Stock Picking Algorithms: The Machine Advantage

Tomer Solel is a Financial Analyst  at I Know First. He graduated from Cal Poly Pomona with a bachelor’s degree in applied mathematics.

Stock Picking Algorithms

Stock Picking Algorithms


  • Investors pick stocks in different ways such as from the news, from companies that make products they like, from balance sheets, using fundamental analysis or technical analysis, but all of those methods have different risks and different issues
  • The algorithmic method is the best method since it is unbiased and it detects market movements in its predictions. It is the method used at I Know First.
  • I Know First provides daily forecasts for different time horizons, giving each stock a signal and a predictability, and the September 27th case study is a good example.
  • In that fourteen day forecast, I Know First correctly predicted 8 of 10 stocks for the long position, returning a great average and crushing the S&P 500’s performance.
  • There is no magic in stock picking algorithms, but although the algorithm has many uncertainties, it still makes very advanced predictions.



How does one pick stocks? Those who seek adventures might look for companies in the news, or those companies that are rumored to be taking over. If they hear positive news, they buy the stock, with hopes that the stock will go higher. Some of us are lucky, but most of us aren’t, and, therefore, we lose money. Is luck related to it at all?

Other people simply buy stocks in companies that make products which they like. Most Apple (NASDAQ: AAPL) investors belong to that category. Investing in Apple is a win-win situation for them as either they can win big, or they can, at least, own stocks in a company that they like.

Some of us are knowledgeable in accounting. We look through different balance sheets to try and find out what a company is really worth putting our money in. This is known as value investing, a technique known to be employed by the great Warren Buffett. The dividend investing strategy is a conservative form of this technique.

Fundamental analysis is important. However, the common argument against it is that the current stock price already reflects the known fundamentals. Therefore, buying one stock has no advantage over buying another. Those who solely rely on fundamentals are missing out on a big chunk of information that is encoded in the daily price movements.

Another type of investors is technical analysts. Those investors are the ones who like to read charts. They rely on chart patterns and known indicators to find oversold or overbought stocks. They use past patterns in order to predict future patterns. But since the world around us is constantly changing, we can’t expect a pattern to repeat itself. In addition, these patterns and indicators tend to lose effectiveness when too many investors use them to trade at the same time.

Stock Picking Algorithms

The Algorithmic Method

At I Know First, we use computers, mathematics, and self-learning algorithms to pick stocks. Markets move in waves, and our algorithms are designed to detect and predict these waves. Each algorithmic forecast has many inputs from many different sources, with each input affecting the outcome. The output of each stock is an up or down signal, along with its predictability. The use of mathematics and algorithms eliminates a big problem that all of us share, being emotional. When it comes to money decisions, we make decisions that are irrational and too impulsive. We buy when we should be selling, and sell when we should be buying. Most of us are not capable of reading balance sheets, comprehending the news, and, therefore, we can’t make informed decisions about buying or selling.

The Forecast Table Explained

The “heat map” table below is the forecast which was issued on the morning of September 27th, 2017 before the market opened. Each cell in the table represents either a stock or an index. Along with the ticker, there are two numbers, the signal in the middle right and the predictability in the bottom left. The stocks are arranged by the strengths of their signals, left to right and top to bottom.

On the top left corner of the table is the strongest up signal, which in this case is Xoma Corporation (NASDAQ: XOMA), followed by Alnylam Pharmaceuticals Inc. (NASDAQ: ALNY), followed by 8 more stocks rounding out the top 10.

The signal is the move direction of the forecast, with a positive up signal or a negative down signal. Predictability refers to the historical correlation between the forecast and the actual result. Thus, it is related to the probability of the predicted move occurring. Green cells are for an up signal while red cells are for a down signal, helping us identify the direction of the market.

Case Analysis of the September 27 Forecast:

In this exercise, we try to find out whether our forecast could have been anticipated by conventional analysis methods. Among the “I Know First” last year top 10 stock picks, 8 of the 10 stocks in the long position went in the predicted direction in accordance with the algorithm, for the fourteen day time horizon from 09/27/17 until 10/11/17.

First, the top 10 long position stocks:

Xoma Corporation (XOMA) had the strongest upward signal of the September 27th forecast. It went up, ending on October 11th with a significant 18.04% gain. Alnylam Pharmaceuticals (ALNY) went up as predicted, with a nice 11.42% gain. Oramed Pharmaceuticals (ORMP) increased by 4.99%, Palatin Technologies Inc. (PTN) increased by an amazing 32.28%, Insmed, Inc. (INSM) increased by 3.56%, AstraZeneca PLC (AZN) increased by 3.72%, and Bristol-Myers Squibb Company (BMY) increased by 4.10%. But the most amazing prediction for the two weeks by far was for Celsion Corporation (CLSN), which increased by an astounding 274.36%. The algorithm predicted the rise of these stocks, which is something that visible fundamentals could not have done. Opko Health Inc. (OPK) and Taro Pharmaceutical Industries (TARO) were the only 2 stocks for this forecast for the year that the algorithm failed to predict, but the algorithm still had a great return of 34.81%, outperforming the S&P 500 by an enormous 32.47%.


In conclusion, stock picking algorithms simply are not magically going to do the job for investors. One can follow the fundamentals, the daily price moves, the company reports, the news, the competitors, the suppliers, the patents, the lawsuits, the weather, and more to predict the stock market. Some forecasts could have been produced by those who follow what I mentioned above. However, algorithms are completely different, using computers and logic.

The algorithm does not know everything that is out there. It still has a lot of uncertainty, and therefore what it does is it makes predictions based on the data available. A lot of pieces are simply left out when creating a forecast, the noise is separated from the relevant data, but still this algorithm gives you a good idea of existing opportunities. Combined with other forms of stock analysis, the algorithm is an additional tool in making clever investment decisions.