Quant 101 Learn2Quant HK, 14 September Vinesh Jha CEO, ExtractAlpha

Transcription

1 Quant 101 Learn2Quant HK, 14 September 2017 Vinesh Jha CEO, ExtractAlpha

2 Data-driven investment is the future ExtractAlpha confidential. Do not copy or distribute. 2

3 Agenda Ø Quick intro to ExtractAlpha Ø Quant vs discretionary Ø Types of quants Ø Recent trends. Flows, quantamental Ø The quant process Ø Alpha, portfolio construction and risk management, trading, portfolio analysis ØPitfalls and biases Ø Tools of the trade ExtractAlpha confidential. Do not copy or distribute. 3

4 Quick intro to ExtractAlpha Ø FinTech independent research provider founded 2013 Ø Team is ex-starmine and Morgan Stanley prop trading (PDT) Ø We build models and data sets designed for institutional investors, to help them take advantage of alternative data Ø To do this, we scout interesting data sets, do lots of quant factor analysis on them, and build them into easy-to-ingest models (rankings) which we then license to funds Ø Currently 4 model products, clients include some of the largest global quant funds ExtractAlpha confidential. Do not copy or distribute. 4

5 Defining quant Ø Quantitative fund management is data driven and as systematic as possible Ø Increased repeatability, reduced emotion and mistakes Ø Human judgment is not removed, but rather built into the process instead of being built into each position Ø Anecdotes don t constitute evidence Ø Quant processes allow a manager to scale a portfolio take many low-conviction bets rather than a handful of higher-conviction bets Ø So diversification is key Ø Beyond this, there is no single definition of quant, and there are many different flavors ExtractAlpha confidential. Do not copy or distribute. 5

6 Quant vs Discretionary portfolio management Quant Discretionary Breadth Wide Narrow Depth Shallower Deeper Scales with Computing power Manpower Constraints Liquidity, data, expertise Attention, speed, expertise Replicability High Low Integration of risk management High Low ExtractAlpha confidential. Do not copy or distribute. 6

7 Some flavors of quant Ø Long horizon equity Ø Risk premia, smart beta, accounting anomalies, value momentum Ø Long-only or market neutral Ø Low Sharpe / high capacity Ø Stat arb equity Ø More technically driven but can leverage earnings, news, etc. Ø Typically market neutral Ø Mid Sharpe / mid capacity Ø HFT Ø No overnight risk, smaller positions, high Sharpe / low capacity Ø Quant macro Ø Multiple different horizons, typically technical in nature Ø Fewer instruments to trade and less homogeneity -> more time series models ExtractAlpha confidential. Do not copy or distribute. 7

8 Recent trends Ø Quant funds (hedge & long only) have experienced strong inflows Ø Smart Beta and Risk Premia products have proliferated (ETFs, structured products) Ø Quantamental investment has become more popular, but means many different things: Ø Integrating new data sources into a discretionary process Ø Using quant models or factors for screening portfolios Ø Using risk models and quant tools for portfolio management and performance attribution ExtractAlpha confidential. Do not copy or distribute. 8

9 The quant process Ø Finding alpha in traditional or non-traditional data sets Ø The quant process allows for straightforward integration of unusual datasets Ø Backtesting covered in Part 2 Ø What did I know, when, and would it have helped me? How robust and scalable is it? Ø If you can t backtest something, you may not be able to use it in a truly quant process! Ø Portfolio construction, risk management, and attribution Ø Using a risk model to understand, measure, and possibly control risk Ø An optimization process can create portfolios which control for delta, risk, turnover, position sizes, etc. Ø Trading Ø T cost models and awareness; trading algorithms ExtractAlpha confidential. Do not copy or distribute. 9

10 Tools of the trade Ø Data analysis Ø R, Python, Matlab Ø Data sources Ø Need basics like market data, fundamentals Ø Backtest engine covered in Part 2 Ø Optimizers and risk models Ø BARRA, Northfield, Axioma, etc Ø Transaction cost models Ø ITG, Northfield, etc ExtractAlpha confidential. Do not copy or distribute. 10

11 Pitfalls Ø Intuition: butter in Bangladesh Ø Sample size matters! Ø Lookahead and survivorship biases Ø Timing assumptions Ø Universe construction Ø In and out of sample testing Ø Realistic assumptions Ø Liquidity, transaction costs, short sales Ø Overfitting! Ø Parsimony is key ExtractAlpha confidential. Do not copy or distribute. 11

12 Best practices Ø Practice, practice, practice Ø Eyeballing and examining the data Ø Hypothesis testing. Portfolios, event studies Ø Focusing on turnover, risk attributes, and robustness Ø Being honest and disciplined with in sample and out of sample testing ExtractAlpha confidential. Do not copy or distribute. 12

13 Bibliography Ø Deutsche Bank Seven Sins of Quantitative Investing Ø R. Narang Inside the Black Box: The Simple Truth About Quantitative and High Frequency Trading Ø V. Jha Being Systematic article series ExtractAlpha confidential. Do not copy or distribute. 13