Top Stock Picks Based On AI: Beating the S&P 500 by 30%. Equity Curve Analysis and Portfolio Construction

Executive Summary

top stock picks
Figure 1. I Know First 5 Top Stock Picks Portfolio Performance VS S&P 500

In this algorithm performance evaluation report, we examine the performance of a sample portfolio constructed on 3 days algorithmic forecasts for US stocks universe. The back-testing is performed on trading data spanning from January 1 to August 2 2019, while compared to the S&P 500 performance as benchmark. The results suggest that the stock market predictions generated by the I Know First AI Algorithm provided solid basis for portfolio strategy and provided 20.63% return implying 3.89% premium over the S&P 500 return of 16.74%.

Note that the results presented in this report constitute an example of one of many possible strategies utilizing I Know First algorithmic forecast over specific time period and cannot be directly related to future stock market and individual assets’ financial performance. The following report provides an extensive explanation of our methodology and a detailed analysis of the performance metrics that we obtained during the evaluation.

About the I Know First Algorithm

stock market predictions

The I Know First self-learning algorithm analyzes, models, and provides stock market predictions for the capital markets, including stocks, bonds, currencies, commodities and interest rates. The algorithm is based on Artificial Intelligence (AI) and Machine Learning (ML) and incorporates elements of Artificial Neural Networks and Genetic Algorithms.

The system outputs the predicted trend as a number, positive or negative, along with a wave chart that predicts how the waves will overlap with the predicted trend. Consequently, the trader can decide which direction to trade, when to enter the trade, and when to exit the trade. The model is 100% empirical, based only on factual data, thereby avoiding any biases or emotions that may accompany human assumptions. I Know First’s model only involves the human factor in building the mathematical framework and providing the initial set of inputs and outputs to the system. The algorithm produces a forecast with a signal and a predictability indicator. The signal is the number in the middle of the box. The predictability is the number at the bottom of the box. At the top, a specific asset is identified. This format is consistent across all predictions.

Stock forecast example

Our algorithm provides two independent indicators for the index – signal and predictability.

The signal is the predicted strength and direction of the movement of the index. This is measured from –inf to +inf.

The predictability indicates our confidence in the signal. The predictability is a Pearson correlation coefficient relating to past algorithmic performance and actual market movement, measured from -1 to 1. You can find a detailed description of our heatmap here.

Portfolio Simulation Methodology

For the explanation purposes of portfolio simulation methodology that is used in this study we denote forecast time horizon by T, number of top assets among the available ones in a forecast heatmap used to comprise portfolio as N.

Initially, US$10.000 are used to open long and short positions (initial weights of positions are equal) in accordance with I Know First forecast generated by the algorithm on the eve of the first simulation date. The N assets for the positions are selected using predictability ranking. At the end of the first trading day the positions’ value is fixed and the new algorithmic forecast is analyzed. the following re-balancing technique is used:

  1. The amount available for re-balancing is equal to 1/T of the total portfolio value
  2. For the assets that appeared in N assets of the previous day forecast and stayed in N assets of the new forecast weights are not changed
  3. For the assets that appeared in N assets of the previous day forecast, but are not in N assets of the new forecast the weight is reduced by its weight multiplied by 1/T
  4. For the assets that did not appear in N assets of the previous day forecast and appeared in N assets of the new forecast the weight is set up to product of 1/N and 1/T

The same process is performed each day until the end of simulation. At the end date all of the positions are liquidated and the profit (return) is realized and compared to the return of the appropriate benchmark.

Top Stock Picks Portfolio Return and Hit Ratio Analysis

For each of the trading days we calculate the total return of the portfolio by taking simple percentage difference between the portfolio’s value at the last trading day close and the previous trading day close values. The hit ratio is calculated daily based on the number of profitable trades during a day – either long or short ones – to the total number of positions in the portfolio held during a trading day.

Assets Investment pool – US Stocks Universe and S&P 500 Index

For the purposes of this evaluation we take US stocks universe as a pool of assets and perform selection from it using 10 top stock picks predictability filtering. Reasonably, we take as a benchmark the most well-known index – S&P 500. When thinking of index funds as benchmarks for the whole economy, many experts tend to gravitate towards checking the S&P 500. This prominent index, followed by millions throughout the globe, has historically shined a light on the movements in the stock market. What the index does, in essence, is to choose the 500 largest publicly traded companies by order of market capitalization and produces a quarterly list of corporations to be tracked.

Top Stock Picks Investment Strategy Evaluation

The considered strategy is back-tested on period spanning from January 1, 2019 to August 2, 2019. The distribution of the total daily trading results appear to be bell-shaped with mean return of 0.56%, while the corresponding hit ratio appears to be bell-shaped and being skewed to right with mean hit ratio of 52.8%.

top stock picks

The above results suggest, that even withing the context of highly volatile market conditions when the hit ratio is expected to be affected, the algorithm provided significant number of predictions for stocks that yielded more returns outbalancing negative effects from “miss” trades. As a result the distribution of daily trades presented in Figure 2 has more weight on the positive side (right side) of the histogram.

Conclusions – Top Stock Picks Portfolio

The portfolio simulation results show that I Know First portfolio performance being based solely on top stock stock picks without any risk hedging assets of different classes or nature, although being lower than the benchmark over considerable time period, demonstrated clear upward trend since March 2019 with relatively lower volatility compared to S&P 500. Finally, since end of July 2019 the strategy performs significantly better than the benchmark ending up with out-performance by 3.89% in comparison with S&P 500 on August 2, 2019. The standard deviation of the daily change of portfolio’s value amounted to 0.52% compared to the similar S&P 500 metric of 0.72%, which is showing the relative stability improvement over the market benchmark by almost 30%.