Examining Execution Stability at IEX
Individual executions are the fundamental units of trading, and the first step in executing a trade is identifying a plan. We can classify the strategies available to the trader into three broad categories based on what price they seek to execute at relative to the National Best Bid and Offer (NBBO)
- Spread-cross — e.g: buy a stock by lifting the offer, essentially giving up edge to the seller in return for immediacy and a high degree of certainty around the trade price.
- Spread-capture — e.g: try to buy a stock at the bid or less than the mid. This strategy is a slightly greedy one, as you’re looking for a deal relative to the best displayed offer. Of course, deal-seeking comes with additional risk of opportunity cost from not finding a willing counterparty.
- Compromise — e.g: try to buy a stock at the midpoint of the NBBO. Here you are not looking for the best possible price relative to the market but instead are trying to meet your counterparty halfway. Naturally, this option falls at a point between the previous two on the risk/reward curve.
Using this simple framework comes with complications: not all executions within these categories are equal. For example, a midpoint trade that happens immediately before the midpoint changes is not the same kind of balanced compromise as a trade at the midpoint of a market that is not in transition¹.
Execution stability is an important factor in evaluating how well the intent of a trading strategy is realized.
At IEX, we believe that exchange design can play a big role in execution performance. Since our inception, we’ve focused on building technology designed to help traders avoid adverse selection and transact as intended relative to prices that are not imminently changing.
And the data shows that we are successful in doing so.
Stable Midpoint Volume
In August 2020, 95.5% of IEX’s midpoint volume was executed in a stable market, defined as executions not occurring in the 2ms periods before NBBO changes. This compares to 74.8% for Nasdaq, 85.3% for EDGX, 78.9% for NYSE, 77.7% for Arca, and 75.8% for BATS.
The main contributor to IEX’s market-leading level of stable midpoint volume is our Discretionary Peg order type (D-Peg). D-Peg orders benefit from IEX’s Speed Bump and our Crumbling Quote Indicator (i.e., the CQI or IEX Signal).
The Speed Bump delays all incoming orders by 350 microseconds (μs) which is designed to give IEX time to determine the most up-to-date NBBO to which we should peg resting orders. The CQI goes further and tries to anticipate a changing NBBO before it happens, allowing resting D-Peg orders to avoid executing during these unstable moments.
These two features are designed to work in tandem to enhance the stability of D-Peg trades, but it’s natural to wonder how they individually contribute to this performance. The following is one way to decompose their impact.
Technology Tag-Team
Consider an incoming order that could potentially trade with a resting D-Peg order at IEX. The way this order interacts with the Speed Bump and CQI depends on the state of the market when the incoming order arrived.
Consider the following cases:
- (Speed Bump-Protected) The incoming order arrives within 350μs after IEX sees an NBBO move. In this case, without the Speed Bump, we would have pegged a resting D-Peg order at an undesirable price (one that would have changed within 350μs). The 350μs window afforded by the Speed Bump allows IEX to price the D-Peg order at the midpoint of the newly established, current NBBO.
- (CQI-Protected) The incoming order arrives between 2ms before and 350μs after IEX sees an NBBO move AND following a correct CQI prediction. In this case, without the CQI, a D-Peg order could have executed at a midpoint that would be stale within 2ms. With the CQI functionality, the D-Peg order would stay booked 1 MPV outside the NBBO, where it is less likely to execute.
- (Unprotected) The incoming order arrives between 2ms before and 350μs after IEX sees an NBBO move WITHOUT a correct CQI prediction. When IEX is unable to anticipate an upcoming price change, an incoming counterparty has the potential to generate an execution at an unstable price.
- (Stable NBBO, no protection needed²) The incoming order doesn’t arrive in a (2ms before, 350μs after) window around an NBBO move. In this case, IEX doesn’t need to do anything special to execute at the intended pegged price, since it is not changing.
Note that all marketable-to-mid orders fall within at least one of these cases. For orders that fall into more than one case, we’ll assign them using the priority: CQI-Protected > Speed Bump-protected > Unprotected > Stable.
Scenario Breakdown
The following visualizations break down the count of incoming marketable-to-mid³ orders from August 2020 into the above four categories. Each vertical line represents the breakdown of an individual security. First, let’s look at a set of well-known symbols:
The symbols are sorted by the percentage of orders in the “stable” category; you can see that across symbols, there is a diversity in the percentage of incoming orders received during a stable market.
Some symbols, such as GOVT, rarely tick, and therefore most order arrival times are inherently stable and neither the Speed Bump nor CQI play a big role in helping Members trade at the midpoint in a stable market.
Other tickers, such as AMZN, are thinly quoted with wide spreads, so the Speed Bump gives IEX the ability to observe the NBBO changes when they happen, while the CQI plays a small role⁴.
Finally, there are some heavily quoted tickers, such as SPY, with more structured quote changes, where the CQI interacts with more incoming orders than the Speed Bump alone.
Expanding our symbol universe to those that traded an average daily notional value of at least $50M over the month of August 2020, we see that this kind of diversity is widespread:
The Speed Bump and the CQI’s relevance in preventing adverse selection varies across symbols, but it is the combination of these two technologies that allows us to raise the floor on all symbols by a significant degree. That is, without the Speed Bump or CQI, all of the counterparty orders outside the darkest-blue section could potentially generate unstable executions with our resting D-Peg orders.
Another way to think about it is: if IEX did not have the Speed Bump or the CQI, the entire graphic above the darkest blue segment would be red.
The Next Frontier in Stable Executions
By appropriately filtering incoming orders across all symbols, the Speed Bump and CQI play a significant role in boosting IEX’s stable midpoint volume percentage to 95.5%.
With stability so high, the natural question is: Where do we go from here?
The answer is: to displayed trading. While IEX’s focus to this point has been on pegged (non-displayed) orders, displayed orders also experience significant levels of adverse selection.
In fact, for the aggressive (spread-crossing) contra orders that are eligible to trade with displayed liquidity (a subset of the marketable-to-mid orders considered above), the diversity in the analogous four-category breakdown is even more dramatic. Since displayed limit orders are not pegged to the NBBO, we’ll discard the “speed-bump-protected” category — aggressive orders received in the windows surrounding NBBO changes will either need to be caught by the CQI, or they will count as “unprotected.”
When we decompose aggressive orders into the three remaining categories by symbol, we see that the diversity across symbols is even greater. Here is the previously used set of well-known symbols:
(This time, we sort by percentage of aggressive orders in the CQI-protected category). And here is the full set of symbols that traded an average daily notional value of at least $50M over the month of August 2020:
You can see that, compared to counterparties at the midpoint, you have significantly fewer opportunities for a stable execution when your contra is an aggressive order. For most symbols, the fraction of stable contra-orders is less than 30%!
However, notice that many of the unstable contra orders are anticipated by the CQI for a large portion of symbols. For example, in TWTR, just 28% of the spread-crossing counterparties arrived during a stable market. Of the remaining unstable orders, nearly 2/3 arrived following a CQI prediction! If we could prevent displayed limit orders from trading with these aggressive orders that were predicted by the CQI, we would have a chance at significantly reducing the number of unstable executions, allowing the orders to achieve true spread capture.
On August 26, 2020, the SEC granted approval to IEX’s Discretionary Limit (D-Limit) order type, which expands the application of the CQI to displayed orders. We view this as a major innovation in displayed trading, and we’re excited to see how its introduction on October 1 affects the way market participants look for spread capture.
IEX is optimistic that using CQI to reduce D-Limit’s executions during moments of instability will present a competitive option to resting displayed limit orders. We look forward to iterating on this new direction for displayed trading and continuing to develop innovative ways to provide high-quality executions for all market participants.
[1] Similarly, buying at the bid is not really achieving true spread capture if that bid becomes the new offer from the NBBO spread shifting downward immediately after the execution. See: Who Ever Said Buy High and Sell Low for an in-depth discussion on how stability can cause the performance of the three categories to blend into each other.
[2] As a probabilistic model, the CQI has the potential to “incorrectly” apply protection. False negatives and false positives could restrict a resting D-Peg order’s discretion in categories 3 + 4 respectively, preventing trades that might have been stable or even in the resting order’s favor. Despite the potential for these missed opportunities, IEX is able to trade more stable midpoint volume than any other exchange.
[3] Marketable-to-mid meaning “priced at least as aggressively as the midpoint of the NBBO.” Such orders are eligible to transact with resting D-Peg orders.
[4] In thinly-quoted names, there is often a single stock exchange at the NBB/NBO. As a result, quote changes are often atomic events that are less predictable. For more, see “lonely quotes” here.
About IEX Exchange
IEX Exchange is a U.S. stock exchange designed to provide superior execution quality and set a new standard for trading. Since launching in 2016, the exchange has grown to be one of the largest exchange operators globally, by notional value traded. To learn more about IEX Exchange and other IEX businesses visit iextrading.com or search IEX.
©2020 IEX Group, Inc. and its subsidiaries. Neither the information, nor any opinion expressed herein constitutes a solicitation or offer to buy or sell any securities or provide any investment advice or service. The information herein is believed to be reliable, but the Firm makes no representation as to the accuracy or completeness of, and undertakes no duty to update, information herein and any and all liability is expressly disclaimed relating to or resulting from use of this information.
The trademarks, service marks, and logos (collectively, the “Trademarks”) displayed on this document include registered and common law Trademarks owned by third parties. IEX’s use of any thirty-party Trademark does not imply any affiliation with or endorsement by its owner(s) nor that such third party has granted any license or right to use such Trademark without written permission from the owner(s). This document has not been prepared, approved, or licensed by any of such third party owner(s).
This document may include only a partial description of the IEX product or functionality set forth herein. For a detailed explanation of such product or functionality, please refer to the IEX Rule Book posted on the IEX website: www.iextrading.com
