(866)-482-2177

sales@alphatheory.com

REQUEST A DEMO

SYSTEM REQUIREMENTS


Please note the following System Requirements. Further, please limit the number of open applications (particularly price streaming applications) while logged in to Alpha Theory™.


Recommended System Specifications
Processor: Dual Core or Quad-Core 2.4GHz or faster
RAM: 4GB+
Browser: Google Chrome 30+
Screen Resolution: 1280 x 1024 or greater
Internet Access: Business Class High-Speed


Minimum System Requirements
Processor: Intel Pentium-M 2.0Ghz or equivalent
RAM: 2GB+
Browser: Google Chrome, Mozilla Firefox, Internet Explorer 9+ (without Compatibility View), Safari
Screen Resolution: 1024 x 768 or greater
Internet Access: High-Speed

Alpha Theory Blog - News and Insights

December 29, 2020

Optimizing Usage for Optimal Returns (Part 2) - Position Level Analysis

 

This article was co-written by Billy Armfield, Data Scientist of Alpha Theory, and Cameron Hight, CEO of Alpha Theory.

 

In our last edition, “Optimizing Usage for Optimal Returns”, we explored the relationship between forecast freshness and portfolio coverage on one-day forward returns at the fund level. Freshness and performance showed a nearly monotonic relationship. Coverage, on the other hand, had a more parabolic shape with the lowest performance coming from the middle of the range. To explore these relationships further in this edition, we are investigating data at the ticker level instead of the fund level. We ask:

1. Does coverage show more predictive power at the ticker level than it did at the fund level?

2. Is freshness correlated with a performance at the ticker level?

 

Coverage

Ten years of data on Alpha theory clients have shown that process-oriented investing yields higher returns. A large part of the process centers around entering and updating scenarios forecasting future stock prices, and the probability of them reaching that price. Implicit in this philosophy is the idea that if you are going to make an investment, it should be supported by research. As part of our exploration of coverage and freshness at the ticker level, we regressed coverage against a one-day forward price change. In this case, coverage is treated as a binary variable based on whether a forecast has been made on the position or not.

 

B1

 

Unlike the results of measuring coverage by the fund, when measured at the ticker level, there is a distinct positive relationship with one-day forward price change. Simply stated, positions with price targets are more likely to outperform positions without price targets. 

 

Freshness

There is more than one way to bake a Christmas cookie, so we first examined which measure of freshness has a stronger relationship with one-day forward price change. We examined two variations. The first measures freshness in terms of the number of days since a forecast was last updated (DSLU). The second treats it as a binary feature, where its value is 1 if price targets were updated in the last ninety days, and zero otherwise. We regressed both features against a one-day forward price change for all positions in Alpha Theory’s historical database. The results for the DSLU method can be found in figure 1, and the results for the binary method in figure 2.

 

Figure 1 (Days Since Last Update)

B2

 

Figure 2 (Binary: Updated in last 90 days)

B3

 

The number of days since the last update has a negative coefficient and t-statistic, which makes sense, given that one might reasonably expect a lower degree of certainty of positive future returns when forecasts are out of date. While intuitive, its lower t-statistic and higher p-value means this relationship deserves further investigation before drawing any conclusions. The binary feature, however, has more conclusive results. Having coverage no older than ninety days does have a positive relationship with the one-day forward price change, with a higher t-statistic and lower p-value.

 

The importance of creating price targets and keeping coverage fresh can be summarized by the annualized price change of fresh positions vs. stale and uncovered positions.

 

B4

 

Scattered, Fresh, and Covered

Based on the analysis above, it is well worth the time to create price targets and keep them fresh as they are both important predictors of future returns. When thinking through how your firm is going to improve in 2021, use these empirical proofs as evidence that the team should seek process as a proven way to improve returns.

 

November 30, 2020

Don't Be A Turkey: Optimizing Usage for Optimal Returns

 

This article was co-written by Billy Armfield, Data Scientist of Alpha Theory, and Cameron Hight, CEO of Alpha Theory.

 

Alpha Theory makes it easier for process-oriented investment managers to do their work, from uploading price targets through Excel Connect, to monitoring Optimal Position Size in the application. Previous analyses have shown a strong positive relationship between increased usage of the app and increased ROIC. Those results showed that for usage, like turkey, if some is good, then more is better. In that spirit, we decided to investigate different forms of usage to see what gives the greatest return on investment.

 

Two common ways in which we measure client usage are coverage (how much of the portfolio has research?) and freshness (how recently were your price targets updated?). In the same way we help our clients optimize exposure to their positions, we also want to make sure they are optimizing how they spend time engaging with Alpha Theory.

 

We ask:

1. Which element of usage has the strongest relationship with returns?

2. How much is enough?

 

Coverage

There are several ways to think about coverage. The first way we will examine is to think of it as the percentage of names in the book which has research. Simple enough. But this method can provide a false sense of security. Say, for example, you have a five-position portfolio weighted at 80%, 5%, 5%, 5%, 5%. You have price targets for your smaller positions, but not the big ones, giving you coverage of 80% under this definition. But are you really covered?

 

The second method of measuring coverage is by calculating the ratio of the gross exposure of the portfolio which has research, to the gross exposure of the portfolio as a whole. Both methods have their place, but we were interested to know which was more helpful as a metric to track and optimize.

 

An Experiment

In order to determine which measure of coverage is more helpful, we regressed both forms of coverage against one day forward returns for a sample of Alpha Theory client funds. The results for the first method are in figure 1, and the results for the second method are in figure 2.

 

Figure 1 (% of Positions with Price Targets): 

Picture1

 

Figure 2 (% of Exposure with Price Targets):

Picture2

 

The results for the exposure-based method (figure 2) show a higher coefficient, larger t-statistic, and lower p-value than method 1, giving us a high degree of confidence there is a stronger relationship between one day forward returns and the exposure-based method (p-value < 0.05 is statistically significant for our purposes). This measure of exposure is what we will use going forward.

 

Freshness

We should ask the same questions about freshness that we have for coverage. The first way we can think about freshness is to take the ratio of positions which have had their price targets updated in the last 90 days to all positions with price targets. The second is to take the ratio of gross exposure of positions with research updated less than 90 days to gross exposure of the portfolio.

 

Figure 3 (% of Positions with Price Targets that are Fresh):

Picture3

 

Figure 4 (% of Exposure with Price Targets that are Fresh):

Picture4

 

The exposure-based method is more predictive here as well, with a higher t-stat and lower p-value. These results make sense – small positions can be placeholders and get less attention. The positions with large exposure have targets subjected to greater diligence.

 

Comparing the Variables

We can now compare the results of regression against one day forward returns for the two variables under consideration.

 

Figure 5: Coverage

Picture5

Figure 6: Freshness

Picture6

 

In figures 5 and 6, we see that both coverage and freshness have t-statistics and p-values which are statistically significant, and both variables have positive coefficients. Of the elements of usage examined, higher coverage is most strongly related to higher one day forward returns, followed by freshness.

 

How much is enough?

Unlike at Thanksgiving dinner, we can estimate how much is enough of each. Focusing on Coverage and Freshness, we examined various thresholds to see how the average client performed above and below it. The 85% and 95% thresholds resulted in the greatest difference in annualized ROIC for clients above and below. More clients have been able to reach the upper echelons of coverage than freshness, which makes sense – over time, and given the reasonably low turnover, it isn’t unfathomable that you can make forecasts on > 95% of the exposure of your book.

 

Picture7

 

Keeping everything fresh, however, requires strict discipline, and an emphasis on process. But it appears as if the hard work pays off: the four funds with freshness over 95% have average annualized ROIC 6.6% higher than the average for all the other funds under consideration.

 

What is interesting is the persistent relationship for Coverage where higher Coverage always results in better returns but Freshness reverses course at the 70% level. This phenomenon is counterintuitive given the statistical relationship discovered in the regression. It is clear that more research is required and in next month’s blog post we’ll investigate this discovery and explore what can be learned from examining on a by-position basis instead of by-firm.