In the context of algorithmic trading, one often encounters situations where end users have additional information that could influence decisions regarding the capital allocation to specific financial instruments in their portfolios. For instance, consider a fund manager with extensive experience in macroeconomics and geopolitics who has insights about upcoming events. This fund manager may want to adjust their portfolio by selling certain stocks and buying others based on their knowledge.
However, a challenge arises when the fund manager directly intervenes in the existing portfolio. Such interference can compromise data integrity, making it difficult to accurately assess system performance in subsequent decision-making processes. It’s important to note that the input from experts often consists of sentiment, which cannot be translated into logical sequences for algorithmic formulation.
The following is the recommendations to strike a balance between the algorithm integrity and expert input.
Do Allow Interference
To benefit from expert insights, it’s essential to allow adjustments to existing portfolios. Other than that, there’s no alternative method to unlock valuable information that can enhance the current portfolio. To maintain transparency and minimize uncertainties, it’s advisable to publish a target portfolio along with the rationale before taking action. This practice prevents emotional interference but also helps subsequent analyses. It might seem counterintuitive to document expert ideas before implementation, it’s a recommended approach for sustainable growth.
Customize Expert Inputs in a Semi-algo
To ensure the integrity of the original algorithm, any input from experts should be put into a separate semi-algorithm. This process involves technically separating two trading accounts with parallel accounting, which ensures that data integrity is maintained for both algorithms.
To simplify, consider this in two steps: First, the original algorithm operates without any external interference; second, based on the original algorithm’s portfolio, an expert uses the semi-algorithm to execute actions. These actions are tracked separately within the semi-algorithm and do not directly affect the original algorithm. Notably, the semi-algorithm can sell financial instruments from the original algorithm, which is not a common practice. In simpler terms, the semi-algorithm borrows assets from the original algorithm then performs its own actions.
Quantify Expert Performance
Based on expert-matched orders, all criteria including return rate, maximum drawdown, Sharpe ratio, and Information Ratio, can be employed in a typical algorithm. After a thorough analysis of expert performance, decisions can be made on how to utilize their input to enhance the overall system performance. It’s important to note that when two algorithms run in parallel on the same account, the total return must equal the sum of the original algorithm and the expert’s semi-algorithm.
Review Expert Performance
This phase is critical to develop the future system. Does the expert’s input enhance or reduce the system’s performance? A transparent answer must be made.
In an ideal scenario, an expert’s insights lead to significant alpha generation. In such cases, adding the expert’s decision into manual trading becomes the ideal choice. Even better, if feasible, integrating the decision-making process of experts as a feature within the algorithm itself is the ultimate goal.
However, when the results from expert input are lackluster, or even negative, a different approach should be adopted. Fully relying on the original algorithm, without any expert judgments, may be the optimal decision. This strategy maintains performance while saving valuable time for the expert. Either way, it’s advisable to quantify the benefits from expert inputs, even though this approach can be challenging in certain circumstances.
In summary, expert input can be seen as an independent algorithm running alongside the live trading system. The key distinction lies in the expert’s ability to leverage the current portfolio as input to optimize trading performance. Evaluating expert input with metrics like the Sharpe ratio, return rates, and maximum drawdown allows us to quantify their overall value for future decision-making.