Oct 04, 2021
One year ago this month, IEX Exchange launched “D-Limit,” a new order type that was designed to do something that hadn’t been done in many years: make the quality of displayed trading better.
For years, displayed trading has been on the decline, as exchanges have created an environment in which institutional investors and market makers have their orders adversely selected, leaving them with a poor execution and more and more reason to trade in the dark or off-exchange instead.
In the year since launch, D-Limit has flipped that story on its head. As broker-dealers have experimented with this order type, they have been using it more and more as they see the results for themselves.
In most of our analysis of D-Limit performance, we’ve focused on markouts, showing that D-Limit was achieving its primary objective of reducing adverse selection. On average, D-Limit trades were less likely to have immediate “buyer’s remorse” than displayed trades on other exchanges.
In this blog, we want to share more data around a unique aspect of D-Limit performance — pre-trade Price Improvement — and demonstrate that, a year later, D-Limit remains ahead of the pack.
Usually, we think of price improvement as something that can only happen to aggressive, taking orders — like if the offer was $11 and I tried to buy it at that price, but ran into a dark order and traded at $10.90. Historically, that dynamic did not apply to resting limit orders since, by definition, the order’s limit price was the best you could do.
Not so with D-Limit, where traders can end up buying for less or selling for more.
D-Limit orders initially rest at their limit price. But if the IEX Signal fires — indicating that the price is unstable — the order is repriced to 1 tick outside that unstable price. And it can do this repeatedly if the price keeps crumbling. Ultimately, a D-Limit order can end up selling for much more or buying for much less than its original limit price.
D-Limit orders sometimes trade at a price better than the one at which they were initially resting.
This unique aspect of D-Limit’s functionality has proven over the past year to be particularly useful for institutional brokers looking for better displayed trading results. Several of them have created strategies predicated on “leaning into” the PI aspect of D-Limit, often keeping their orders on the book through multiple re-pricings and “letting the winners run.”
Okay, but how often is this really happening? Does “pre-trade price improvement” really impact performance overall, or is it just a nice surprise when it happens?
The answers are “a lot” and “yes.” Let us explain.
The first thing to note is that pre-trade price improvement is completely additive to markout performance, which is the traditional way to compare displayed trading on different exchanges.
Markouts look at what happens to the price of a stock after the trade occurs. However, the race doesn’t start when the trade goes off. You can take the results of the markout “race” that happens after the trade and just add those “in the money ticks” to the results. It’s like getting a huge head start in that race.
With pre-trade price improvement factored in, D-Limit outperforms lit trading on every other exchange by an even larger margin. We show this data expressed both in mils and % of spread, to control for the different types of symbols that may be traded across venues. Additionally, we include the trading cost (conservatively assuming top-tier fees and rebates) to truly provide an “all-in” comparison.
Source: NYSE TAQ & IEX Market Data
Source: NYSE TAQ & IEX Market Data
In Q3 2021, ~4% of D-Limit volume benefited from pre-trade PI, and when it did, the savings amounted to over 2 cents per share. That adds up to $2.5mm in aggregate, $1.9mm of which was realized by institutional brokers and their clients.
It’s also important to note that we are only calculating price improvement that we can trace back to the original re-priced order. There are also scenarios where a broker’s order was posted on IEX, re-priced, and then replaced by a new order at a new price. Trades belonging to that new order (also at a better price than the initial limit) would not be part of the price improvement calculation, since we wouldn’t have the visibility to string these orders together. Therefore, in practice, it’s possible that the price improvement received thanks to D-Limit is even higher.
With these results, broker-dealers are discovering that incorporating D-Limit into their strategies can make a big impact on their aggregate results.
Because passive orders on exchanges have typically been adversely selected, many firms have focused on minimizing time to fill and maximizing fill rates.
But now that there is an order type designed to prevent adverse selection for those orders, there is an opportunity to look at a broader set of performance factors, taking into account pre-trade price improvement, understanding how this functionality is balanced with fill rates,3 and optimizing to make it work for you. This is why, although we are primarily showing D-Limit’s outperformance via markouts, we ultimately believe that it can help brokers and their clients achieve better parent level performance.
One year in, we’re seeing the fruits of the work and effort that broker-dealers have done to maximize this opportunity for their clients. We’ve had numerous brokers try D-Limit, which usually involves running experiments and A/B tests to see if D-Limit delivers better performance, both on a trade-by-trade basis as well as at the parent-order level.
We consistently receive feedback that D-Limit is delivering superior results and look forward to working with more of our members to continue growing adoption.
If you are interested in more information about D-Limit and its performance, please reach out to your IEX Exchange contact.
 A markout calculates the magnitude and direction of a market move after you trade. Often, they compare the price at which a trade occurred to the midpoint of the NBBO at various points in time. For example, if you buy 100 shares at $10.00 and 1 second later the market is $10.00 x $10.01, you have a markout of 50 mils per share. Often, markouts are used to measure adverse selection, or buyer’s remorse. For example, buying (selling) right before the price goes down (up) usually means the market timing for the trade was sub-optimal. Not surprising, adverse selection is very common for trades that take place while the Signal is on.
 Here is an example to bring this to life. Market is 10 x 11 cents. There is a displayed bid posted at 10 on Exchange A and on IEX Exchange. Our Signal fires, and D-Limit moves to 9. A trade goes off at 10 on exchange A and, after the market moves down to 9x10, a trade goes off at 9 on IEX. Let’s even say that after that trade, the market keeps moving lower and becomes 8x9. The markout on Exchange A is -50 mils (10 vs 9.5). The markout on IEX is -50 mils too (9 vs 8.5, the new midpoint). However, the order that traded on IEX had an original limit of 10 and ended up executing at 9. That is a penny of price improvement completely separate from the markout experience. In both instances, there was a trade on the bid, after which the market moved down. However, in the case of IEX, that trade was executed at a penny better, something the markout alone wouldn’t show.
 The price improvement opportunity provided by D-Limit can also be measured against the lower fill rate “opportunity cost” resulting from the repricing of D-Limit. For more on this, please see our blog post “D-Limit Performance & the Fill Rates Race.”