Category: Investing Insight

Investing insight to make you a better investor.

Sentiment Analysis of News Articles

Markets trend, until they don’t. Markets consolidate until they start trending again. Since the beginning of markets, traders have been trying to figure out when these turning points occur. At StockViz, we have approached this problem from a risk-management angle rather than a prediction angle. We assume that we cannot predict, but we can definitely prepare.

Risk-management, however, is not free. Whether you hedge or employ stop-losses, there is a cost involved. What if we can reduce this cost by employing market sentiment as an input to our risk-management models? One approach could be to widen stop-loss levels when the market is “extremely bearish” and tighten them when they get “extremely bullish.”

To get things started, we have setup a weekly sentiment roundup that looks at a few popular news sources ranked using different lexicons. The first step would be create a time series of these values to see how they relate to market returns.

Stay tuned!

Categories Tags

Trend Following vs. Trend Prediction, Part I

Traditional equity momentum strategies are variations of algos that try to figure out “trending” stocks so that they can be ranked to create a long/short portfolio. The key thing to remember is that these algos are following a trend, the prediction that a trending stock will continue to trend is implicit. However, using machine learning techniques, stocks can be ranked based on their predicted returns over a future time frame.

The simplest momentum strategy looks only at a price series. However, it quickly runs into problems when additional factors are overlaid on top of basic momentum. For example, you may want to filter out volatile stocks out the basket. You can do this either by setting a maximum volatility level or by weighing both momentum and volatility to arrive at a combined rank. Either approach leads to ad hoc decisions of cut off levels and the ratio with which to weigh each of those factors. Luckily, typical machine learning algorithms can work with multiple factors and weigh them based on the training set you supply.

The biggest drawback of using machine learning is that the larger the number of factors/features you use, the less explainable the resulting model becomes. As a trader, if you want to use any of these models, you should have a fairly good idea of what is going in, how the model is setup and what exactly is the model getting trained with.

As a first step in taking a crack at this, we have setup four machine learning algos. Two of them use SVRs and the other two use LR to train on data that is either return-series only or a combination of returns and volatility. You can have look at them here.

We will have more of these machine learning models out as we ramp up our understanding of these models. Stay tuned!

Categories Tags ,

Transaction Cost Analysis of a Momentum Strategy

Momentum strategies have been on a tear over the last few years and have generally out-performed pure-value strategies. When we compare momentum returns with mutual funds, the most common criticism we encounter is that mutual fund returns are after transaction costs whereas our “Theme” returns are before transaction costs.

Momo (Relative) v1.1 vs ABSL S&M Fund (Annualized returns are 85.50% and 35.12%, respectively.)

The challenge we face in showing post-cost returns is that we offer different brokerage slabs to different types of clients, making a one-cost-fits-all analysis impossible. However, we can show how different brokerage slabs impact returns.

A gross return of 83.82% translates to returns of 74.20%, 69.58% and 65.08% for brokerage slabs of 0.1%, 0.05% and 0% respectively (STT of 0.1% was assumed.) Momentum out-performs even after transactions costs.

Categories Tags ,

Should You Buy What Mutual Funds Buy? [Update]

Back in November-2015, we had concluded that it does not make sense following mutual fund entries and exits from individual stocks. Using an expanded data set of 260 funds, we still reach the same conclusion. Median returns over 10, 20 and 50 days on additions were 0.49%, 1.81% and 4.88% whereas exits clocked -0.01%, 0.90% and 3.27%.

Direct equity investors would do well to ignore what fund managers a doing.

Code and results are on github.

Categories , Tags

Can Beta Dispersion be used for Market-Timing?

The paper Beta Dispersion and Market-Timing (SSRN) argues that one can predict crashes by tracking the dispersion of betas of the constituents of an index. The intuition presented in the paper is that when beta dispersion is high, any shock to the high beta stocks could spill over to the low beta stocks and create a broad market correction.

Although the paper proceeds to present a back-test on the US S&P 500 index, there some questions that need to be answered before deploying this strategy:

  1. What is the performance if you remove 2000 and 2008 from the data? Perhaps most of the out-performance can be attributed to skipping these two periods purely due to chance?
  2. Are the results robust over different markets? Perhaps it is unique to the US?
  3. What happens if you change the look-back period of beta calculations? Perhaps it is being data-mined?
  4. What happens if the calculations are continuous rather than sampled at the end of the month? Perhaps its an end-of-the-month effect?

Unfortunately, we don’t have a robust data-set to put this theory to test. However, the chart of the cumulative returns of the NIFTY 100 index vs. the beta-dispersion of its components does not lead to the same conclusion made in the paper.

The code for this analysis is on github.

Categories , Tags