Tag: quant

Practical Momentum – Conclusion

Recap

We began the exploration of a practical way to execute momentum using derivatives. We found that:

  1. A lookback period of one year works best (Part I)
  2. Because of survivorship bias, long-short underperforms long-only (Part II)
  3. Hedging with single-name put options doesn’t work(Part III)
  4. Larger long-only portfolios have smaller drawdowns and better performance than smaller long-only portfolios (Part III)

Conclusion

The way things stand, Momentum is best executed using a broad basket of stocks. There is no mechanical way to maintain a momentum driven derivative portfolio. You can explore long-only equity momentum here.

Categories , Tags ,

Practical Momentum Part III – Hedging

Introduction

In Part II of our Practical Momentum series, we saw how adding a volatility adjustment significantly improved portfolio returns. However, we were left with a nagging observation that long-only returns were much higher than long-short returns. The problem with a long-only futures portfolio is that draw-downs can wipe you out. But what if we hedged the portfolio?

You can hedge a portfolio in two ways: (a) buy individual put options, and (b) calculate the beta of the portfolio and short an appropriate multiple of NIFTY futures. The problem with option (b) is that it will not protect you against idiosyncratic risk. For example, say you are long a pharma stock and the USFDA issues an import alert, the stock will tank irrespective of the NIFTY. So for the purposes of this simulation, we will try option (a)

Hedged Long-Only Momentum

With 5 long-futures hedged with long put-options below the purchase price:

black line shows long-only; red shows hedged long-only

hedged.momentum

A portfolio hedged with single-name put options performs poorly:

  • There is always a d between the option payout and the underlying
  • ?-decay eats away more of the option value than the protection it offers

Another way to make draw-downs shallower is to diversify. When we increased the number of stocks in our long-only equity momentum portfolio from 10 to 20, it reduced portfolio volatility and boosted returns. Here’s how a 10-count long-only momentum portfolio compares with the 5 from above:

black line shows a 5-item long-only portfolio returns; red shows 10
five10.momentum

Conclusion

The problem with leveraged momentum is that losses can wipe you out. Hedging it with single-name options doesn’t work. Are we stuck with unlevered momentum? We will explore this in the next post. Stay tuned!

Categories , Tags , ,

Practical Momentum, Part II – Volatility Adjustment

Introduction

Previously, we ran back-tests on long-only and long-short momentum algorithm over a couple of look-back periods. We found that (a) momentum with a one-year look-back period out-performed one with a 100-day look-back, and (b) a long-only portfolio significantly out-performed a long-short portfolio. We hypothesize that this is probably because the universe of stocks that we are forced to consider was heavily plucked. But what if we added a volatility metric into the mix to smooth out draw-downs?

Long-only Momentum

First, lets take a look at the long-only portfolio; both with a one-year look-back:

The red line is the volatility adjusted momentum; black is naive momentum; and green is buy & hold Nifty
long-only-momentum.volatility.2005-2010

long-only-momentum.volatility.2011-2014.

By year:

long-only-momentum2
Adding volatility into the mix did nothing to drawdowns but boosted returns considerably – with volatility adjusted momentum out-performing the naive version in 7 out of 10 years.

Long-short Momentum

Long-short ended up under-performing long-only once again:

The red line is long-short momentum; black is long-only momentum; and green is buy & hold Nifty
long-short-momentum.volatility.2005-2010

long-short-momentum.volatility.2011-2014

By year:

long-short-momentum2

Conclusion

Over the long run, long-only momentum with volatility adjustment outperformed the long-short version. However, while long-only tanked with the rest of the market in 2008, long-short was in the green. So if you are one of those guys who ask “how did this strategy perform in 2008?” Well, it performed pretty well. But would you have stuck by it when it got shellacked in 2013?

The problem with steep drawdowns is that it makes implementing the strategy with derivatives or leverage difficult. Margin calls might force you to abandon the strategy just before it turns. Next, we will explore a hedged long-only momentum strategy. Stay tuned!

Categories , Tags ,

Practical Momentum, Part I

Introduction

Momentum effects are one of the premier anomalies in the market and we have been running an equity long-only momentum strategy since 2013 with returns of +64.44% vs. Nifty’s +27.25% so far. Given the success of long-only equity momentum, we were curious as to how a long-short version of it would perform in India given our unique constraints. And also investigate if its success could be replicated using derivatives.

Typically, academic research that discuss momentum tide over the difficulty involved in shorting stocks. In India, you can only short stocks through SLBS in quantities that are multiples of the lot-size. And only those stocks that are allowed in the F&O segment can be borrowed for selling short. In order to overcome these constraints, we restricted our universe of stocks to only those that have been in the F&O segment since Jan 2004. There are grand total of 97 stocks that fit this criteria.

The biggest problem with choosing such a restricted universe is survivorship bias. One can argue that the stocks that survived from 2004 through 2015 and had enough liquidity to be listed in F&O would have stronger long-term momentum than those that do not. If this is true, then it doesn’t make sense going short. We will see if this hypothesis is confirmed in our back-test.

Long-only Momentum

Typically, momentum strategies are run using a one-year look-back period. We wanted to check what kind of impact shorter look-back periods had on overall returns. The following results are for going long (equally weighted) the top 5 stocks in our universe at the beginning of every month and holding it for one month.

The red line is the one-year look-back momentum; black is 100-day look-back momentum; and green is buy & hold Nifty
long-only-momentum.2005-2010

long-only-momentum.2011-2014

By year:

long-only-momentum

Long-short Momentum

You would think that shorting “weak” stocks should give returns comparable to going long “strong” stocks. But that doesn’t seem to be the case. The short-portfolio was always a drag on performance and made returns more volatile.

long-short-momentum.2005-2010

long-short-momentum.2011-2014

Conclusion

A long-only momentum strategy with a one-year look-back beat the pants out of both the Nifty and the long-short strategy. This could be because the pool of stocks in F&O show strong survivorship bias. We will continue to investigate if the short portfolio can be made more efficient. Stay tuned!

Categories , Tags ,

Machine Learning Long-Short Trend Following

Introduction

Our previous post discussed how a simple SMA On/Off Switch based tactical algo can be enhanced by a volatility metric. We generated significant alpha by following a simple rule:

Go short if either or the volatility signal or the 50-DMA indicates a negative bias and long otherwise.

But what if we trained a machine on the same data and allowed it to decide when to go long and short?

Support Vector Machines

We fed an SVM our volatility metric and the percentage distance from 50-day SMA. A 5-year training set was used to predict the next year daily long/short. We will not delve into the details of how SVMs work, Wikipedia does a decent job introducing the concept.

Performance

To make it easier to compare, we plot the wealth-charts for the NIFTY and BANKNIFTY indices side-by-side.

The black line is the Machine Learning Long-Short Model and the blue line is buy-and-hold. NIFTY and BANKNIFTY since 2011:

nifty.machine.learning.2011

banknifty.machine.learning.2011

NIFTY and BANKNIFTY since 2013:

nifty.machine.learning.2013

banknifty.machine.learning.2013

Cumulative Returns

Buy-and-hold has two big advantages over a trading strategy: transaction costs and tax treatment. Here is how the different strategies compare with buy and hold:

NIFTY SVM

BANKNIFTY SVM

It appears that the ML(V + 50-DMA) Long Short strategy works better on the BankNifty than on the Nifty. The out-performance of the ML model on the BankNifty more than compensates for transaction costs and taxation.

Conclusion

The ML model outperformed the NIFTY by an average of 12% in the last 4-years and the BANKNIFTY by 94% in the same period. The out-performance on the BANKNIFTY is considerable enough to warrant further exploration.

Categories , Tags , , ,