Details#

On this page, we provide a full technical explanation of how Grandiat works, so that you can have complete confidence in the results.

Step 1: Strategy Creation#

In the strategy creation step, Grandiat converts your investment thesis, written in plain English, into an investment strategy. An investment thesis is a statement of the form, “stocks with [some characteristic] outperform the market.” An investment strategy is a procedure that tells you what stocks to include in your portfolio at any given time using the information available at that time.

Step 1a: Factor Construction#

First, Grandiat uses AI to represent “[the characteristic]” in your thesis as a series of mathematical expression that we call “factors.”

For example, if your thesis is “stocks with higher earnings outperform the market,” Grandiat may represent “higher earnings” using a single factor:

1. net_income_0 / market_cap_0 (where net_income_0 is the earnings for the most recent time period and market_cap_0 is the market cap for the most recent time period)

If your thesis is “stocks higher earnings and growing revenues outperform the market,” Grandiat may represent “higher earnings and growing revenues” using two factors:

1. net_income_0 / market_cap_0 (same as above)
2. revenue_0 / revenue_1 (where revenue_0 is the revenue for the most recent time period and revenue_1 is the revenue for the previous time period)

Step 1b: Portfolio Construction#

With a set of factors in hand, Grandiat can create the investment strategy. As we mentioned, an investment strategy is a procedure that tells you what stocks to include in your portfolio at any given time using the information available at that time. For a given set of factors, the associated strategy is as follows. On the last trading day of each month, stocks are ranked by each of the factors, using all data available as of the end of the previous day. The average ranking over all factors is computed for each stock. The top n stocks (where n is a configurable parameter) are bought at the close of the trading day (equally-weighted) and sold one month later at the end of the last trading day of the next month.

Step 2: Strategy Testing#

Once we have defined an investment strategy, the job of AI is done and traditional data processing algorithms take over to test the strategy. For the last trading day of every month, starting from December 1999, a dataset is prepared consisting of the information known as of the day before. The strategy procedure described in Step 1b above is applied to this dataset in order to simulate the performance the strategy would have had that month.

Step 3: Strategy Implementation#

When you are satisfied with your strategy and you are ready to implement, you can download the portfolio of stocks that you should buy at the end of the current month. You should hold these stocks for one month and repeat the process the next month. For many strategies, the portfolio may not change too much from month to month, and you will only have to make a few changes every month. If you use Interactive Brokers, we provide a file that can be uploaded to the desktop software every month in order to automate the order creation (see here for details).