Anatomy of a Momentum Strategy

#1
For the past couple of years, I have been trying to crack an ideal momentum strategy that would give the best risk/reward profile. I read a lot of books & research papers, conducted backtests, paper traded, as well as live traded. Below are my observations.

How to measure momentum - Classically ROC is used and it works just fine. One can also use regression (linear or exponential) to measure slope and fit.

Time frame & false signals - Shorter the timeframe to measure momentum, more the false signals. Momentum strategies work best when your look-back period is large. Most academic paper look at 12 months (leaving the latest month to account for short term reversal) and my observation is anything from 3-months to 12-months can work.

Control for volatility - extremely high volatility stocks can fall just as fast leading to earth-shattering drawdowns. 2 ways to control - avoid super high ADX stocks (>50) or avoid stocks with linear regression fit less than 0.6 or 0.7. Yes, you will miss out on some crazy returns but will control DD massively

Stocks to avoid - stocks which keep circuiting. They are a time bomb waiting to explode. They look great in backtests but try exiting a lower-circuiting stock and you'll know the problem.

When to enter - One can wait for a slight oversold(RSI(3) < 50) condition to make an entry or just enter on signal. My personal experience has been that trying time a momentum stock leads to more lost opportunities than downside protection.

When to exit - Stop losses are a suicide in a momentum strategy. This may sound controversial but let me explain. The returns curve of a momentum strategy by nature is a right-skewed. This means, traded long enough you will have a lot of trades with small losses and very few trades with extremely high gains (100%+). These few large gains are by nature high volatile stocks. Now if you apply stop loss, you'll exit these super profitable trades on which the entire strategy relies to make a lot of money.

Market Regime filter - It may sound silly but the number 1 requirement for momentum is 'momentum'. Unless the market is trending, you will not make money with the strategy. A classic filter is Nifty being above its 200-day MA (there are a few variations to this). I personally use 75-day MA since I am not comfortable with 30%+ DD. Typically, you'll be in market only 50% of the time over 5-10 year period.

Positions - Anything between 10 to 20.

Position sizing - The difference between equal size and risk parity positioning is not so great. This has been proved by research as well (Jagdeesh, Titman et. al.)

Measuring success - Sharpe ratio is not a good measure because of large positive standard deviation. An ideal measure can be CAR/MaxDD - a value above 1 should be great.

Strategies I have backtested/traded -
1. Stocks on the Move by Andreas Clenow
2. Quantitative Momentum by Wesley Gray
3. The 30-min stock trader by Lawrense Bensdorp

Would love to get more recommendation on books

If someone has traded on lower time-frames successfully, please do share ( I have been breaking my head for a long time on this)

P.S. Backtesting done on Amibroker. Survivorship Bias not accounted for.
 

stoch

Active Member
#3
Stock (and especially currency pair) prices generally have no memory. Read what is Markov property of time series. All momentum that is observed in prices happens in very short periods of time (1-3 sec) where slippage will on average eat your profit.
 

Similar threads