Inside the performance dynamics of two IEX Exchange order types.
In this blog we discuss the performance dynamics of two IEX order types: Midpoint Peg (M-Peg) and Discretionary Peg (D-Peg). If you would like a backgrounder on the baseline functionality of each order type, you can start here. This post focuses on the distinctions between the order types and how understanding those nuances may be able to help Members improve theirtrading experience on IEX.
While this analysis does not provide a clear “formula” for when to use one order type over the other, we hope it will help users optimize for different strategies. In going through the data, it becomes evident that when Members are focused on queue priority (e.g., when using liquidity seeking and other aggressive strategies), they may want to lean toward M-Peg, while D-Peg has been more helpful when there is a greater focus on price improvement (e.g., for more passive or price sensitive strategies).
Additionally, the data highlights the types of names where these advantages have been most pronounced. M-Peg’s advantage in queue priority is most impactful in tighter-spread names, while D-Peg’s price improvement advantage stands out most in wider-spread names. At the same time, it is tighter-spread names that benefit the most from the D-Peg’s Signal functionality in terms of markouts.
In this piece, we first analyze queue priority, then break down pre-trade and spread capture price improvement.[1]
Queue Priority
One very important distinction between M- and D-Peg is that M-Peg rests at the midpoint and thus has price priority over D-Peg, which rests one tick below the bid or above the offer and has discretion to trade to the midpoint.[2] This is an intentional feature of D-Peg as it allows more potential for price improvement (discussed below), but it comes at the cost of its queue priority.
To understand where queue priority is most important, we looked at how often D-Peg cedes priority to M-Peg orders because of the order book priority dynamics.[3] To do this, we calculated the percentage of D-Peg orders and shares that traded after an M-Peg[4], solely because of the M-Peg’s price priority (i.e., the resting D-peg had time priority). As you can see below, it is less likely that M-Pegs take priority over D-Pegs in the wider-spread names where queues are relatively less competitive. For example, in the $0.01 spread bin, 15% of D-Peg shares traded after an M-Peg whereas in the $0.25 spread bin that number was 11%.
To directly compare M-Peg’s queue position advantage versus D-Peg, we look at “hit rates,” or the percentage of orders with at least one trade. We chose to use hit rate instead of fill rate, because it normalizes for order size. Again, we see a greater M-Peg hit rate advantage in the tighter-spread versus wider-spread names.
As traders weigh the strengths of M-Peg and D-Peg, this data suggests that M-Peg’s queue priority is more impactful in tighter spreads, meaning the opportunity cost of using D-Peg instead of M-Peg is highest in those names. In more aggressive, liquidity seeking strategies, M-Peg’s priority can help users avoid being queue jumped and increase their probability of fill.
Price Improvement
Price Improvement (PI) can be achieved by all pegged order types, but D-Peg’s unique features allow it to maximize probability of PI, setting it apart from nearly all other order types.
D-Peg has three features that enable PI: discretion to the midpoint (unique to D-Peg), pegging relative to the NBBO (possible for all pegged order types including M-Peg), and Signal protection (unique to D-Peg). Looking at different measurements and sources of PI helps determine where D-Peg’s PI is most impactful versus M-Peg.
Discretionary Price Improvement: Spread Capture PI
D-Peg’s discretion to the midpoint (i.e., its ability to step up only as far as necessary to meet a counterparty) allows it to capture spread in situations where M-Peg is unable to.
We look at PI relative to the midpoint at the time-of-trade to measure spread capture PI. This analysis isolates the amount of PI that D-Pegs achieved by executing at prices better than the midpoint, excluding the benefit of floating with the quote.
For example, if the market is $10.00 x $10.10 with the midpoint at $10.05 and the stock price changes to $9.99 x $10.09, the new midpoint is $10.04. If a sell order comes in to cross the spread, D-Peg will execute at $9.99, that is $0.05 of PI since it was $0.05 better than the midpoint at the time of trade ($10.04). The methodology allows us to isolate when spread capture occurred.
Below, we show the percentage of D-Peg trades and volume that achieve PI through spread capture. As you can see, ~9% of D-Peg trades and ~7 of D-Peg volume executed at a price better than the mid, achieving an average savings of 14.2 mils/share. M-Pegs receive no PI in this way because they can’t execute at a price better than the midpoint.
To see exactly where D-Pegs benefit most from spread capture PI, we look at the proportion of D-Pegs that achieve time-of-trade PI and the average improvement by spread. We find a higher percentage of D-Pegs achieved PI and higher notional PI in wider-spread symbols. This means that the discretionary feature of D-Peg is likely more useful in wider spreads because there are more price points within the spread at which to trade.
Pegging to the NBBO: Pre-Trade PI
Both M-Peg and D-Peg are pegged relative to the NBBO, which means that they update their resting prices when the NBBO changes. When the quote changes, M-Peg updates to the new midpoint while D-Peg moves to one tick outside the new NBBO.
As detailed in this previous post, the Speed Bump is designed to give IEX an advantage in effectively pegging orders compared to other exchanges. Exchanges are responsible for managing the pricing of pegged orders. Because all incoming orders on IEX must pass through the Speed Bump, IEX has that extra time to consume, digest, and incorporate quote updates from other exchanges. This is crucially important for pegged order types as it designed to ensure their prices are pegged to the most up-to-date NBBO. The idea is to have the most accurate pricing information to allow IEX to reprice our pegged orders before a counterparty can interact with a resting order at a stale price.
Since both order types have this functionality, we compare them side-by-side to see if one order type receives more of this benefit than the other. We measure this type of PI using pre-trade price improvement.
A higher percentage of D-Peg than M-Peg volume achieves pre-trade PI at every spread. Though both are pegged relative to the NBBO, M-Peg’s relatively more aggressive midpoint price means it executes before D-Peg. As M-Peg interacts with aggressive contras first, it has less time and fewer opportunities to achieve PI. Additionally, D-Peg has Signal protection, which prevents it from trading in situations where M-Peg will, giving D-Peg even more time and opportunity for PI.
For example, imagine the market is $10.00 x $10.10. There is an M-Peg buy order for 100 shares (pegged at $10.05) and a D-Peg buy order for 100 shares (pegged at $9.99). A seller of 100 shares comes in with a $10.00 limit. First the M-Peg buy order will be filled at $10.05. Next, the quote becomes unstable and the Signal fires, which turns off D-Peg’s discretion to $10.05. Another seller comes in with 100 shares and a $10.05 limit, but this order does not trade. The Signal turns off. Then the market moves to $9.99 x $10.09, making the D-Peg reprice to $9.98 (one tick below the bid). A seller comes in with 100 shares and a $10.00 limit. The D-Peg steps up and trades at $10.00.
In this example, the M-Peg did not have a chance to achieve PI because it traded before the D-Peg. The D-Peg gained PI because the Signal fired, preventing it from trading at the midpoint, and allowing it to trade at an even more passive price.
Below, we look at the percentage of D-Peg volume that achieves pre-trade PI when the Signal fires over the life of the trade. This aims to show situations like the example above when, because the Signal fired, D-Peg was able to avoid buying moments before the bid or offer crumbled. We find a much larger percentage of D-Peg volume achieves PI when a Signal fire occurred during the trade versus when no Signal fire occurred. This suggests that the Signal plays an important role in generating D-Peg PI.
Another way to measure this Signal benefit is by using markouts, a common metric for performance that measures the change in price of a stock a given amount of time after a trade occurs.
For example, if you buy at $10.00 and the stock immediately goes to $9.99, you just “gave up” $0.01 of potential price improvement that you could have gotten if you had waited and traded at that more passive price. This scenario would generate a negative markout because the price is lower moments after the trade.
D-Peg is designed to avoid trading between the NBBO and the midpoint when the Signal is “on,” but M-Pegs can trade during that time, and, on average, 5% of M-Pegs trade within 2ms of a Signal fire.[6] The Signal helps D-Peg outperform M-Peg in markouts at all time horizons.
Because M-Peg trades both when the Signal is “on” and “off,” we can directly compare M-Peg “on” versus “off” markouts to solidify the Signal is behind this D-Peg outperformance. As expected, we find that markouts are significantly worse when the Signal is “on,” indicating missed PI.
When viewing this by spread, we find tighter-spread names generate the largest delta between D-Peg and M-Peg markouts (D-Peg - M-Peg). The Signal is powered by data from changes in many of the venues that make up the NBBO, so, generally, more venues at the inside improves Signal accuracy. Since tight-spreads mean there are more venues at the inside, tighter-spreads can also equate to improved Signal protection. As we can see, the larger markout delta in tighter-spread names means that the Signal enables D-Pegs to achieve PI more effectively in these names.
Conclusion
There is a lot to consider when determining which midpoint order type best fits a strategy.
Our analysis aims to shed light on the strengths of each order type and where their functionality differences are most impactful. In aggressive strategies, M-Peg’s queue priority is more likely to outweigh the benefits of D-Peg’s PI. In passive strategies, D-Peg’s PI and better markouts can be more impactful than higher probability of fill.
We have laid out a framework to compare D-Peg and M-Peg, but we know traders are also weighing when to use D-Peg and M-Peg versus other exchanges. From our vantage point, when comparing performance between IEX and other venues, it is probably best to compare D-Peg to what firms experience in ATSs, and M-Peg to traditional exchange midpoint posting. Also, while M-Peg and D-Peg may be stronger than one another in certain situations, both order types exist in an ecosystem where protection and performance are top priorities. This protection allows IEX to achieve the #1 position in stable midpoint volume versus the eight largest U.S. equity midpoint exchanges (see Appendix).
This analysis covers just a few ways that we compare order types and performance on IEX. We would love to partner more closely to find these stats for specific symbols, times of day, and strategies. If you are interested in running any of these numbers for specific situations, please reach out to us and we would be happy to help!
[1] For this analysis we look at M-Peg and D-Peg orders that rested on IEX. M-Peg and D-Peg can both remove liquidity on entry, making them act more like M-Peg and D-Peg IOC orders, and so we do not look at those orders in this blog.
[2] There is one situation where M-Peg could have a lower limit price than D-Peg, yet maintain its price priority because of its higher booked price. This would be if the M-Peg’s limit was less than the midpoint in a name but more aggressive than the bid (offer). In this situation, even if a D-Peg has the ability to trade at the midpoint which is more aggressive than where the M-Peg is resting, the M-Peg would still maintain price priority over the D-Peg. This is a fairly rare situation so for the sake of this note we will always assume that M-Pegs are resting at the midpoint.
[3] Simply, we are looking for times when M-Pegs executed over D-Pegs solely because of their price priority. We look at this data by spread bin and price bin to get the most comprehensive picture of when D-Peg is most susceptible to being queue jumped by M-Peg (i.e., in what situations might you want to use an M-Peg to make sure that you’re first in line to get a fill).
[4] In this scenario, “queue jumped” refers to situations in which an M-Peg execution came from an order that arrived after a D-Peg.
[5] “Non-Prop” Member firms refers to Agency or Full-Service Broker-Dealers. IEX classifications throughout are on a best efforts basis by Member firms’ trading sessions.
[6] IEX Market Data. Date Range: 10/01/2020–12/31/2020.
Appendix:
Far less of IEX’s midpoint volume trades when the quote is unstable, or within 2ms of a quote change, compared to other exchanges. When looking overall or by spread bin, IEX maintains the #1 position versus the eight largest U.S. equity midpoint exchanges.
Sean Spector
Tori Dewey