Nosegbe: The primary goal of active strategies is to outperform a relevant passive benchmark, meaning deliver better risk-adjusted return and alpha. We believe and have demonstrated that in-depth fundamental analysis of companies can be a core input to strategies that deliver superior alpha. We have developed proprietary technologies that automate the entire investment process—from analysis, to portfolio construction, to trade execution—with a clear objective: deliver better returns at lower risk. Our melding of fundamental analysis, quantitative methods, and technology is complementary, enabling the disciplined execution of our alpha-generating investment strategies without human biases.
Hortz: How did you go about developing your sentiment & fundamental Indicators?
Nosegbe: The development of our sentiment and fundamental indicators were driven by clear objectives: 1) the indicators and signals must deliver better return while reducing risk, 2) the achievement of better outcome must be statistically and investment significant, 3) the alpha value of the indicators must persist across market caps, across investment styles, across sectors, across time, and 4) the indicators must have the flexibility to be used by asset managers as alpha-generating input to their unique portfolio management strategies.
Hortz: What kind and level of predictive capability do your indicators exhibit?
Nosegbe: To demonstrate the predictive value of our indicators, we begin by asking the question: selected from a universe, for example the S&P 500, what is the probability of companies in your portfolio outperforming their peers over the next 20 days? For our Fundamental Grade “A” classifier, the probability of outperforming over 20 days is 51% to 53%. We then ask the question, what is the probability of outperforming in the long run? In the long run, rolling three-years, the probability of Fundamental Grade “A”’ companies (as a class) outperforming the benchmark is 99%, delivering 2.5% alpha on the average and beta less than one.
It is important to note that the analysis of financial statement data, like 10Ks, 10Qs and 8Ks, is the sole input to the algorithmic model that yielded our Fundamental Grade “A” outperformance. The model had no knowledge of stock price, analysts’ opinion, or other external data. This is our baseline. Performance can be further improved when other user-defined variables with predictive power (e.g., Valspresso’s sentiment data or dynamic sector allocations) are added to the baseline model.
Hortz: How did you design your artificial intelligence system to automate the entire investment process—from analysis to portfolio construction to trade execution? Why automate the entire process?
Ty Seddon: We chose to automate the entire process in order to align investment objectives, remove human bias, and respond more quickly to changing conditions. By automating the entire process, we feel like we have developed a more holistic and cohesive understanding that allows us to uncover new insights.
We apply different AI techniques to different classes of problems within the entire value chain of the investment process. They roughly fall into 3 categories: Interpretation, Prediction, and Design Optimization.
Fundamental analysis is primarily an Interpretation task. We believe that transparency is important, so we don’t use any black box machine learning algorithms for this Interpretation phase. We need to be able to explain why a company is rated as it is. To do this, we use an expert system to perform deep fundamental analysis on every company, every day.