Optimizing Short-Term Trading Predictions Of Your Algo
Today let’s look at a very interesting paper from the quant minds of Charles-Albert Lehalle & Ngoc-Minh Dang. Entitled, ‘Using Short-Term Predictions for Participation-Rate Driven Trading Algorithms.’ Though the paper was originally published in 2012, we still think it is quite relevant for the markets of 2023.
The authors “propose a decomposition of algorithm’s a priori performance, from which we separate contributions from different factors” and “show that, in combining estimations on volume and price and always taking into account the price-impact effect, one is able to optimize the execution in a sequential manner.”
The authors believe that actual success or failure of the trade relies on one of 3 principles which they look to break down:
- firstly, performance related to the algorithm’s constraints
- secondly, performance related to the execution context
- lastly, performance related to the quality of execution
The authors justify this decomposition with the belief that a, “strategy’s behaviors must depend on not only the execution context but also its own performance. In reality, the execution context consists of the future joint distribution (price, volume), while the performance is captured by: (i) the (price) slippage with some benchmarks: Percentage of Volume (PVOL), Volume Weighted Average Price (VWAP), Implementation Shortfall (IS); and (ii) the volume slippage (for instant in PVOL benchmark).”
An implementation shortfall refers to the difference between the expected performance of an investment strategy or portfolio. And the actual performance achieved. This can occur for a variety of reasons, including market conditions, changes in investment strategy, or errors in implementation.
One common cause of implementation shortfall is the failure to fully execute an investment strategy due to market conditions.
For example, if an investor wants to buy a stock at a certain price. However, the stock’s price jumps before the trade can become executed. As a result, the investor may not be able to buy the stock at the intended price. As a result, an implementation shortfall. Similarly, if an investor plans to sell a stock at a certain price. However, the stock’s price drops before the trade can become executed. As a result, the investor may not be able to sell the stock at the intended price. Resulting in an implementation shortfall.
Another cause of implementation shortfall is changes in investment strategy. For example, if an investor is planning to hold a stock for a long period of time but decides to sell it earlier than planned due to changing market conditions, the investor may not be able to achieve the intended returns, resulting in an implementation shortfall.
Finally, implementation shortfall can occur due to errors in implementation. For example, if an investor mistakenly buys a stock at a higher price than intended, or if a portfolio manager fails to execute a trade in a timely manner, the investor or portfolio manager may not be able to achieve the intended returns, resulting in an implementation shortfall.
At the end of the day, the authors attempt to:
“decompose the realized performance into different components, through which to reduce the sub-optimal ones.”
And that, “These decompositions can now be used to understand and control the performance of any trading algorithm having a multi-criteria constraints (usually one on prices and another on volumes or participation rate).”