Volume constraints are widely used as a tool to prevent an algorithm from having excessive market impact. Volume constraints allow the trader to specify the maximum fraction of volume the algorithm can reach at any point in the execution. For example, if the trader sets a volume limit of 5%, the algorithm could trade a quantity no greater than 5 for each 100 that execute in the market, even if it means the algorithm cannot complete. In the case of a 5% volume limit on a 1,000-contract order, the algorithm would only complete the order if at least 20,000 contracts are being traded in the market. If market volume is only 12,000 contracts, the algorithm could only trade 600 at most, with the remaining 400 shares left unexecuted at the end time.
The most common reason for using a volume constraint is to prevent the order from being a significantly larger percentage of volume than initially anticipated. For example, an investment manager may decide to buy 100,000 shares of a stock based on the assumption that it typically trades in excess of 1 million shares a day, making the order less than 10% of volume. Without a constraint, if market volume is significantly below that forecast, the investment manager could wind up as a very large fraction of volume. In that case, say, if realized market volume is only half anticipated market volume, the manager may also want to reduce his quantity by half, thereby limiting the order’s footprint and corresponding market impact.
But volume constraints also have certain limitations. In most broker algorithms, they simply compare the volume limit to the realized participation rate at the present time.[1] If the volume constraint is set at 10% and the market volume over the life of the order so far is 500,000 shares, then the algorithm can trade up to 50,000 shares, i.e., 10% of the market’s volume. Once it reaches that 50,000-share quantity, the algorithm will not send out additional orders until more market volume trades. For liquid stocks that trade “smoothly,” with relatively few extreme volume spikes, those constraints generally work as intended. But for stocks where volume is sporadic or episodic, this can be problematic.
Consider an illiquid stock that trades only periodically throughout the day. Suppose that, for the first 10 minutes of a VWAP order with a 15-minute duration, the stock has traded nothing. Suddenly, volume spikes for a few seconds as an aggressive trader sweeps the limit order book and executes 10,000 shares. The algorithm now has slack to trade and will be free to execute. If the volume constraint were set to 30%, the algorithm would be free to execute 3,000 shares.
But suppose further that, following this short burst of activity, the market resumes being illiquid for the remaining 5 minutes, with the limit order book remaining thin and other trading virtually nonexistent. Because of the past burst in volume, the algorithm is free to trade 3,000 shares regardless of whether that is a much larger fraction of the market while it is actually trading. For example, if no additional volume trades over the remaining life of the order, the algorithm would nevertheless continue to trade, effectively being 100% of interval volume while it is trading! The issue here is that the volume limit is calculated by simply comparing the participation rate over the entire life of the order to the volume limit. While this ensures the algorithm does not violate the volume limit in aggregate, it does not prevent the order from being a much larger percentage of market volume over subperiods.
Therefore, a related issue is that, when a volume constraint is hit, a fixed schedule algorithm can actually start to behave like a Percent Of Volume (POV) algorithm. If a VWAP algorithm is currently at its volume limit, it will wait for market volume before sending out additional orders, as a POV algorithm does. And, like a POV algorithm, the VWAP algorithm becomes reactive to volume, sending more orders out as volume takes place. So, in some scenarios, the behavior of the two algorithms becomes quite similar, but this does not mean they act identically. If market volume picks up substantially, as in our example above, a POV algorithm would increase its trading proportionately immediately, while a VWAP algorithm would simply resume trading along its (potentially realigned) schedule.
As an example, suppose a trader sends a VWAP order with a 2% volume limit on an order that turns out to be 10% of volume. In this case, the algorithm will track the market volume more closely than the volume schedule, because the volume constraint is likely to be binding for most, if not all, of the order life. The more restrictive the volume limit, the more a VWAP order will resemble a POV order, with all the potential for adverse consequences as well.
In practice, the misuse of volume constraints is one of the most common mistakes algorithm users make. Ironically, it is most often made by cost-sensitive traders with relatively small alphas that they are trying to preserve by controlling execution costs via tight volume constraints. But, in our example above, we observe that, if constraints are set too tight, the algorithm trades much more like a POV algorithm, whose reactive nature results in more spread crossing and larger impact. In such a scenario, this cost-sensitive trader’s desire to limit costs may actually increase their costs.
This article was adapted from my book Algorithmic Trading: A Practitioner’s Guide.
Footnotes [1] The algorithm will generally include any “open” orders in its calculation as well, since they have the potential to execute.
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 info@bacidore.com or via our webpage www.bacidore.com.
Please check out our new book Algorithmic Trading: A Practitioner's Guide, available on now on Kindle and paperback. 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 2021, The Bacidore Group, LLC. All Rights Reserved.
Comments