top of page
  • Writer's pictureJeff Bacidore

Why Use a Portfolio Trading Algorithm? (And why not)

In recent posts, I have been focused on algorithm nuances that can have disproportionate effects on algorithm performance. In this post, I am going to move in the opposite direction and discuss a much broader topic: portfolio trading algorithms. Specifically, this post outlines how portfolio trading algorithms can add value over the VWAP/TWAP algorithm, which is commonly used to implement portfolio trades. I chose this topic primarily because it has come up with clients repeatedly recently. But also because, frankly, I don’t think I can stomach reading – let alone writing – another article on payment for order flow. (But if you can, please take a look at my relatively recent post, Payment for Order Flow: My Two Cents (per hundred)).

Portfolio Trading Algorithms

Portfolio trading algorithms manage multiple orders simultaneously, trading them in a coordinated manner meant to optimize the cost/risk trade-off of the aggregate execution. With that said, portfolio traders often rely on VWAP/TWAP algorithms to implement their trades. At first blush, this may seem somewhat counterintuitive – why would someone use a single-order algorithm to execute a basket? The answer lies in the fact that most portfolio trades are risk-balanced. A trade basket that is balanced at the start of the trade would generally stay balanced throughout the execution since the individual algorithm will be following roughly the same schedule. An imbalanced basket would see its risk fall in proportion to the common (TWAP) or nearly common (VWAP) schedule. How quickly the risk declines to zero is attributable solely to the trader’s choice of trading horizon – the shorter the horizon, the faster the risk is reduced.

To add value over these schedule-based single-order algorithms, and thereby justify their complexity, portfolio algorithms must do more than simply maintain risk profiles over time or shorten their trading horizons by trading more aggressively. So, how do portfolio algorithms add value?

First, a portfolio algorithm can create individual trading schedules that best balance cost against risk on an aggregate level. Only in relatively rare cases will a portfolio algorithm be able to no significant value. This generally occurs only when 1) the basket is initially risk-balanced, and 2) the basket contains orders that have similar risk and cost characteristics. With regard to (1), many quant funds, for example, generate portfolio trades that are risk-balanced, as the fund is likely moving from one risk-balanced position to another. Thus, for these trades, it is challenging for portfolio algorithms to add value over TWAP/VWAP. But when the basket is not well-balanced from a risk perspective, portfolio algorithms can add considerable value by trading the basket strategically and focusing on managing any risk imbalances in a cost-sensitive way. With regard to (2), orders within a basket often have very different risk and cost characteristics. Ideally, they would be traded differently to efficiently manage risk, something a portfolio algorithm can accommodate, but TWAP/VWAP cannot.

Second, a portfolio algorithm can add value by trading certain positions opportunistically, and then dynamically adapting its future trading to rebalance the portfolio. TWAP/VWAP cannot accommodate opportunistic trading natively. Any “hole” blown into a portfolio trade via an opportunistic execution will only be plugged slowly, at a rate proportional to the TWAP/VWAP curves as noted earlier.

Third, a portfolio algorithm can add value by accommodating order-level volume constraints more rationally from a risk perspective than single-order algorithms. For example, suppose the portfolio algorithm begins with a risk-balanced basket, but the trader applies volume constraints to the orders. Depending on the realized volume and the strictness of the limits, the basket could become significantly imbalanced if orders were routed to single-order algorithms. But portfolio algorithms can incorporate volume constraints in two ways. First, when determining the initial trading strategy for each order, they can evaluate the likelihood that orders will hit a volume constraint during the execution and adapt the initial plan accordingly. Second, as the algorithm trades and some orders unexpectedly hit constraints, the algorithm can update its schedules by reoptimizing, taking such complications into account. Single-order algorithms, on the other hand, have no mechanism to eliminate constraint-induced portfolio risk imbalances.

Fourth, a portfolio algorithm can allow for portfolio-level constraints that single-order algorithms cannot accommodate. For example, suppose a portfolio manager plans to use the proceeds of the sell orders to fund the buy orders, which is an exogenous funding requirement. If the sell side of the basket is significantly more constrained than the buy side (via price and/or volume constraints), single-order algorithms may end the day with less cash than they need to fund the buys. This can be costly and operationally burdensome. A portfolio algorithm, by contrast, could slow down the buys to ensure the dollars bought are roughly in line with those sold.

Fifth, a portfolio algorithm can be more sophisticated about intraday cost and risk patterns. VWAP algorithms, for example, tend to trade more around the open in U.S. equities, despite evidence that the costs of trading are often highest at that time. Portfolio trading algorithms are not bound to follow an arbitrary fixed schedule, so they can incorporate intraday variations in cost and risk into their optimizations.[1]

Finally, a portfolio algorithm can add value by buying and selling other assets to manage risk in a more automated fashion. For example, suppose a trader is executing a large equity sell basket, which has significant exposure to market downturns. A portfolio algorithm could initiate a hedge by selling the appropriate amount of a correlated futures contract immediately (e.g., Russell E-minis to hedge a small-cap sell basket), and then unwinding the futures position in tandem as the algorithm executes the original basket. In this scenario, the algorithm effectively executes the “beta” part of the sell trade immediately using liquid futures, so any market-related losses incurred are offset by gains to the short futures position.

With this in mind, why wouldn’t someone use a portfolio trading algorithm? Perhaps the biggest drawback is that shortfall algorithms generally tend to front-load executions, as a means to reducing the aggregate execution risk. Bigger baskets and riskier baskets get front-loaded more, all else equal. So, while having the algorithm trade individual orders differently is a key benefit, the fact that the algorithm also likes to trade everything more quickly makes them less attractive. Another issue is that some trades may be comprised of multiple non-identical trade lists, aggregated for workflow reasons. Optimizing the aggregate trade may be suboptimal for some, if not all, accounts. This alone is typically a deal breaker. Other common complaints are the opaqueness of the algorithm and other workflow issues (e.g., communicating lists). In the aggregate, then, the common front-loading coupled with the fact that the risk/cost trade-off provided by VWAP/TWAP is often close to “optimal” has stymied the broader use of these cutting edge algos. And, as a designer of some of these algorithms, the fact that I am left writing blogs from a NYC office and not feet up drinking mai tais in a place with a name starting with “Cabo” is testament to the fact that these things just haven’t lived up to their promise. At least, not yet.

This article was adapted from the author's book Algorithmic Trading: A Practitioner's Guide.

References [1] See, for example, Bacidore, Wu, and Xu (2013). “Balancing Execution Risk and Trading Cost in Portfolio Trading Algorithms.” Journal of Trading, 8(4), 37-43.

The author is the Founder and President of The Bacidore Group, LLC and author of the new book Algorithmic Trading: A Practitioner's Guide. For more information on how the Bacidore Group can help improve trading performance as well as measure that performance, please feel free to contact us at or via our webpage

Please check out our new book Algorithmic Trading: A Practitioner's Guide, available on now on Amazon. Click Here for more details.

For an overview of our other services, please click HERE.

And please check out our other blog posts available HERE.

And if you'd like to receive notification of new posts, please join our mailing list below.

Copyright 2022, The Bacidore Group, LLC. All Rights Reserved.


bottom of page