Episode 153 ... Xiao Qiao (64:39)

Studied engineering, math, finance, statistics
[Is he first generation Chinese-American or did he grow up in China? Moved to America as a child? Fluent English, but non-native to my ear]
Liked Applied Mathematics
Both parents are engineers (Dad a mechanical engineer)
Did materials science engineering, didn't like working in a lab
Went to grad school for finance, Ph.D. program, University of Chicago
Probability theory, applied statistics, linear algebra
Asset pricing research
Preferred to go into industry rather than become an academic
Was teaching assistant for Eugene Fama in his asset pricing class
Fama demanded that students be both precise and concise
Lars Hansen taught him to take economics seriously, always apply economic thinking
Five years to complete a Ph.D. in finance at Chicago
Enjoyed playing blackjack in college
Blair Hull's personal assistant emailed him about market timing paper Hull was writing
Hit it off with Blair Hull, agreed to co-author paper on market timing
A Practitioner's Defense of Return Predictability
Team play for blackjack important, MIT team employed this method
Don't play if you don't have an edge
Bet table minimum when the house has the edge
Bet size hugely important once you have an edge
Most people who try to play to win don't have an edge
Academic research divided about market timing
Return predictability (can we forecast returns?)
Forecasting equity premium, combining return predictors
Modeling six month ahead excess returns
Examples of return predictors: PE ratio, CAKE ratio, variance risk premium, etc.
Forecasting horizon shortened to one month for next white paper
Return Predictability and Market-Timing: A One-Month Model
Reads a lot of academic research papers for his current job
Likes Frank Diebold's blog
[I guess he went to U Penn undergrad]
[Losing listeners in the late 30 minutes mark as talk strays into CAPM]
Try ideas yourself, don't just read about ideas
Simple linear regression goes a long way toward telling you if your hypothesis is any good
Lower frequency risk premiums: holding period in months, execution doesn't matter much
R, Python, Matlab code will be fast enough to implement
Higher frequency trading: holding period is intraday or days... execution becomes very important
Execution is where you make a significant fraction of your profits if your holding period is short
C++ necessary then ... code needs to be fast enough to trade in real time
Backtests are inevitably overfitted ... need to use out of sample data to test your model
Look at: Returns, Cumulative returns, Sharpe Ratios, Volatility, Cumulative Drawdown, Maximum Drawdown, Tail Correlation
Tail events, extreme events ... look how your model does during these
Trading costs, transaction costs, turnover, slippage are supremely important ... academics ignore these factors ("market microstructure")
Intermediaries are all-important, can make or break your returns ... practitioners are well aware of this, academics ignore it
Risk premiums important, but market impact, liquidity equally important
Rebalancing a portfolio during illiquid periods will affect returns
Frequency of rebalancing also important
His website
This Xiao Qiao should not be confused with Zhou Yu's wife, the anime girl
He is not on Twitter