Community

QOTW (Sept 13th 2016): How do algorithm behaviors vary across strategies?

How do algorithm behaviors vary across strategies?

 

Chart 1: Average Behavior Statistics by Algo Strategy

 

Chart 2: Spread Capture and Venue Type Distributions by Algo Strategy

  • Algorithmic trading performance is inextricably tied to the instructions an algo is given and the constraints under which it operates. We’ve shown in prior research and posts that usage varies, sometimes dramatically, between different strategies. With that in mind, we highlight behaviors observed in our execution peer database for US Large Cap trading to compare and contrast the outcomes achieved by each algorithmic trading strategy.
  • Average fill sizes were largest in dark algos, almost 2.5 times greater than fills in scheduled algos. No strategy had a large proportion of block fills (>=1,000 shares); dark and opportunistic algos were highest at 2% of total fills.
  • Dark and IS algo orders were much shorter than scheduled algo orders: 4 and 5 minutes, respectively, compared to 107 minutes.
  • Dark algos tended to fill most aggressively within the available volume with an average limit participation rate of 61%. Opportunistic algos followed at 37% while scheduled algos had the lowest average participation rate at 6%.
  • In the aggregate, dark algos had the highest proportion of fills executed at the mid point and scheduled algos crossed the spread most frequently.

US Large Cap transactions from 1/1/2015 through 12/31/2015 were included in the analysis. Placements without valid execution prices were excluded along with those that were placed outside of continuous trading hours. Scheduled strategies include VWAP, TWAP, and volume participation strategies.

 

Refer to ITG Peer Analytics for information on available Peer Related Analytics.

You may have a question or want to share your view with others in our community, please join the Incubator community and submit your entry or send us an email

 

Comments

You must be logged in to post a comment.

Please sign in or sign up.