Thursday, 1 June 2017

IEX statistics for May: the devil is in the details

A reader may know by now that I'm not a huge fan of the IEX hubris and hypocrisy emanating from their general direction concerning their speed-bump. The smart HFT will continue to worry that such market structure debauchery will harm the market in the longer term even if it provides opportunity in the short term. HFTs rely on healthy markets. IEX's Dark Fader does not promote market health.

As you can see in the following table, IEX's dark and expensive share restaurant continues to darken. Not one day in May had over twenty percent of displayed volume trading as a proportion of total shares handled.

Lit / total handled
May 17.5%
April 18.7%
March 19.8%

IEX had an improved, albeit small, market share in May of around 2.2%. The devil is in the detail and in the following chart, the May detail devils are highlighted for your amusement or apprehension.
May details are devilish
(click to enlarge)
You'll see that in the unlikely event that a linear relationship was to hold, you would expect IEX to be completely dark if it was to grow to 8% of the market. This may be a consequence the SEC did not intend.

A more traditional view of lit volume traded versus total shares handled is the following chart:

(click to enlarge)

To me the surprise is not why IEX lit activity is below twenty percent, it is why investor naivety is such that IEX trades at all.

Sunshine continues to be a great disinfectant. Let's all hope for the sunrise as the IEX dark moon rises.

Happy trading,



Older IEX related meanderings:


  1. Not sure I follow the "IEX is bad. See, most of their executions are dark" logic:

    - If IEX changed to a maker/taker model and offered the biggest rebate on the street, they'd probably see a huge rise in their lit percentage. They'd still have the speed bump. But would you all of the sudden consider them a "good" exchange just because their lit percentage improved? My guess is no, so I'm not sure the logic follows.

    - Furthermore, isn't the problem here really broker conflicts of interest with respect to Rebates/Payment for Order Flow? I've read your defense of rebates as well as your criticism of PFOF. From what I understand, you are okay with ECN 'Rebates' because the poster of liquidity takes on more risk than a taker. That makes sense in an HFT world. However if a broker acting as an agent with no risk, is making routing decisions based on ECN rebates/fees that they do not pass back to the client, isn't that strictly a payment for order flow situation (which you seem to be opposed to)?

    - Isn't IEX only "relatively" expensive because they don't pay participants to post on them as other exchanges do (Free vs. say, -$.0020)? Would you also say routing to IEX vs. traditional maker/taker models to take liquidity is relatively cheap (generally FREE vs. $.003/share)? If IEX is relatively cheap to take, than anyone taking should have it near the top of their routing table. Since there is very little lit IEX liquidity (see your stats), it's near the top of most routing tables, and it's now a protected quote, shouldn't brokers with best-ex requirements be posting lit on IEX all the time? After all -- it's not costing them anything.

    - So you may believe IEX is "bad", but is this data really indicative of that? I think you've actually exposed a different problem...

    1. 1/2

      Thanks Curious George for the comment.

      I'm not fond of dark in public. Price discovery is harmed. Like passive index funds, parasitic behaviour is OK in moderation but needs to be regulated so the parasite does not destroy the host. Jack Bogle suggested a safe upper limit of 75% for passive indexing. IEX is over Bogle’s limit.

      I'm harsher in my view. I don't even like icebergs. I see them in the same class as spoofing. A lie about your size.

      I do think DPEG & PPEG with the dark fading algo from IEX is innovative. However, it is the wrong kind of innovation. It thwarts asset manager, brokers, and traders applying their own innovation due to the speed advantage IEX give themselves and no other. IEX algos are special & suck a bit due to their attempt at one size fits all and the false positive rate. Innovation that kills innovation.

      DPEG is interesting. I'm not a fan of midpoint matches as giving up half the spread goes against my DNA. A limit may be placed on an IEX DPEG that may limit you to BBO and thus give yourself an advantage over others due to the dark fader's speed advantage. IEX has gone on & made this explicit by adding the dark fading algo to the primary peg so you don't need to do this in a strange roundabout way. In their docs they call it stepping up to the price when things are stable rather than calling it a dark fade which is the more correct term as markets are stable more often than not. Misleading conduct from IEX.

      All IEX orders bleed leaks to the SIP. For dark orders the quotes will not bleed as they are dark, but the trades will leak. Many care about this, some will not.

      I don't buy into the argument rebates are the only thing holding them back. On the lit side, being a slow exchange with leakage of information to competitors is more of a concern. On the dark side, a round trip of 18 mills is just plain expensive, and if they offered a big rebate with the same nett on a round turn, they'd still be expensive.

      I'm agnostic about rebates on the make or take side. They are just one of a few ways to favour types of customer behaviour. There are only so many things you can do to compete with sub-pennies. Attracting price making to an exchange can be hard & rebates are one such mechanism.

      Not passing on rebates is different to not having rebates. Such bad behaviour should be exposed. I'm not a fan of PFOF but it serves the market well with benefits going to the customer, broker, and wholesaler. Hard to compete against it with the sub-penny rule in place.

    2. 2/2

      Posting lit on IEX is dangerous due to slowness creating inherent risks, SIP information leakage, & the expense of trying to get fair access close to their pop. It is messy. You're better off with simpler & fairer access at Nasdaq or Bats in their co-los.

      Taking lit liquidity is cheap at IEX as you describe, if you enjoy giving up the spread & quotes actually remains in the hall of mirrors once you navigate the speed-bump. Plus you have the opportunity for price improvement from the dark pegs. For a 0.01 spread, always paying 0.01 or 0.005 for mid, is expensive in my book, but yes it is cheaper than paying 0.013 for an alt take of the price. Your competitors listening on the SIP will know about your trade before you. That is impolite. Not a sustainable way to trade in my book.

      The badness of IEX comes from a number of areas: their thwarting of price discovery, added risk of a slow exchange, preventing customer innovation, complex order types, misleading statements around their fading, lack of fair co-lo, information leakage via the SIP, the expensiveness of trading, reliance on orders that prevent trading during times of change, and their harassing treatment of some market commentators such as Mr Weisberger. Combine that with the narrative around their origin story, i.e. RBC's Thor, speed bumps, and "puzzle master" simple order bias, that will save the world & they just annoy anyone with a decent recollection of market history.

      IEX is more cult than scientific endeavour.

  2. 1/2 Your points are well taken and you clearly have a very strong grasp of the concepts at play -- well beyond my current understanding (which is why I like your blog so much). That being said, nearly all of your points seem to reflect the HFT sentiment. IEX, flawed as it may or may not be, has made it clear that their aim is to improve the experience of investors who don't have the resources to effectively navigate the NMS system that exists today. So I'd ask you to consider the experience of say, a retail trader rather than HFT participant in your analysis:

    SIP Leakage -- You bring up this point a lot and I believe it's the most misguided of the IEX criticisms. Do you agree that IEX router basic ensures that you execute against all participants at the NBBO before anyone can back away? Participants with enough resources can do that on their own (probably better), but many participants cannot, and it is a decidedly better router than Nasdaq, Bats, etc. at the NBBO. Just look at the fill rates. So can some HFT see my fills before I do when I route using IEX? Maybe, but who cares? I already got my fill (or at least a better fill than if I missed some participants at the NBBO). Now, let's say for some reason this leakage happens before my fill is done. Do you think this "leakage" could be profited on in some way? If I can be gamed from SIP leakage, don't think you think I'd be much worse off if I was leaking information via a direct feed? Since some exchanges have attribution on their direct feeds, Shouldn't this be a much bigger concern?

    Slowness -- My executions come back with a timestamp in seconds. What's a few hundred mics? Not to mention wholesalers take much longer to fill, and haven't they been known to price trades off stale quotes?

    Expensive Cost -- I pay $X a trader to my broker who doesn't pass back ECN fees. I understand that they aren't going to post me on an inverted rebate venue, or go out of there way to pay a $.0030 fee, but shouldn't they be willing to post/take for Free? The only cost to me in this sense is opportunity cost (not adverse selection from my perspective), so shouldn't my broker be trying to maximize the likelihood of my fill?

    Complex Order Types -- Is this really the exchange you want to critique for having complex order types? BATS has thousands of order combinations across four exchanges. Shouldn't we start there?

  3. 2/2 Misleading Statements -- I believe ARCA has IOIs "On" at the port level by default on their basic routable order. NSDQ STGY strategy sends to dark pools but good luck deciphering that from their description. Both pass back those executions at the full route rate. What are your thoughts on BATS "feedback" mechanism? Citadel, well you know that one. So is the DPEG a midpoint that backs away? I think so, but the rules stipulate that they have to call it a primary peg with discretion. Again I think if misleading statements is bothering you, IEX should be way down your list, don't you think?

    Lack of fair co-lo -- I'm actually completely loss on this one. What do you mean by this? Even IEX's own router is outside the POP. How is that more unfair than selling real-estate to the most wealthy participants and forcing everyone else to wait outside for the data?

    So As for the science, IEX first wave fill rates are constantly over 99%. A recent paper highlighted several benefits of posting on IEX vs. their competitors and exposed the conflicts of interest that prevent brokers from posting there (this was written by IEX employees, for full disclosure, but I haven't seen a rebuttal). Most notably BATS' own stats indicate that IEX has the best execution quality of the exchanges. Are we to disregard these facts and studies (a.k.a. science) because rebate seeking brokers won't post on IEX? Or because of a theoretical, post-fill SIP leakage that I believe Mr. Weisberger would say is not actionable? Do you think Citadel, as a large user of IEX, is drinking the IEX juice?

    I agree with you that too much dark liquidity is bad for the market, and clearly IEX is very dark. I just don't necessarily think IEX, a pseudo-utility, is necessarily to blame for that. So to label IEX as "bad" as a result of the stats above, while ignoring much of the positive data, seems unfairly critical to me. Do you think that makes me a cult follower?

  4. 1/2

    Retail is an interesting conundrum. NMS actually serves small direct retail pretty well. Brokers handled this remarkably efficiently. PFOF is a big winner there. Place a trade at TD Ameritrade, it flows through to Citadel as a wholesaler and you often get a price improvement with reviewable SEC audit reports if you care about the details. The opaqueness and complexity is concerning but in the SEC we trust.

    IEX itself does not argue so much about direct retail anymore. IEX makes an argument for retail via institutional funds or asset managers and that is a reasonable perspective, but it is a different perspective to direct retail as it becomes more about systemic or algorithmic efficiency.

    SIP leakage – routing is a different issue. As you point out there are plenty of good routers out there. IEX doesn’t publish enough valid information to make informed decisions. For example, KCG’s previous work has shown that sometimes more than 100% can be filled at some exchanges but IEX caps this out in their statistics, so it is hard to make a judgment. IEX publishes to the SIP before the speed-bump, so other participants will learn of your trades before you get your fill, but this is not a big deal if your routing does a latency adjusted simultaneous hits, so it depends. It means you have to have equipment at Mahwah, Carteret, & Seacaucus if you want to maximise the use of information. Less than ideal for those that care. Many asset managers with large orders for their retail clients bases would care.

    Slowness – see the Speed-bump 101 post. If PFOF price off stale quotes they get fined. That has been illegal since Nov 2015 thanks to a particular footnote.

    Expensive cost – broker has best execution obligations, if they aren’t fulfilling these, get a different broker.

    Complex order types – this is a real problem for all exchanges. I like ouch over ufo at Nasdaq, but all exchanges have this problem at various venuse and places in their stacks, especially when routing options invade the scene. The point here is that in Flash Boys the IEX “puzzle masters” promised to not go down the complex order path. They did. It is an industry wide problem, but it is pertinent that IEX broke their simple order type promise.

    Misleading statement – they always bother me.

    1. 2/2

      Lack of fair co-lo – co-lo has been one of the great levelers in the trading playground. It was better for HFTs before they came into being as an HFT could gain clever geographical edges. HFTs generally prefer co-lo as they then will not be suspicious someone has got an edge over them. It works both ways. IEX says they will never do co-lo implying it is evil when it is one of the great fairness measures. IEX is just plain weird on this one. You need to be as close as possible to their pop, so you need to locate in NY4 or NY5 to as close as possible. You need co-lo, it is just that someone else owns it. Worse, because of their SIP leakage, you’ll also need co-lo at Mahwah and Carteret with RF links to minimise your information disadvantage thanks to IEX’s poor trading architecture. IEX is a real mess on this one.

      Bats stats on the IEX best execution are flawed for a couple of reasons. I’ve written about this previously. Firstly, they include dark orders so the midpoints are often what is being measured and someone’s gain is another’s pain. This is too often glossed over. It is that old immediacy versus patience argument yet again. Secondly, due to the dark component which includes dark fading protected orders, they exclude the times of stress of tick changes which prevents a like to like comparison. If you exclude the hard edges and just do the soft centre, of course things will look better. It’s a d’oh. The IEX trade statistics thus become meaningless without more information. This is not peculiar to IEX, there needs to be much better disclosure industry wide on not just trades and quotes but routing too.

      IEX is not to blame for taking advantage of SEC’s interpretation of NMS rules. I think all exchanges should add dark Fraken-pool to their repertoire as it is allowed. I don’t like this, but parasitic darkness can add certain types of efficiency. Having a better alternative to IEX’s Franken-pool would be good. Fundamentally I’d like them all to go away, but I can’t see SEC wising up and kicking IEX’s Franken pool out, so the next best thing will be competition on that front.

      NYSE American has an equally flawed architecture, as it is a clone, but at least it may compete on price and provide better efficiency. It will still leak to the SIP. It will also have IEX’s dark fading flaws if it clones it with a similar logistic regression adjusted for its location. SNAFU.

      I don’t know what a cult follower looks like. I do know that the markets are hard to understand. There is much more to it that I don’t understand and also much that many of us trader types keep secret, often widely held secrets. I worry about what I know I don’t know and what I don’t know I don’t know. I have much to worry about in both categories.

      Reg NMS is pretty good by world standards. It is a bit flawed but really quite excellent, especially for direct retail. Look across the screen to the FX, bond, and other OTC markets if you really want to find ugly markets. Then again there is always the soft-ponzi of Bitcoin ;-)...

  5. Thanks for improving my understanding. Definitely appreciate your perspective on fair co-lo... hadn't thought about it that way before. In the end, I certainly think IEX has flaws as you point out, but I don't think it's as binary as IEX is bad and BATS is good, but it seems we agree on that.

    My one hangup is still here:

    "Expensive cost – broker has best execution obligations, if they aren’t fulfilling these, get a different broker"

    I think your original stats support that hypothesis that most brokers aren't fulfilling best ex, not that that IEX is a bad exchange. Perhaps there are risks to posting on IEX vs. other exchanges, but don't you think a large percentage of participants across the industry would be willing to overlook those risks if IEX kicked them back a rebate?

    Have a great weekend.