Two years after launching the Midpoint Extended Life Order (M-ELO), we re-visited the design to ensure it continues to deliver on its early success and stated purpose. The holding period is core to M-ELO success and is what makes it such an effective purpose-built order type, consistently delivering the performance customers need. When M-ELO was initially designed, a half-second was selected to incorporate an easy-to-understand human element. We always planned on revisiting this holding period with a data driven approach to recalibrate and optimize M-ELO opportunities and executions. With months of analysis completed, we will be reducing the M-ELO holding period to 10ms, which in effect will open up M-ELO to use cases that are presently unavailable.
When comparing against similar order types and venues that utilize a delay mechanism, M-ELO consistently outperforms. IEX recently issued a piece discussing its D-Peg order and how it performs against displayed bids and offers. Since midpoint and displayed trading are quite different, it is appropriate to compare IEX’s D-Peg to a product that is a closer competitor, such as M-ELO. The results show that M-ELO has an average 1-second markout of -1% of the spread, which is 3% better than IEX’s D-Peg, and a 1-minute markout 9% better
As a result of M-ELO’s performance and trading experience, our customers are looking to expand use cases for more M-ELO executions based on their trading experience to date. As such, we have revisited key design assumptions to increase the opportunity set for our clients. While it is not possible to test directly the impact of different holding periods on execution quality, we can rely on the following two proxies: midpoint-to-midpoint executions on the Nasdaq Continuous Book and M-ELO executions.
First, we analyze the time-to-execution and the cost of delay of midpoint-to-midpoint trades that took place on the Nasdaq Continuous Book. The cost of delay is measured as the difference in basis points between the midpoint execution price and the prevailing midpoint after 50, 200 and 500 milliseconds. Our results suggest that the median time-to-first execution on non-tick constrained, liquid securities with low M-ELO market share is 50ms, while its considerably higher for securities with high M-ELO trading activity. Furthermore, the average cost of a delay is zero regardless of the holding period, while the volume weighted difference between the costs of delay calculated using 50ms and 500ms holding period is only 0.02bps. The result is driven by the tails of the distributions.
In the next step, we compared each M-ELO execution price with the prevailing midpoint of the NBBO available at 100ms, 200ms, 400ms, 450ms and 490ms before the execution time. The below table suggests that:
- If M-ELO’s holding period was reduced on securities having an average daily dollar volume higher than $250 million and bid-ask spread higher than $0.02, would continue to yield impressive markouts.
- If M-ELO’s holding period was reduced on securities having an average daily dollar volume higher than $25 million and bid-ask spread higher than $0.02, we did not observe a material difference in markouts.
- If M-ELO’s timer was reduced on all securities, M-ELO would remain the highest performing order type amongst its peers with a slight positive increase of 0.10 basis points.
We conclude that reducing the M-ELO holding period to 10ms will increase clients’ opportunity to implement additional use cases for M-ELO while sustaining its strong performance characteristics. If we were to view, the impact on a dollar cost per share basis, for three actively traded stocks, there is close to zero cost incurred for reducing the holding period to 10ms.
Armed with this data, the M-ELO holding period will be reduced to 10ms. M-ELO will remain a purpose-built solution with one of the longest delay mechanisms in the industry, and enable clients to incorporate the order type across a broader set of opportunities.
 Our sample includes all liquidity providing (the first entered order in any execution) orders that were entered 1 millisecond after an NBBO update and executed within 3 minutes. There are 912,482 M-ELOs. We assume that the markouts at 0 milliseconds is 0% M-ELO orders (similar to the IEX methodology).
 Since we are interested in understanding how long it takes before a midpoint to be executed by another midpoint, our time-to-execution is measured as the time difference between the first execution message of the order and its entry time.
 500ms is M-ELO’s current holding period