June 15th, 2022

Boosted Insights Product Update: Equity Explorer Offers Unprecedented Explainability For Institutional Investors

Written By: Joshua Pantony

One of the biggest challenges with incorporating AI into traditional institutional investing is the lack of transparency behind the machine’s decision-making process. Portfolio managers and analysts have to be able to trust (and explain) all machine recommendations. Here at Boosted.ai, we believe investing begins and ends with humans. The context around the machine’s outputs is just as important as the outputs themselves, which is why we invest heavily into explaining the machine’s thinking behind every stock pick. As part of our commitment to our institutional investor clients to always push to make AI more explainable, we’re excited to announce the release of Equity Explorer. 

What is Equity Explorer? 

Equity Explorer is the first research tool of its kind that lets institutional investors fully lean into their capital markets expertise. In the past, our machine learning platform Boosted Insights explained the stock picks by highlighting top features that drove the ratings, but now we can zoom in even further. Equity Explorer breaks down the different patterns – or combinations of features – that the machine evaluated to give users a comprehensive understanding of what went into a stock pick. 

Equity Explorer benefits: 

  1. Stock Screening – Discover fundamental and technical variables that generate buy/sell signals, which may go unnoticed by the human eye. 
  2. Monitor Market Changes – Be alerted of shifts in market conditions through monitoring changes in pattern performance. 
  3. Stock Discovery – Easily find other stocks that follow a pattern of interest and discover sector/industry trends. 

How Equity Explorer helps asset managers with stock screening

In our new Equity Explorer page, asset managers can see the different patterns the machine evaluated and the buy/sell signals they generate. Note, there are thousands of different patterns evaluated in the machine learning process, but we show you the unique clusters of similar patterns to make the information more digestible. 

The patterns surfaced are all able to be explored in depth. Users can see how the pattern performed during the training and live period. We measure performance by Excess Return and Hit Rate, which we define as: 

  • Excess Return – The pattern’s average return compared to the benchmark.
  • Hit Rate – The pattern’s rate of positive return. 

Although there is great depth to Equity Explorer, we also know that asset managers want to be able to see, at a glance, exactly how any stock in their universe is performing within their models. We use Equity Explorer to surface high level insights as well, like the hit rate batting and slugging average and the excess return batting and slugging averages. 

  • Hit Rate Batting Average – A measure of how accurate the patterns for this security are, where we predict the hit rate should be above 50% if it has been in live trading that counts as a hit.  Similarly, if we predict the hit rate should be below 50% and it has been in live trading that is also a hit.  The total number of hits are divided by the number of patterns.  Minimum score: 0%, Maximum score 100%.
  • Excess Return Batting Average – A measure of how accurate the patterns for this security are, where we predict the excess return should be above 0% if it has been in live trading that counts as a hit.  Similarly, if we predict the excess return should be below 0% and it has been in live trading that is also a hit.  The total number of hits are divided by the number of patterns.  Minimum score: 0%, Maximum score 100%.
  • Hit Rate Slugging Percentage – A measure of how accurate the patterns for this security are adjusted by the magnitude of the signal.  Using the same “hit” methodology as batting average, we also adjust for the strength of the signal for the pattern.  Hits are multiplied by the score for the pattern and then divided by the absolute sum of the scores for all patterns.  Minimum score: -100%, Maximum score 100% – we would want to see shorts with a negative Slugging Percentage and longs with a positive Slugging Percentage.
  • Excess Return Slugging Percentage – A measure of how accurate the patterns for this security are adjusted by the magnitude of the signal.  Using the same “hit” methodology as batting average, we also adjust for the strength of the signal for the pattern.  Hits are multiplied by the score for the pattern and then divided by the absolute sum of the scores for all patterns.  Minimum score: -100%, Maximum score 100% – we would want to see shorts with a negative Slugging Percentage and longs with a positive Slugging Percentage.

How Equity Explorer helps investment managers monitor market changes

One of the benefits of Equity Explorer is surfacing changes in market conditions by looking at pattern anomalies. Here are some things asset managers can look out for: 

  • Big changes with excess returns and hit rates between training and live period could be indicators that something changed in the market causing certain patterns to either be more favorable or unfavorable in picking stocks. It may be worth adjusting the portfolio’s exposure to some of these factors. 

  • Discrepancies with excess returns and buy/sell signals could mean that certain patterns that were indicators for buy or sell may no longer work in the current environment. When there is negative excess returns but the machine is indicating that is a buy, or vice versa, then it is a signal to adjust either the model or portfolio settings. 

To easily find some of these market changes, Equity Explorer has various sorting options to quickly surface these anomalies. By default, we sort by patterns the machine finds most interesting. Sorting by absolute delta for excess return or hit rate will surface patterns with the biggest discrepancies between the training and live period. 

How Equity Explorer helps asset managers through stock discovery

Equity Explorer showcases the different features that form any pattern and how they contribute to the overall buy/sell signal. This is where the investment manager’s capital markets expertise is critical – their teams assess the patterns and understand the features and interactions that impact their stock universe. A team may notice a particular pattern, say 60M CAPM Beta or EMA (50D) appears frequently with positive results. This is a signal for them to further study these results to see if their machine learning models can offer important insights. 

Clicking on the Securities tab, the investment manager can see a list of stocks that follow the pattern. Looking at the list, it can show multiple stocks that fall within a similar sector or industry, which shows there is a high exposure to this particular pattern. 

Equity Explorer (Advanced Mode)

One additional view to Equity Explorer is Advanced Mode. The most important additions to the page are Cluster Signals and Instance %

  • Cluster Signal – The patterns you see can be found in other patterns as well. Cluster Signal is the total buy/sell rating for all patterns that include this particular combination of features. 
  • Instance % – Instance % shows how many stocks follow this pattern and how often. 

Difference in Instance % between live and training periods is an important metric to look for as it directly relates to market fluctuations that made the pattern more or less relevant.

Takeaways

Given the volatility in the market today, it is more important than ever to get ahead of market changes and understand the drivers behind the movements. Equity Explorer advances the way Institutional Investors work alongside AI. Whether it is for stock screening, market awareness, or research, this breakthrough research tool will give investors greater confidence in incorporating AI into their workflow and their investment decisions. If you want to learn more, please reach out to us here for a demo! 

Leave a Reply

Back to Blog