Wednesday, 27 June 2018

IEX updates

iex_update.md IEX update


It’s been at least fifteen minutes since I complained about IEX. It must be time for another meander.

IEX starts pricing off the SIP

Mr Brad Katsuyama’s misleading of investors would be amusing if it wasn’t so dangerous. His false and misleading statements have undermined confidence in the US National Market System over the last few years.

Do you remember the classic Lewis, Katsuyama, O’Brien CNBC debate where Mr Katsuyama accused BATS Mr William O’Brien of pricing off the SIP and disadvantaging his customers? Mr O’Brien denied it as it slipped his mind in the heat of the moment the acquired Direct Edge did that. O’Brien was no longer at BATS four months later. Despite Mr Katsuyama’s continued misrepresentations, he remains employed.

Well, IEX has decided to start using the SIP to contribute to its pricing.

File No
SR-IEX-2018-10
Date
2018-05-09
Description
Proposed rule change to amend rule 11.410(a) to update the market data source that the Exchange will use to determine the Top of Book quotation for NYSE National in anticipation of its planned re-launch.

… the Exchange proposes to amend and update the table specifying the primary and secondary sources for NYSE National (“XCIS”) in anticipation of the planned re-launch of XCIS on May 21, 2018.9 As proposed, the Exchange will use securities information processor (“SIP”) data, i.e., CQS SIP data for securities reported under the Consolidated Quotation Services and Consolidated Tape Association plans and UTDF SIP data for securities reported under the Nasdaq Unlisted Trading Privileges national market system plan, to determine XCIS Top of Book quotes. No secondary source is proposed to be specified as SIP data will be used exclusively.

Perhaps it’s not that important. In the context of Mr Katsuyama’s misdemeanours, perhaps it is.

I wish IEX would apply their own Rule 3.14.0 to staff, including Mr Katsuyama:

Rule 3.140. False Statements
No Member or applicant for membership, or person associated with a Member or applicant for membership, shall make any false statements or misrepresentations in any application, report or other communication to the Exchange. No Member or person associated with a Member shall make any false statement or misrepresentation to any Exchange committee, officer, the Board or any designated self-regulatory organization in connection with any matter within the jurisdiction of the Exchange.

IEX keeps markets simple by making its pricing more complex

IEX long ago promised to keep order types and pricing simple. This was to ensure your average punter would not be disadvantaged from not understanding the complexity. An admirable goal IEX has failed at. IEX continue to increase complexity making life worse for your average investor.

File No
SR-IEX-2018-09
Date
2018-04-20
Description
Proposed rule change to charge a more deterministic fee of $0.0003 per share for executions at or above $1.00 that result from removing liquidity with an order that is executable at the far side of the NBBO.

IEX added a Spread-Crossing Remove Fee at 3 mills, $0.0003 per share, to ameliorate part of the rape and pillage they introduced with the Crumbling Quote Remove Fee Indication of 30 mills. The Crumbling Quote Remove Fee Indicator adds great uncertainty to trading at IEX. You don’t know if you’re going to be hit with a high fee or not. IEX’s crummy crumbling quote voodoo is often wrong causing massive fee charges to appear on your statement for no real reason.

It’s quite terrible you can’t determine the crummy crumbling fee a priori. You can’t even calculate with rigour ex ante.

You get hit with a high fee for a trade. Is it correct? You have no way of knowing as the crummy crumbling quote is dependent on the feeds IEX gets internally within their own parameters of delay, jitter, and determinism, or lack thereof. You can’t deterministically parse chaotic information you don’t have. You can’t check your own fees. Does the SEC check the fees every month? I don’t think so.

IEX had to refund customers for incorrectly charged fees earlier this year. No one knew.

IEX fees - can’t predetermine them, can’t audit them
click to enlarge

At least this new fee introduced is lower than the expensive dark fee you get pinged with if you stumble over a midpoint in the dark.

I did like this line from the fee reasoning on page 14,

Furthermore, the proposed Spread-Crossing Remove Fee is substantially lower than the fee for removing liquidity on competing exchanges with a “maker-taker” fee structure (i.e., that provide a rebate to liquidity adders and charge liquidity removers)

One sided misrepresentation from IEX again. The nett fee IEX takes is much higher than other exchanges. IEX is the expensive dark exchange.

Clarity in pricing

The rules were so clear, IEX had to issue on the 20th June a further update to add clarity to the simple rules they have, effective 1st July. Yes, Minister.

File No
SR-IEX-2018-11
Date
2018-06-20
Description
Proposed Rule Change to Modify the Structure of its Fee Schedule and Make Several Conforming and Clarifying Changes, pursuant to IEX Rule 15.110(A) and (C).
Let’s check out the new simple and clear fee structure:

IEX fees - you should check out the fine print
click to enlarge

The fee schedule is quite misleading as for some of your stocks the quote instability charge that you pay may be more likely when the quote is stable rather than when it is unstable, according to their own formulae. If you put it in capitals and pervert the meaning of Quote Instability by subverting its normal definition, thereby turning it into your own proper noun, then you have a a nice marketing message. Science be damned.

IEX aren’t the only offender in the complex pricing stakes, but their impossible to know a priori, or audit ex ante, CQRFI fee takes it to a new level in the market of the absurd.

IEX crumbling quote indicator changes

IEX changed the crumbling quote formulae a while ago as they do from time to time. It is still a bad idea and a bad formula. I think many undergraduates could come up with a better toxin in a few days if they were appropriately directed with the same data. It is what it is:

File No
SR-IEX-2018-07
Date
2018-04-03
Description
Proposed rule change to amend Rule 11.190(g) to incrementally optimize and enhance the effectiveness of the quote instability calculation in determining whether a crumbling quote exists.

One funny thing about this update is the sheepish admission by IEX they have been indicating quote instability on one side it is infact the other side of the NBBO that is crumbling. And yes, you will have paid an excessive 30 mill fee for trading on the stable side of the market. You couldn’t make this stuff up if it wasn’t true.

Digest this bit first from page 8,

When the System determines that a quote, either the Protected NBB or the Protected NBO, is unstable, the determination remains in effect at that price level for two (2) milliseconds. The System will only treat one side of the Protected NBBO as unstable in a particular security at any given time

Do you see the problem? The market could step up and then switch back around pretty easily, even in just 100 microseconds. You could end up with a variety of stocks locked into Quote Instability mode on opposite sides of the market for two milliseconds. Crazy stuff, right?

IEX was woken a little from their slumber and decided that if their often false crumbling quote switches around then they will too rather than keeping it locked at the wrong side.

The indicator is still often false, but at least that is something.

IEX describes the changes starting on page eleven,

Rule 11.190(g)(1) provides in part that when the System determines that a quote, either the Protected NBB or the Protected NBO is unstable, the determination remains in effect at that price level for two (2) milliseconds. The Exchange proposes to revise the time limitation on how long each determination remains in effect, and reorganize certain existing rule text for clarity. As proposed, when the System determines that either the Protected NBB or the Protected NBO in a particular security is unstable, the determination remains in effect at that price level for two (2) milliseconds, unless a new determination is made before the end of the two (2) millisecond period. Only one determination may be in effect at any given time for a particular security. A new determination may be made after at least 200 microseconds has elapsed since a preceding determination, or a price change on either side of the Protected NBBO occurs, whichever is first. If a new determination is made, the original determination is no longer in effect. A new determination can be at either the Protected NBB or the Protected NBO and at the same or different price level as the original determination. Based upon our analysis of market data, as described above, the Exchange believes that changes to the time limitation would provide for a more dynamic methodology for quote instability determinations thereby incrementally increasing the accuracy of the formula in predicting a crumbling quote by expanding the scope of the model to additional situations where a crumbling quote exists at a different price point, or again at the same price point within two (2) milliseconds. For example, suppose that the NBBO is currently $10.03 by $10.04 in a particular security, and the System determines that the NBB is unstable. This determination goes into effect, with an expiration time set two (2) milliseconds in the future. Now suppose that one (1) millisecond later, the NBB falls to $10.02 and the System determines that this new NBB is unstable. As proposed once the System makes a new determination that the NBB of $10.02 is unstable, even though the prior determination at $10.03 has not expired, the new determination will overwrite the old determination, and its expiration time will be set to two (2) milliseconds in the future from the time of this determination.

The dark and expensive exchange that should be an ATS and not an exchange, is still deploying some wriggle room chicanery with their 200-microsecond from the proceeding determination thing, which they really shouldn’t need given they have a 350 microsecond delay in their stupid system.

And they have to change some of the variable definitions in their dumbly inaccurate, often false, crumbling quote indicator,

The Exchange proposes to revise five of the quote stability variables currently specified in subparagraph (1)(A)(i)(b) of Rule 11.190(g). Specifically, the Exchange proposes to revise variables NC, EPosPrev, ENegPrev and Delta to be calculated over a time window looking back from the time of calculation to one (1) millisecond ago or the most recent PBBO change on the near side (rather than on either side), whichever happened more recently. Based on our analysis of market data, as described above, the Exchange identified that for each variable, considering the maximum change over the time window defined in this manner is a more accurate indicator of a crumbling quote than the current approach. Similarly, the Exchange proposes to revise variable FC to be calculated over a time window looking back from the time of calculation to one (1) millisecond ago or the most recent PBBO change on the far side (rather than on either side), whichever happened more recently. Based on our analysis of market data, as described above, the Exchange identified that for this variable, considering the maximum change over the time window described in this manner is a more accurate indicator of a crumbling quote than the current approach.

You also have to keep in mind IEX is acting in the future due to their internal lack of 350 microsecond delay implying that one millisecond look-back is really 650 microseconds of look-back from their lack of perspective.

IEX should be forced to change the term from “quote instability” or “quote stability” as that is a clear misrepresentation of the facts and misleading, regardless of their intention.

Thanks again for signing up for our “Stability Operation.” Just because you’ll be losing two perfectly good legs and be confined to a wheelchair doesn’t matter. Many people may need a wheelchair in the future, so we’re enhancing the stability of everyone. No, we don’t think it is at all misleading. Being more stable on a flat surface that takes a wheelchair will be a boon to everyone. Yes, even gymnasts and rock climbers. It was right there in the EULA you clicked the OK button on. Stop complaining. The regulator approved it.

Double squeak from double speak indeed.

Anyhow, just to sound Presidential, here is the new failing, often fake, Quote Instability Factor and friends from the failing IEX’s new logistic regression formula:

Definitions
Spread
Protected Best Offer minus Protected Best Bid
Protected Quotations, Protected NBB, Protected NBO, Protected NBBO
includes quotations from not all of the exchanges, just these ones: XNYS, ARCX, XNGS, XBOS, BATS, BATY, EDGX, EDGA.
  1. N = the number of Protected Quotations on the near side of the market, i.e. Protected NBB for buy orders and Protected NBO for sell orders.
  2. F = the number of Protected Quotations on the far side of the market, i.e. Protected NBO for buy orders and Protected NBB for sell orders.
  1. NC = the number of Protected Quotations on the near side of the market minus the maximum number of Protected Quotations on the near side at any point since one (1) millisecond ago or the most recent PBBO change on the near side, whichever happened more recently.
  2. FC = the number of Protected Quotations on the far side of the market minus the minimum number of Protected Quotations on the far side at any point since one (1) millisecond ago or the most recent PBBO change on the far side, whichever happened more recently.
  3. EPos = a Boolean indicator that equals 1 if the most recent quotation update was a quotation of a protected market joining the near side of the market at the same price.
  4. ENeg = a Boolean indicator that equals 1 if the most recent quotation update was a quotation of a protected market moving away from the near side of market that was previously at the same price.
  5. EPosPrev = a Boolean indicator that equals 1 if the second most recent quotation update was a quotation of a protected market joining the near side of the market at the same price AND the second most recent quotation update occurred since one (1) millisecond ago or the most recent PBBO change on the near side, whichever happened more recently.
  6. ENegPrev = a Boolean indicator that equals 1 if the second most recent quotation update was a quotation of a protected market moving away from the near side of market that was previously at the same price AND the second most recent quotation update occurred since one (1) millisecond ago or the most recent PBBO change on the near side, whichever happened more recently.
  7. Delta = the number of these three (3) venues that moved away from the near side of the market on the same side of the market and were at the same price at any point since one (1) millisecond ago or the most recent PBBO change on the near side, whichever happened more recently: XNGS, EDGX, BATS.

Now, it’s relatively easy to get a feel for what this is doing without thinking too hard. If a DPEG is a buy order and the market’s protected bid quote volume is large, then N is big. The QIF formula will use resulting in one over a big number, which will result in a near zero, or low, number. It will be likely be less than 0.19, IEX’s fake Quote Instability Factor threshold. That is, lots of volume on your side of the market means the market is stable and not in a position that indicates a crumbling quote. Quite natural.

Though please remember when referencing the formulae, now is 350 microseconds into the future, as per the use of non-speed bumped data in a speed bumped exchange.

From the IEX Signal 2.0 whitepaper, IEX told us,

“On our example day of December 15, 2016, our current formula resulted in about 1 million true positives and 975,000 false positives. This new candidate formula would have produced about 2 million true positives and 2.1 million false positives.”

That’s a stupid number of false positives. It is good if you can fleece charge your customers big fat fees when you’re wrong 😇 👀.

How false is this new calibration of the fake indicator? For which stocks? IEX does not tell us. We can be sure that such a synthetic silly sausage shop standard doesn’t fit all stocks the same. QIF will be more fake for some than others. One size does not fill all. The universe of stocks IEX hinders your trading with is broad. For example, liquid and illiquid stocks behave quite differently. Somehow I don’t think you’re suprised.

Conclusion

You may be surprised to hear I’m not a fan of a dark and expensive exchange. Especially one where you can’t work out your fees in advance or even after you’re charged.

IEX impedes price discovery and harms public markets.

Fortunately, despite the harm IEX has done to the NMS, it’s marginal impact is economically small due to its small market share. Unfortunately, the qualitative burden of IEX is large due to the noise and misleading statements from IEX the industry has had to try to counter. The additional execution infrastructure costs the industry has had to bear is wasteful at best. IEX is a frustrating little baby who thinks its soiled nappy is innovation.

All of this for an exchange that leaks information to the well connected trader via both the SIP and significant clients’ holdings. No other exchange comes close to the level of bad latency arbitrage exposed by IEX’s displayed market and dark executions.

IEX also gives a greater advantage to faster traders than other exchanges as it does not have a fair co-location facility. Shorter cables and better positions help.

The SEC should send back IEX to the land of the ATS, where it belongs. There’s no shame in that. Many a good ATS serves a useful purpose. Then, perhaps, we could all get along.


Happy trading,

–Matt.


No comments:

Post a Comment