The Gateway to Algorithmic and Automated Trading

Critical thinking

Published in Automated Trader Magazine Issue 35 Winter 2015

Sal Arnuk started electronic trading some 20 years ago at Instinet, where he met Joe Saluzzi. The pair have since started their own brokerage, Themis Trading, and become big voices in the market structure debate - attracting praise and criticism alike for taking a stance against increasing complexity in the equities markets.

Sal Arnuk, Themis Trading

"All of a sudden, a goldman Sachs trader gets arrested at Newark airport and everybody wants to know what the hell high frequency trading is."

Automated Trader stopped by Themis Trading's offices, a slice of Wall Street located just off an idyllic New Jersey Main Street, to talk to Arnuk about his biggest beefs and how he navigates in the darkest pools.

Automated Trader: You have become known for being a vocal critic of US market structure. How did that start?

Sal Arnuk: The market was changing and adapting, and there were people figuring it out while we were focusing on what a company's earnings were going to be. But people were learning to master the microstructure, and we were caught behind the times.

So we very quickly educated ourselves and started improving the way we interacted with the market.

When we dug into this, we found things that were hurting our execution - order types, hidden order leakage. The first things we found were pretty unsophisticated things we were doing wrong.

We were vocal about those things we found with our clients, and even on an STA (Security Traders Association) conference call. We also wrote a whitepaper or two, and some people read (them), and a lot of people didn't.

All of a sudden, a Goldman Sachs trader gets arrested at Newark airport and everybody wants to know what the hell high frequency trading is. So our whitepaper goes viral, we started our blog, we wrote our book "Broken Markets". So we have been trying to broaden the debate.

AT: What do you say to your own critics for taking that stance?

SA: Ironically, when people talk about us as "old timers" or saying that we can't compete, in our entire career we were never market makers, we were never specialists on the floor defending the specialist system, defending the market maker spreads. Of course not. Our whole business, our whole careers at Instinet, is built on selling against those wide spreads, taking out the middle man and bringing buyers and sellers together directly, sourcing liquidity in the cleanest way possible and leveraging technology to do it.

We feel that somewhere technology took a wrong turn and inserted the maximum number of intermediaries between natural buyers and sellers, instead of bringing them together directly, and it led to all kinds of distortions - especially with payment for order flow - which is why the markets have gotten complex.

AT: How does Themis Trading fit in as a brokerage in this environment?

SA: We have 40 or 50 institutional accounts, which haven't really changed over the course of 15 years. We are proud of that. Until 2006, no one had ever heard of us and we loved that, because we have a very loyal, small group of clients. We will never be goldman Sachs, we will never be a large brokerage firm. A lot of that is by choice, and a lot of it is because we believe our business model is not scalable. Everything about Wall Street is: find a niche and scale it. We fill a niche but we recognise that it is not scalable because to replicate what I do, and what (our team) does, to try to create an algo that thinks like we do defeats the purpose. Sooner or later that algo becomes predictable and predictability is the enemy.

Everything about how we execute here is trying to execute in the cleanest way possible, take out the intermediaries and we source liquidity wherever it resides. I don't care if it is the New York Stock Exchange, IEx or a market place in Kabul. Wherever the liquidity is, we want to find out without going through 19 hops to get there, because along those 19 hops bread crumbs get dropped. There is slippage. There is information leakage. And as a result market impact suffers.

To get an order to buy (thousands of) shares, you can't just come in and post that on an exchange, and at the same time, if I were just to throw it in the average broker-dealer smart order router, that order is going to get shown in millisecond time periods to an ELP (electronic liquidity provider). It's going to get winged around through a pinwheel of other dark pools. So, if I just go in to a dark router and think that my order is dark, it is really naïve.

AT: So what is your approach to execution?

SA: We have direct pipes to exchanges. We also have third party pipes to exchanges, so we use Knight's pipes, or Credit Suisse's suite of tools for example. We probably have 10 different third party broker dealer suites as well as our own direct connections, including IEx, BIDS, Level ATS.

The way we execute changes from order to order, but always in the back of our minds, we need to be unpredictable. I would love nothing more for my own pocketbook to create a good housekeeping seal-of-approval algo and then shop it out. But that is just not something that I think would be effective. As soon as you make it systematic, it is going to get figured out by some of the very savvy people in the market place.

AT: One of the biggest areas of concern for buy side right now is how orders are being handled by exchanges in the US. What does that look like from your vantage point?

SA: The buy side is very frustrated and in general there has been a tendency to not want to speak out. For one, they don't want to be in the regulatory spotlight and create enemies. And if you are on the buy side and you come out and you say all of these things - that clients have been hurt - then they have a fiduciary (obligation).

The easiest thing for a buy side trader today would be to just take an order and put it in a scheduling algo - like VWAP or percentage of volume algo. They are pretty standard, and you will never get a great execution.

Let's say someone gives me (hundreds of thousands) of shares to buy in a (retailer). As soon as that goes into a VWAP algo, it is broken up into time buckets over the day, maybe it's six, maybe it's 60 time buckets. In each of those buckets, let's say it's a five-minute time window, the stock can trade. So even though the spread now is a penny wide, the stock can trade (with a wider spread) in that five-minute window. In every one of those windows the algo first tries to bid on different exchanges and all the while gets shown to ELPs in the dark on the way to the exchange. So it leaks, and then, lo and behold, in every five-minute window it is buying at four and five cents, when it would have been better off crossing the spread right out of the gate at cent one or two.

The broker is trying to minimise its costs, or there are payment for order flow distortions that exist in the market place, because it is better to receive a rebate than pay a take fee, which is expensive for a broker-dealer.

So in the end you can have a (multi-cent) slippage, not a tenth of a penny, or half a penny, it is three cents (for example) on every child order. When algos behave this way, and are so systematic, that is a real danger to execution.

AT: Won't transaction cost analysis pick that up?

SA: The trader will get back a report and it is an average execution and it is roughly where it traded throughout the day, but by the systematic way that (the algo) executes, it's signaled along the way and affected the benchmark.

So it looks OK because it is close to the benchmark, but the benchmark has been distorted, the benchmark has been corrupted. Maybe they overpaid at the end of the day eight cents higher or 16 cents higher than they would have if they were trying to VWAP in a clean manner. So there are these types of concepts and ways of executing and costs that are not picked up by modern TCA (transaction cost analysis).

If regulators take out the incentives, then a lot of this bad stuff goes away without them having to pass complicated rules.

going back to the buy side, the reason they are requesting that a broker-dealer give all the data, or want to see where each child order is routed, where it gets executed - all the things that they measure and are taking ownership of, kudos. It is hard and it
is a shame because if you think how much money the buy side has to invest in surveillance tools and police tools to monitor their trades to make sure they are not getting ripped off. Those resources could be used to seek out companies to invest in.

So their cost equation is changing. They are having to spend so much more resource on being policemen instead of seeking out good investments.

It is a shame that the regulators, the cops on the beat, are forcing this downstream instead of taking care of it for everybody in terms of creating best practices and simple rules that make common sense.

Why do we need to have three years of data and study to see whether it is a conflict of interest paying someone to execute, getting paid to give a customer's order to someone else so they can trade against it?

When we execute, we are tracking real-time where our orders get routed and where the ultimate execution comes back. We are able to discern patterns on the fly as opposed to looking at the last three months of data and making a generalisation.

AT: Regulatory scrutiny of these issues has been increasing - are you optimistic on what can be achieved?

SA: If you look at how our regulatory authorities in the US have been addressing issues, it has been very reactionary and band aids - a circuit breaker throw in, a limit up limit down - which worked pretty good by the way. But they are all band aids and they will all work until they don't during the next black swan and no one can predict - including us - when something bad will happen next.

And these things happen because it is so complex and there is so much fragmentation and we have taken the easiest thing in the world which is matching supply and demand and made it into a complex Rubik's cube.

The market can truly decide if we take away these obvious distortions - paying someone, buying someone's order is a conflict of interest. If you eliminate payment for order flow verywhere the market of true competition and true capitalism will sort out the correct number of fragmentation and innovation that the market demands or needs for matching supply and demand, which is not rocket science.

We loved it when (SEC chair) Mary Schapiro said 'we are going to look into order types' or (ICE exchange CEO Jeffrey) Sprecher. But it's one thing to hear in a speech that there is bad behavior and then it's another thing to actually watch what they do and what they file for, and what the regulators approve.

Even if you know all the order types, that doesn't quite tell the true story because there are modifiers. So, New York Stock Exchange (now owned by ICE) applied for an add liquidity only and post only ISO (inter market sweep order).

Haim Bodek (whistle-blower on order types) and Bloomberg Tradebook commented to the SEC saying don't do it and it went through anyway. This add liquidity only, which on the other exchanges is typically used only for midpoint pegs, now it can be added to any order type because the precedent was just set with NYSE. So what is going to happen? I have a feeling you are going to read a lot more order types over the next few months because all the other exchanges will file now that NYSE got the green light.

So what is my view on market structure? It is the same, it is getting more complicated, it will not get simpler. The only good thing investors have is they are relying on themselves more and looking into it and that is a great deterrent to bad behaviour in and of itself.

AT: Do you see Themis Trading branching out past equities into other asset classes?

SA: Currently no, we have our niche. We feel like we found our role in terms of educating the market place on our little corner. That is not to say we won't branch out into derivatives in the future.

It depends on what the market demands, but one thing I know we won't do is we are not going to be an also-ran and we will not follow policies and mechanisms that we think are flawed.