
What does it mean to maintain a (zero-investment) portfolio when there are differential rates of return?
Should you use rebalanced zero-investment portfolios, or actual portfolio returns?
“Easy” question in a perfect capital market without transaction costs
Good Weights?
Use index (futures) as cheap hedging/market timing?
Can we use ETFs as cheap inputs?
Can we trade events?
Are non-investment like ``shorts’’ (holes)?
How should you handle outlier rates of return in the backtests?
Should you worry about data errors?
The gorilla in the room is that we do not know if our historical behavior is representative of the future.
Did we just find spurious correlations from trying too many signals?
Are there factors that have not reared their heads? That could wipe us out?
Is the experience from the last x years still even relevant or has the world changed?
How many years should we/you choose for data evals?
As X-costs have come down, what is their effect then vs. now? How does it influence backtests?
What do other investors and funds do?
What would be an expected performance pattern on strategies?
It is not clear whether these are inefficiencies or risk factors. That is, it is not known whether the historical ability to earn higher rates of return in them was due to stupidity by the rest of the market or due to a risk that other investors really disliked. In either case, from 1963-2018, these helped investors earn positive average rates of return on zero-investment portfolios.
## 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
## 21.6 13.2 -82.2 6.0 7.4 1.4 7.9 1.7 20.7 -20.4 5.0 9.3
mean(umd)
## [1] -0.7
Momentum has worked most of the time, but seems to blow up every once in a while (see 2009!)
Has not worked for last 12 years
Data is from Ken French’s website. Look at 2009!
Go long last month’s losers
Go short last month’s winners
Could be due to transaction costs and due to stale prices
## 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
## -14.7 0.8 -9.2 -5.3 -8.5 9.7 1.5 -1.6 -9.6 22.9 -13.9 -9.4
This is supposed to give you a positive performance, but since 2007
print(mean(hml))
## [1] -3.108
Value has worked for a long time, but not for a while
Has not worked for last 12 years
Value was Buffett’s main strategy (before he moved into private companies)
Value was based on a Dodd-Graham book from 1934
## 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018
## -7.19 3.20 9.23 13.80 -6.01 -1.15 7.27 -8.08 -4.05 6.60 -4.77 -3.32
This is supposed to give you a positive performance, but since 2007
mean(smb)
## [1] 0.4608
Was not described until around 2013
Recently added by Fama-French to prominent factor model
RMW (robust minus weak operating profitability)
Related to (net) issuing activity
Was not described until around 2013
Recently added by Fama-French to prominent factor model
Investment (Conservative minus aggressive)
Not a monthly factor in the ordinary sense
Market factor rate of return each January only was
## 2007 2008 2009 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019
## 1.4 -6.4 -8.1 -3.4 2.0 5.0 6.0 -3.3 -3.1 -5.8 1.9 5.6 8.4
mean(jan)
## [1] 0.01538
Note that unlike the previous numbers, this is monthly rate of return, not an annual one.
Can be exploited better with reversal stocks and just the first half of January
Given all these problems, should you just switch over to qualitative investments?
This is just stupid.
It’s like arguing that because we have evidence that falling 10 meters will kill you, we are better off jumping out from 100 meters
If you do qualitative research, you can use it as a quantitative input signal!
Transaction Cost Analysis
Turnover analysis
Use of Collateral (shorting costs)
Year-by-Year Performance
Risk analysis (idiosyncratic and factor risk)
Sharpe Ratios
Longest Time To Recover
Longest Negative Stretch
Worst Performance