- Barclay Group performance CTA indexes (1996-2001): Systematic Traders beats Discretionary Traders consistently and by a wide margin.
- The key beauty of a mechanical trading system is it removes the irrational human emotional element.
- Pioneers of quantitative trading: William Gann (How to Trade in Commodities), Richard Donchian, Welles Wilder, Thomas DeMark (The New Science of Technical trading, New Market Timing Techniques). Recent traders (Monroe Trout, John Henry, Ken Griffin, Jim Simons).
- Rosenberg, Reid, Lanstein study (1985): mechanical long/ short buying losers selling winners mean reversion strategy produced 1.1% per month with profits in 43 out of 46 months. The strategy was optimized to eliminate any cap, value, or industry bets.
- Shiller believes the markets are 5-13x more volatile than is justified by pure fundamentals (cash flows). The excess volatility is attributed to investor psychology.
- “Prices move first and fundamentals follow”. Fundamentals drive performance in the long-term. Bubbles and burst do not move in a straight line—there are spectacular rises and devastating declines in the shorter-term within the secular trend. Irrational, emotion-driven volatility is hard to exploit using fundamental analysis.
- Need only a slight edge. 52% daily accuracy results in 72% positive years; 55% accuracy results in 94% positive years.
- Grossman and Stiglitz: markets cannot be long-term strong form efficient—efficiency breeds less research which creates inefficiency.
- Trading strategies should deteriorate (exponentially) over time—competition and “half life” concept. Use best fit regression line of equity curve and standard error bands to determine when the strategy should be cutoff.
- As the number of trials increase and the bet size decreases, the distribution of returns narrows. The potential for large returns (without leverage) also decreases.
- Diversification is most effective when correlations are lower than around 0.5. Correlations of 0.75 have limited diversification benefits.
- K-Ratio improves upon weakness of Sharpe Ratio, which does not factor in autocorrelation (timing of returns). Calculation: slope of equity curve regression / (standard error of slope * # of observations).
- Treynor/ Mazuy alpha measure: use t-stat of beta statistic to squared benchmark returns.
- Need for optimization: no consensus. Some feel scrupulously fitting data yields unrealistic expectations; others feel it allows for fine-tuning. Best to optimize to test parameter stability and then stick with parameters that perform well over the long-term.
- If volatilities aren’t balanced when trading more than one market (asset class), one volatile market can dominate performance. Scale dollar amounts or notional values to match expected volatilities within and among markets.
- Study results indicate markets tend to trend roughly 60% of the time (50% is random walk baseline), thus trend-following strategies may be profitable.
- Trend-following strategies (buy on strength, sell on weakness; misses tops and bottoms; can get whipsawed in range-bound sideway markets):
- Channel Breakout: enter long (short) on highest (lowest) 40-day close; exit on lowest (highest) 20-day close. Very profitable strategy historically but has significantly deteriorated.
- Dual Moving Average Crossover: calculate a 10-day and a 40-day moving average. Enter long (short) when 10-day MA crosses above (below) 40-day MA; exit on opposite signal. Highly profitable strategy. High correlation with Channel Breakout strategy.
- Momentum: enter long (short) when if close/ MA/ oscillator is higher (lower) than 20 days ago; exit on opposite signal. Use 80 days for longer-term momentum. Simplest strategy.
- Volatility Breakout: enter long (short) when daily price change is greater (less) than 2 standard deviations of 100-day trailing daily volatility. Good performance in stocks but not in futures. Moderate 0.5 correlation with trend-following strategies.
- Reversion/Oscillator strategies (looks for reversals and exhaustions; use when signals are short, indicating trend reversal not trend extension):
- Stochastics: enter long (short) on 14-day %K stochastic below 20 (above 80) and then crosses above 20 (below 80) and cross above (below) fast %D stochastic; exit on opposite signal. Unprofitable strategy historically but looks more promising on recent data.
- Relative Strength Index (RSI): enter long (short) when market bottoms (tops) below 30 (above 70); exit on opposite signal. Most popular overbought/oversold oscillator. Very poor performance historically on stocks and futures.
- Moving Average Convergence/ Divergence: MACD line constructed from difference between 12-day and 26-day exponential MA’s; signal line constructed from 9-day MA of MACD line. Enter long (short) when MACD crosses above (below) signal line; exit on opposite signal. Hybrid trend-following and trend-reversal strategy—buy on strength after declines, sell on weakness after rallies. Strong performance in futures markets.
- Remember to consider deterioration of strategies. Trend-reversal may have worked well historically, but exhaustion strategies look more promising recently. Technology and indices tend to trend, but most stocks do not trend.
- Relative value markets tend to mean revert, so trend-following strategies are not profitable.
- Lack of a trading range for a stock causes more volatility. Tech stocks during boom—traders were reluctant to step in due to no experience with historical price range.
- Market-timing strategies incorporating falling interest rates have beat buy-and-hold of S&P 500, with less risk (time in the market). Lower interest rates benefit stocks: 1) cheapens financing for projects; 2) more attractive than low bond yields; 3) home owner refinancings increases investable funds. Mondays used to be worse performing than Fridays, but the Monday effect has largely disappeared recently. The monthly effect is still present—beginning, middle and end of months have higher returns, due to people automatically investing on pay date in retirement plans.
- VIX strategy: enter long (short) when VIX is +2 (-2) standard deviations above (below) 120-day average. Measures complacency and fear. Beats buy-and-hold with only half the time invested in the market.
- No money management technique can turn an unprofitable strategy into a profitable strategy. Poor money management technique can turn a profitable strategy into an unprofitable strategy.
- Leverage/ volatility magnifies returns asymmetry and creates a skewed distribution, affecting the probability distribution of ending wealth. Seek to maximize the median ending wealth, not the mean ending wealth!
- Higher reward-to-risk strategies can accommodate higher leverage; lower reward-to-risk strategies are unable to overcome large losses.
- Optimal bet size is parabolic—median ending wealth increases with bet size only up to a certain point, then decreases. Reference Kelly’s Formula for optimal bet size.
- Markowitz showed that maximizing the median ending wealth is maximizing mean logarithmic (geometric) return. Optimal leverage = (excess) return / variance. Need to continuously adjust to maintain optimal leverage when capital grows or shrinks.
- Most people are risk-averse; research shows most are trading at only one-half to one-third of optimal leverage.
Finished: 28-Aug-2008