This book is a ~160 page tirade against the book “Flash Boys” that has captured everyone’s mindshare with the marketing slogan – ”U.S. stock market is rigged”. The author, Peter Kovac, has worked with a HFT firm for eight years and claims to be an industry insider. Since “Flash Boys” was basically anti-HFT book, it is natural to expect someone from the HFT to criticize the book. So, there we have Peter Kovac with a book length treatment that has similar title with Lewis book but a different tagline.
For a person who has read “Flash Boys”, it is likely that one might be dazzled by Michael Lewis story telling capability. Who doesn’t love a David vs. Goliath story, i.e. Brad &Co vs. Wall Street story? But there is a possibility that one might be too carried away by the arguments in the book. Peter Kovac’s book delves in to the story, strips off the fluff from “Flash Boys” and critically analyzes numbers, examples, arguments, protagonists from the book.
In this post, I will briefly summarize the main points from the book:
Introduction: The Dangers of Speed
The central thesis of “Flash boys” is, “stock exchanges + HFT players+ brokers have rigged the markets”, i.e. intermediaries have screwed investors. However the book does not carry a single interview of a person working in a stock exchange or a HFT firm. If you have read previous books by Michael Lewis, you will see a pattern. He analyzes an event after the fact, picks up a few key people responsible for the event, and writes in great detail about the key people in an entertaining and illuminating way. This book is different and is not an after the fact narration. The underdog who is supposed to have led a revolt is still nowhere near declaring a win. IEX is yet to be granted an exchange status. HFT players are still operating and exchanges have not majorly changed any rules. It looks like Michael Lewis himself succumbed to very element that he criticizes, “speed”, and has gone ahead and published the book, before even seeing whether IEX would really make any difference to the U.S markets or not?
Chapter 1: Spread Networks and the Value of Speed
Since the advent of electronic trading a decade ago, the trading costs and spreads have vastly reduced. The increased speed and automation facilitated more precise prices, and investors have benefited tremendously. Every penny of that reduction in “economic friction” is a cash bonus. Why are market participants still looking to go faster if they can’t make their prices any more precise? The answer is: competition. If they don’t get faster and their competitors do, they lose. Speed doesn’t matter for individual investors, since there is no race to be run. Spread networks story is pretty accurate and the reason people lapped up the service is , speed matters. Lewis claims that intermediaries made between $10B to $22B per year in total profits. How does this stack up to the rest of Wall Street? At this point in his story – early 2010 – Goldman Sachs had just paid $ 16.2 billion to its employees in compensation, out of revenues of $ 45.2 billion for the previous year. In other words, Goldman alone paid more in bonuses and salaries than the total profits of all high-frequency trading firms (plus whoever else he lumped in as “financial intermediaries.”) So, perhaps, they weren’t making “more money than people had ever made on Wall Street.”
Chapter 2: The Education of Brad Katsuyama
Brad starts to suspect the market when the shares of a certain script, Solectron go down as soon as he sells shares. Brad thinks his action should not have caused a price collapse since the firm was going to be acquired and the target price was fixed. The problem with this argument is that the price is never stable even in the case of a merger. There is always a band of traders trying to do “merger arbitrage”. Using the number quoted by Lewis,i.e. $0.0029 per share as the cost of front-running by HFT firms, the author says the money Brad lost was predominantly due to his strategy and not because of front running. It is basic economics 101 playing out where he sells hundreds and thousands of shares increasing the supply that causes the price collapse. Katsuyama’s job was to minimize price impact and when he fails to do that, and ends up blaming HFT traders for front-running his order.
There is a lot of criticism on “maker-taker”model but the key idea behind it simple. When an exchange starts off, it often follows a maker-taker pricing model where a maker is given an incentive to quote. As volumes go up, the exchange may decide to invert the “maker-taker” model and start paying the taker to incentivize access to the markets, in the assumption that the previous players who were paid for providing quotes would still remain as the increased volume would make it lucrative. Sometimes this idea works and sometimes it doesn’t.
The real questions that on must address in this area are:
Fragmentation. By providing another dimension for competition, does this model encourage too many new exchanges? Or does fragmentation occur for other reasons? Would restricting price competition among exchanges impact fragmentation?
Winners and losers. Generally , guys like Katsuyama who take liquidity pay the most under the maker-taker model. If, from a policy perspective, we want to help them at the expense of the market-makers, it makes sense to ban maker-taker. On the other hand, if we don’t want policy to favor one class of traders over another, pricing should be determined by market forces.
Needless complexity. Does the maker-taker model create needless economic complexity in our markets? Or can traders easily account for this in their trading models?
Best execution. Does maker-taker pricing create unhealthy incentives for the “best price” fiduciary responsibility of brokers with respect to their client orders? The fees and rebates flow to the broker, while the price of the shares is passed along to the client.
Instead of addressing the above areas, Lewis messes it up in the book.
Another problem with Brad’s argument is that he believes in the perfect view of market, i.e. if he sees a market at a particular bid and ask, he thinks he can buy or sell as many shares he wants at NBBO. But sadly that is not how markets work. Michael Lewis does criticize flash orders, and rightly so. However the largest critic for flash orders was GETCO, one of the largest HFT players in the world.
“Someone out there was using the fact that stock market orders arrived at different times at different exchanges to front-run orders from one market to another.”
Thor, the tool which Brad and his team used to send orders was a clever tool that foiled the market makers attempt at widening their quotes before a large buy/sell order. When this happens again and again, the only way market makers adjust their quotes is widening their spreads as compared to Pre-Thor days. So, is Thor actually causing market volatility and increased spreads? Will IEX also receive the same response from market makers? Also the whole basis of extrapolating that front running was happening all over U.S markets was based on one trade in citigroup shares. It does not look like a fair extrapolation at all.
Chapter 3: Trying to connect the dots
The author dismisses a lot of things that Ronan Ryan describes, such as “moving the server in the colo by three feet” etc., as funny. The guys who moved the server by three feet in the colo center did not have a clue as to why their strategy worked/failed and wanted to come even closer to the main feed from the exchange. Lewis says Ryan is the world authority on colo. But no one has ever heard about him until “Flash boys” got published. There is a reason for the exchanges offering colo, besides the obvious outcome that it generates profits.
Exchange co-location is regulated by the SEC, and, as such, is required to be available to all market participants. Whether or not one thinks it is currently regulated perfectly, it is regulated – thereby providing, if not a guarantee, at least a possibility of fairness. In an exchange data center, the data is broadcast to all traders simultaneously, providing everyone with an equal footing and a fair chance. If exchange co-location were prohibited, traders would still vie to be next to the exchanges. They would just be housed in private data centers, outside the reach of the SEC. Such facilities could discriminate on pricing, or simply establish a monopoly. Any chance at regulation, or fairness, is gone. There would no longer be a common starting line, but instead a system where, unlike today, some firms actually do get a head start. Further, prohibition of co-location would impact the exchanges’ bottom line.
Any would-be front-runner has to overcome at least five hurdles to rip you off: 1) Determine the price and quantity of shares of your order 2) Buy the same amount of shares you want, before you do 3) Manipulate the market price upward 4) Sell the shares back to you at the higher price 5) Avoid anyone else in the market who could disrupt the scam. With the electronification of the markets last decade and the implementation of Regulation NMS in 2007, the five hurdles above have now become solid barriers to front-running.
The whole argument of “latency tables” in “Flash boys” falls apart as no order is stamped with a broker name and hence even if a HFT player has a latency table, it would be impossible to tease out whose order was being executed on a specific exchange. If one had a set of perfect latency tables for all algorithms used by all traders at every broker in the market, and all brokers had non-random, precise and unique latencies , one would still never have enough data to figure out who sent an order. Contrary to what Lewis implies, it is utterly impossible to identify anything from a single order on the BATS Exchange. One moment there is nothing. The next moment there is a trade. Nobody knows how long it took the order to hit the BATS Exchange, they only know that a single trade occurred. The next trade on the next exchange isn’t going to be terribly helpful either. One could conceivably measure the difference in time between the trades, but it’s basically useless in identifying anyone since that difference would provide a single data point. Perhaps by the third or fourth exchange one could narrow the field of possible brokers and trading algorithms a bit, but by that time the trade is complete. Remember that all these orders were already in transit anyway, so the whole exercise was pointless to begin with – by the time you saw the first trade report come back, the other orders would be long gone. Lewis completely dodges the question of how any strategy with the help of latency tables, would ever reveal the actual quantity and price of an order. Without this information, there’s no front-running possible.
The example of china ETF order being front-run is a ridiculous example cited in the book. Attributing front-running to such a market behavior is totally unreasonable.
Broker routing orders
Can a HFT firm indulge in front running based on router-driven stock quote of 100 shares on BATS? Impossible, even if a front-runner accurately guessed the quantity of an order, and they fended off all competitors, they are more likely than not to be stuck with a massive losing position as their reward.
Spread Networks – CBSX
The fact that start of Spread networks and SIRI stock volume explosion on CBSX coincided, lead Brad and team to conclude that HFT players were using spread networks to arb. The author attributes the reason to CBSX brilliant decision of using inverse maker-taker pricing for a penny stock because market making in SIRI had an extremely low risk per share. SIRI volumes returned to normalcy when other exchanges followed the same inverse maker-taker model. Which argument to believe is up to you, but the latter sounds far more convincing?
Chapter 4: There’s Another Explanation, but it’s Not As Interesting
Lewis ascribes RegNMS as the main reason for unleashing HFT. However HFT was always present before 2007.The majority of HFT firms trading doesn’t depend upon Regulation NMS, maker-taker pricing, or many of the things that Lewis describes as the foundations of “rigged” markets. Lewis argues that high-frequency trading is bad, and we can fix this if we eliminate co-location, maker -taker pricing, and Regulation NMS’s best execution requirements. But apparently none of these items are necessary for high-frequency trading. If they were eliminated, high-frequency trading would still exist. The only difference would be that big bank equity traders like Katsuyama and friends would have much lower costs (no co-location necessary, much lower trading fees) and much more discretion in obtaining “best execution” for their clients.
There are three things you need to know about the SIP: 1)You can watch the SIP, you can’t trade on the SIP 2) The SIP is faster than you but slower than other data (it ought to be faster) 3) It is sometimes used for “trade-through” protection, sometimes not. One thing to note is that it is guaranteed to be slower than the markets’ direct data feeds since it must consolidate all the market data from every source. The author says that SIP is just a cheap way to show approximately what the current market is. It’s great for CNBC or your favorite Internet finance site, and probably adequate for any retail investor who doesn’t live next to an exchange and doesn’t possess super-human reflex times. Professional investors have the choice of viewing the SIP or using direct feeds from the exchanges. Anyone who trades frequently will likely chose direct feeds. Lewis uses AAPL stock example to show how slower SIP is advantageous to HFT players. The author systematically makes a case against the AAPL example.
Why Thor needed a regulatory approval ?
As market-makers, they take the risk of always being ready to buy or sell stock. Market-makers will adjust their prices up or down based on risk, and based on supply and demand. If some trader using Thor bought up all the shares on the NYSE, NASDAQ, and BATS , a market-maker on the Boston Stock Exchange would think that (a) this new surge in demand will push the price higher , and (b) my offer to sell at the current price is a big risk since the price is about to shoot higher. While this market-maker is pondering this, Thor hits them, too. For the idea of Thor was, of course, to jump all the market-makers on all the exchanges at the same time, before even the last ones could react to the new price. The fact that some might see this tactic as rather predatory may have been why RBC’s upper management thought it might be a good idea to seek the SEC’s blessing before widely publicizing Thor. Was Thor predatory? The premise of using latency to trick market-makers into bearing the price impact isn’t illegal, but it does seem to prey on a particular weakness of a particular (essential) class of market participant.
Spreads have decreased because of automation
Lewis says spreads have decreased because of computerization, not because of HFT. Sadly, exchanges being computerized do not lead to spread attenuation. It is HFT players who play the role of market makers that have made spreads shrink.
Dark pools – play ground for HFT players
Dark pools don’t disseminate market data to any of their clients, high-frequency or otherwise. Some dark pool operators make no guarantees about their own trading in the dark pool. For those that promise that their bank’s proprietary traders have no special advantages, it’s still blind faith: unlike the public markets, there are few police on this beat. For this reason, many high-frequency traders choose not to trade in dark pools – they are afraid that the banks’ traders will rip them off in the dark.
Chapter 5: Sergey Aleynikov
There isn’t much in this chapter, except that the author says Lewis should have done some more research in the story. Aleynikov’s claimed that he only took open source code. Open source code is readily available on the Internet – that’s the idea behind it. Why didn’t he just plan on re-downloading the code at his new job? It’s difficult to buy the idea that it would be easier for Aleynikov to transfer the files to a third-party server, later retrieve them, and then “disentangle” all the Goldman proprietary code.
Chapter 6: How To Take Billions From Wall Street
The author wishes that IEX succeeds so that people might think that HFT problem is solved in the markets, a problem that did not exist in the first place.
The pricing used by IEX – $ 0.0009 per share for takers – is far more attractive to a taker than the pricing on any of the major exchanges. Applying Lewis’ logic, this is just another kickback to the big banks (all of whom are on IEX), and IEX is one big “flash trap.” If it seems logically inconsistent, that’s because it is. Realistically, different pricing models are just that: different pricing models. They are set by the exchange to attract business. To claim that high-frequency puppet-masters dictate these pricing structures to the exchanges doesn’t make sense in the case of IEX, NASDAQ, NYSE, or anyone else.
Types of Predatory behavior
The puzzle masters in “Flash boys” categorize the predatory behavior in to three types. 1) front running, 2) rebate arb, 3) slow market arb. Front running is already shown to be impossible in the earlier chapters. The fact that HFT players can get a rebate without doing a trade is impossible. Also slow market arb is again impossible in the examples cited in the book because of trade-through protection provided by RegNMS.
It seems that (1) the tweaks implemented by IEX would not actually prevent the predatory trading that Lewis hypothesizes, but (2) there doesn’t appear to be any predatory trading on IEX. This paradox could be explained two ways. Perhaps the predators are simply scared to show their stripes on IEX. Or perhaps the feared predatory trading is rare to non-existent, and it doesn’t matter that IEX’s defenses are illusory.
Dark pools and broker internalization facilities aren’t unquestionably bad, but it’s hard to make a compelling case for any significant benefit . For professionals in particular, they make it easier to shoot oneself in the foot. For the public, the lack of transparency doesn’t inspire confidence. And for the markets themselves, there is a legitimate question about whether or not they detract from the price discovery process.
Chapter 7: IEX Launches
One of the often cited reasons for taking a stance against HFT players is that they siphon off trades coming from pension funds and such large institutional clients. It might be worth telling all those critics that the same pension funds are also invested in high-frequency trading firms. For example, more than a third of high-frequency trading behemoth KCG / Getco is owned by institutions and mutual funds. CalPERS and CalSTRS, two of the largest pension funds in the country, own stakes in privately held high-frequency firms as part of their private equity portfolio.
This chapter shows so many chinks in Lewis arguments that I did not feel like summarizing as it would have meant basically replicating the whole chapter from the book. After reading this chapter, I was kind of overwhelmed by how much stuff from “Flash boys” was plain wrong.
Chapter 8 : Dinner with Sergy
About Sergy, the author says that a little more effort should have been taken to understand the truth. He says,
I don’t know what was contained in the 500,000 lines of source code that Aleynikov took. I do know that most trading systems use proprietary code for strategies, but rely on open source operating systems and network processing code. This open source software is often modified for the particular network requirements of trading. And these modifications are quite valuable – they make every single order faster or slower , depending on how clever one is. I wish somebody had asked hard questions.
All the books written by Michael Lewis till date other than “Flash Boys” are after the fact narration. In his previous works, one can see a pattern. He identifies few key characters and weaves an entertaining story along with facts so that readers can easily understand the basic theme. It looks like with “Flash Boys”, he has created a catchy narrative but sadly the foundation of the narrative is weak. Peter Kovac has written a 100,000 word book that basically rips apart almost every argument of “Flash boys” and thus defends HFT’s role in today’s markets. Must read for someone who wants to critically analyze the arguments made in “Flash boys”, that says U.S markets are rigged.