The 'Flash Boys' Exchange Is Growing Up | Hacker News Reader

The 'Flash Boys' Exchange Is Growing Up

bloombergview.com

82 pointsevanpw53 comments

I am deeply confused by IEX. It seems like anyone who knows anything about HFT thinks the narrative proposed in Flash Boys is completely ridiculous and mostly incorrect, including what Brad Katsuyama says. Is there something else interesting about IEX that people outside the industry don't really understand, or is it an entire company of people who have deeply flawed ideas about HFT and how it works?

Just because Flash Boys presents modern electronic exchanges incorrectly, does not mean that some people aren't worse off in the modern HFT world.

In particular, large block buyers, who previously got to take advantage of inefficient price discovery to get better prices have been negatively impacted. It is in their best interests to make it possible to buy a very large amount of traded products without the price changing.

IEX is built from the ground up for those market participants and none of them have any misunderstanding about how the market works. They just don't like it.

"IEX is built from the ground up for those participants". This is not really correct. Why? because a huge percentage of IEX volume is HFT, just like on any other exchange or dark pool . Their whole solution is nothing but a marketing gimmick. All the major HFT firms trade on IEX just fine

Can you expand on how IEX's delay isn't preventing HFT from succeeding? I don't really understand enough about the financial markets to know what you mean.

every exchange takes some amount of time to match orders. IEX is no different. Let me ask you this way. Why would you expect IEX delay to prevent HFT from succeeding ?

Most exchanges take considerably less time and there are a lot of HFT strategies that require low latencies in order to cancel certain trades before they happen.

without getting into the nitty gritty of how HFT works, suffice it to say that anywhere between 20 and 40% of IEX volume comes from HFT firms.

Can you please get into the nitty gritty? Why is anyone using IEX if there is so much HFT on it?

Edit: http://blogs.wsj.com/moneybeat/2014/08/11/debate-over-high-f... Seems like there is some debate. IEX claims 17.7% which is still higher than I expected.

why not use it? if it has liquidity and an attractive fee structure, why wouldn't any trader trade on it, HFT or not ?

I believe the discrepancy comes down to whether you count opposing counterparties indivdually or not. i.e. If 6 trades involve 6 non-HFT/non electronic market-maker (referred to as 'naturals') trading against 6 other naturals, and the other 4 involve naturals trading against resting HFT liquidity, you could either count that by counterparty that is HFT - (4 out of 20 counterparties) = 20%, or the number of trades that involve someone taking HFT liquidty - (4 out of 10 trades) = 40%.

So the 17.7% of HFT counterparties squares roughly with 30%+ HFT trades.

Second the call to get into the nitty gritty. I've read "Flash Boys" and consider myself reasonably aware of how HFT works but I'd like to know the technical aspects more in detail (understanding that some of that is, probably, proprietary). Any good resources?

Flash Boys is quite inaccurate with respect to how HFT works. Read this book instead

http://www.amazon.com/Introduction-High-Frequency-Finance-Ra...

Nitty gritty point: Volume is not equal to liquidity. Volume a low order values, provided by pinging, doesn't mean there's liquidity behind it that large block orders provide.

Here's your nitty-gritty:

If a participant places a large buy order into the IEX, this is going to affect the price of that asset, both at the IEX (obviously) and anywhere else it is listed, along with any correlated assets.

Now, the original participant will have bought its shares at the price it wanted at the IEX and be happy about it. However, market makers at other exchanges will not be aware of this upward price shock just yet. They will be blindly quoting sell offers at a price that had just been bought in bulk at the IEX. Whoever can receive the information out of the IEX 'magic shoebox' first and shoot that information over to the other exchanges will be able to make a profit. You can see that it doesn't matter how long the IEX delay is.

The original buyer will be the first to know of course, but they are unlikely to want to take on the risk of all available shares at that price at all exchanges. Thus, there will likely be some left for HFT to take.

Right...but the important function of IEX is that it protects the block order buyer or seller, and it seems to still be doing that, so it sounds like what they are doing is working.

protects from what ?

HFT requires low latency on their reactions, not the entire transaction. In IEX's situation, there are two major cases --

(1) A person decides to cross the spread on IEX first -- in this case, their order will be delayed by 350us (or whatever the shoebox delays), and no one in the world sees the result of their trade until the public data feed updates, in which case all participants are reacting off the same feed, same as any other US equity exchange.

(2) A person hits another exchange first, in which case HFT races to cancel their orders on IEX or cross the spread against another firm's orders on IEX -- yes, their action is delayed by 350us, but so is every other person, so you're otherwise racing to get into the message queue at IEX, and the artificial shoebox delay is irrelevant, since all market participants are equally delayed by it.

So while IEX's shoebox delays the latency on the result of an action, it doesn't make an appreciable difference in the typical HFT strategy. The primary difference is in (1), there is a higher chance that an unsophisticated large block trader (because a sophisticated one will have already set up the necessary delays to hit all exchanges simultaneously) will be able to hit market makers on both IEX and other exchanges before anyone has the time to react, thus being able to take large advantage of their own proprietary knowledge (their own order) some percentage of times more often. The only thing this results in a decrease in liquidity on IEX, but seeing as they're a dark pool/minor player anyways, the difference is negligible.

Yes, that makes sense, but the trade that is done on IEX is still not affected by HFT. So I understand how HFTs can still use IEX to make money, but I don't think that means the 350us delay is a marketing gimmick as vasilipupkin said.

It's a marketing gimmick insofar as it claims to somehow protect investors from HFT. As I said elsewhere, every exchange has latency of processing, theirs is 305micros higher. That doesn't appreciably change anything for anyone.

IEX claims that their delay means that large orders are able to be carried out without the price changing. If this is true, it would seem that IEX is not just a marketing gimmick and is in fact doing what they say they do.

If this is not true, I don't see why anyone would care about IEX at all, and why they would trade on it.

Yes, this claim is untrue. Large orders will move the market the way they always move the market and 350 micro delay won't change it one bit.

they offer lower fees than other exchanges - standard practice to steal away business from competitors. Traders will trade anywhere where there is liquidity.

I'm willing to believe you but do you have any source at all? Are you saying big block traders are staying on IEX because the fees are lower? That just doesn't make sense to me, the fee overhead is almost insignificant with big block trades. It's not a question of do the trades move the market, it's a question of can the trades fully execute before the market has moved. Obviously they move the market.

Look up their fees vs their competitor fees and their volume share of the equity market, and you'll have your answer.

I just want to make sure I'm understanding you properly: you are saying that big block traders are using IEX because it has low fees, not because the IEX system is preventing the pricing from shifting during the order, correct?

yes. Also "big block traders using IEX" is a big statement. IEX has 1.5% of US equity volume and at least 20% of IEX volume are not big block traders. So, they aren't really using IEX that much. But to the extent they are using it, it's because the fees are lower and also it's because it's always nice to put pressure on other exchanges via additional competition for liquidity

Right, that makes sense, but Katsuyama and Lewis both present IEX as a morally superior, cleaner stock market. It makes sense that large block traders would want a different way to do large block trades, but why not just see it as an alternative financial product that has some demand just like the regular stock market, instead of making it a moral issue? I guess my question is, do the people who work at IEX drink the Kool Aid?

I don't know enough about IEX to know whether the below is true. If IEX has this 'magic shoebox' mechanism in place, it might follow that they care about establishing a fair exchange in other ways too. There are hundreds of ways in which HFT firms profit from understanding the architecture of various exchanges. In some (many?) cases, HFT firms know more about the exchange architecture than the exchange does. The 'shoebox' addresses just one (mostly outdated) source of HFT competitive advantage.

Perhaps employees of IEX drink the koolaid because the exchange actively seeks to eliminate architectural areas of unfairness. The only thing publicized is the 'magic shoebox', but perhaps there are many more such features. Again, this is just conjecture.

The large block traders need someone to trade with.

By making it a moral issue, IEX can attempt to trick suckers into joining the market so that the large block traders can take advantage of them.

Watch this video (put out by IEX) and think about how it's trying to manipulate people: https://www.youtube.com/watch?v=v2OZkTesSx0

From some content presented in Flash Boys, a question:

Why would large block traders like to trade with someone pinging the market for 100 buy/sell orders, waiting for discovery, and then changing the price on other markets from information gained from these pings? If the market were rid of these pings, as the 350ms delay is intended to do?

Or to put the same question another way, why would a large block trader not like all trades being done at the same price or at a better or equal price on other exchanges (discovered by IEX) rather than some of the block done at the original price and the rest done at a higher price because 'liquidity' was provided by HFTs driving up/down prices on other markets and then selling/buying from the large block trader this 'liquidity'.

See the comment up thread from kasey_junk:

"In particular, large block buyers, who previously got to take advantage of inefficient price discovery to get better prices have been negatively impacted. It is in their best interests to make it possible to buy a very large amount of traded products without the price changing."

Throwaway because it would certainly not be kosher for me to comment on this, but:

There are people who work at IEX who do not "drink the Kool Aid" and have been critical in private industry discussions about Katsuyama's statements.

Additionally, IEX actively solicits trading in their pool from HFT firms. For one, these same people understand that it is difficult to match the quality of other modern US equity markets without those participants. Moreover, it's hard to build appreciable market share (% of marketwide volume traded) without electronic trading, which is how their business makes money and grows in valuation.

How is the book wrong?

Here is a detailed rebuttal:

http://www.amazon.com/Flash-Boys-Insiders-Perspective-High-F...

Some people (of which I'm one) don't have anything good to say about Flash Boys. On the other hand I don't think there are many people who have bad things to say about IEX or Brad Katsuyama. They are doing their thing and they'll succeed or fail based on their ability to compete with the other exchanges. It's hard to complain about that.

To lay the ground work for what's going on here....

In the US you can be an ATS(Alternative Trading System) or a full blown exchange. All dark pools are ATS's. This gives you a bit more leeway in terms of what you are allowed to do, ie hide orders, trade only at midpoint, trade only at certain volume's etc, even report trades in a slower fashion.

A full blown exchange has alot more scrutiny but comes with one huge caveat, all other exchanges must route orders to you if you have the NBBO(National Best Bid or Offer) as dictated by the SEC's REG NMS 611.

This means if an order to sell comes to the NYSE for 5.00 and takes out the entire first level forcing the NYSE to have a bid of 4.99 and there is still more volume in the order, the NYSE has to send the order to another exchange still showing $5.00 as its bid.

Since the exchanges make money when each share is traded they don't really like to do this( as a side note, this is how flash orders came about, exchanges "flashed the order to a select few market makers for 50 milliseconds before sending it to another exchange giving the market makers a chance to fill it, and keeping the commissions at the exchange, people got pissing about it and it no longer happens).

If IEX transitions from being a dark pool to an exchange this requirement to use their quote as part of the NBBO might mean more liquidity, and hence more profit, for them.

The issue with this is that IEX is the first proposed exchange, atleast to my knowledge, that has an intentional built in delay, meaning that its quotes are by design slow and stale as compared to the other US exchanges.

They are arguing to the SEC that since the NBBO is determined by the SIP(the system that aggregates all other exchanges quotes to determine what the NBBO is), which itself is slow, that this doesn't really matter much.

On the other side of the debate is a literal whose who of the finance world, not just HFT's, arguing the slipper slope defense, ala well if they delay 350 microseconds, what's to stop someone from having a 2 second delay or 1 day delay in showing quote.

I don't really have a dog in this fight but most people I know believe that IEX is going to lose this fight, though there is a loud contingent at the SEC itself that wants IEX to win to act as a test case for what the SEC can do to shape the market microsructure of the US cash equity markets.

As to becoming a place to list your company, I think Levin is right here... no one cares at all what exchange the company is listed on, even the company itself.

EDIT replaced ECN with ATS above

Do you know of they will be required to show book depth/qty or just best bid/offer prices?

This is actually a really interesting question. They would have to show size at the inside otherwise sweeping under Reg NMS would be impossible (you have no way of knowing how many shares to clear the protected quote). Also they already have a prop feed (IEX TOPS) that already shows size/price at the inside.

Under their proposed Rule 11.330, they only have two prop feeds which are the TOPS feed and a last sale feed. So no full depth feed. They would be the only US equities exchange without a full depth feed, but it looks like that's what they want to do.

> All dark pools are ECN's

This isn't quite true. Every dark pool is an ATS (alternative trading system). An ECN is a different kind of ATS that displays its quotes publically, unlike a dark pool. I don't think there are any ECNs left that have any kind of volume (since LavaFlow shut down in January).

Yikes, you are 100% correct. I appreciate the correction!

https://www.sec.gov/divisions/marketreg/mrecn.shtml

I blame wikipedia which says:

> ECNs are sometimes also referred to as alternative trading systems or alternative trading networks.

What would happen if you split the tick interval to three 30 second intervals, where during the first interval new orders are collected in private, the second serves as a 30 second waiting period, and at the third all previously private orders and matches/trades are made public and broadcast?

So collect (30s) -> wait (30s) -> broadcast (30s) -> collect...

Why 30s? Why not 10s? Or 1hr? Or 1day?

Sure. The point is that t >> worst case synchronization latency.

So what you're saying is that traders should play Diplomacy....

One of the better proposals to fix whatever market structure issues we have is to move to "frequent batch auctions", in which a micro-auction would occur every 30 seconds or minute, which would amount to a similar thing. I'm not aware of any research that discusses whether such a system would be workable in the presence of the existing continuous trading market.

There is research on this. It won't work the way you hope. Here is the problem:

What happens if in a single batch you have more buy orders than sell orders at a given price (or vice versa)?

One side eats up all the liquidity on the other side and the price moves appropriately in the next tick?

It would clearly have to be separate from the high frequency discrete market. (c'mon, don't call it continuous!)

On the side where all the orders aren't filled how do you decide who gets filled and who doesn't?

At equal prices, random selection. Anything else seems like it could be games.

Could lead to some funny movements though. Value aliasing! : )

If you think time priority orders are viewed as unfair, think what random matching would mean. It would mean that you could put in an order 6 months ago and watch it repeatedly never get filled as your price point was hit over and over again.

So lets say at tick one 1 see that only 50% of the buy orders were filled.

At tick 2 I want to buy 100 shares. But maybe I'll guess that only 50% will be filled this tick too, so maybe I should submit a buy order for 200 shares instead? This kind of game playing can lead to highly unstable outcomes.

I wonder why they're using a "magic shoebox" instead of a call market a la http://ai.eecs.umich.edu/people/wellman/?p=40.

Because call markets don't actually solve any of the supposed problems of HFT. Unless the call market has a global monopoly (an impossibility that we wouldn't want if it were available) venue arbitrage is still an issue.

Further, no one who proposes call markets ever talks about the matching algorithm in the face of order book imbalances. How you do tie-breakers can either still have speed advantages or has other probably worse disadvantages.

neither does their magic shoebox. It doesn't solve anything and HFT firms aren't meaningfully impacted by the shoebox

Reply on news.ycombinator.com