The Derivative - Decoding Execution Algos with Joe Signorelli & David Don

Episode Date: February 4, 2020

This episode dives deep into the math and people behind what most people take for granted or don’t even know they need to improve their trading – execution algorithms. We cover Orlando as a favorite vacation spot (What&^%),  what/who is a quant, working with students at Notre Dame, UCLA, and Illinois, how even a slow trader can benefit from fast execution, algos named Prowlers, Snipers, Leggers, and….R2-D2 as a vacuum cleaner. RCM-X is a trading technology and risk management services company, and as you might guess from  the name, a subsidiary of RCM Alternatives. We sit down and chat with their top two people - Joe Signorelli, RCM-X Managing partner, and David Don, COO/Head of Algorithmic Trading – and dive into the background of RCM-X, what execution algorithms really are, how they're implemented in modern day trading firms, and why they're important for more firms than you might think. We also discuss the RCM-X University program, which is a cool way this innovative group is bridging the gap between academia and professional trading firms to help build a solid base of future talent of the industry through top college programs. How to follow along with RCM-X: RCM-X website, Twitter, & LinkedIn Joe Signorelli LinkedIn & Contact David Don LinkedIn & Contact And last but not least, don't forget to subscribe to The Derivative, and follow us on Facebook, Twitter, or LinkedIn, and sign-up for our blog digest. Disclaimer: Due to industry regulations hosts and participants will not discuss any company funds on this podcast. All opinions expressed by podcast participants are solely their own opinions and do not reflect the opinions of RCM Alternatives or any of their affiliates. This podcast is for informational purposes only and should not be relied upon for investment decisions. Visitors should not act upon the content or information found here without first seeking appropriate advice from an accountant, financial planner, lawyer or other professional. For more information, visit rcmalternatives.com.

Thanks for listening to The Derivative. This podcast is provided for informational purposes only and should not be relied upon as legal, business, investment, or tax advice. All opinions expressed by podcast participants are solely their own opinions and do not necessarily reflect the opinions of RCM Alternatives, their affiliates, or companies featured. Due to industry regulations, participants on this podcast are instructed not to make specific trade recommendations nor reference past or potential profits, and listeners are reminded that managed futures, commodity trading, and other alternative investments are complex and carry a risk

of substantial losses. As such, they are not suitable for all investors. Welcome to The Derivative by RCM Alternatives, where we dive into what makes alternative investments go, analyze the strategies of unique hedge fund managers, and chat with interesting guests from across the investment world. They have the connectivity to all the traders that connect into the exchanges, which is a very detailed and a lot of technology that makes that happen for thousands and thousands of traders. But with our execution algos and us doing fixed connectivity into these ISVs, we can provide the algos, execution algos, on demand.

Welcome to The Derivative by RCM Alternatives. I'm your host, Jeff Malek, and excited to have two special guests today, Joe Signorelli and David Don of RCMX, two longtime business partners who have built software companies, managed a hedge fund, and now work on a new type of business model, automated trade execution as a service. Hey guys, thanks for joining us today. We're often here to talk to those who come up with the trading ideas, the alpha generation, but you're involved from a little bit different perspective of helping hedge

funds, banks, and the lot execute those orders in all sorts of cool ways. So I'm excited to dive into how that all works. But first, Joe, I'll start with you because I know these stories overlap a little bit tell me how'd you get into this business well first thanks Jeff for having us here the business we're currently running right now which is as you alluded to execution algo business we got into it it was it was more it was a natural role for us to come into this service business from our backgrounds of trading, statistical arbitrage, market neutral trading, market making in the past, even running a couple different hedge funds in the past. We, along the way,

we always built our own technology to service our own trading and it was in about 2008 when David and I decided that we decided to start a service company or a technology company because we thought we had enough expertise and we had a competitive idea and edge that we can build out a service on software which eventually led out to the execution algo business and you're the trading background you were back on the CBOE yeah I back on the CBOE? Yeah, I started on the CBOE in 1986. I was in the OEX pit in the crash of 87, and I migrated to the upstairs trading back in the early 90s, late 80s. I was more or less getting more comfortable trading on the old Schwartz-O-Tron system,

and eventually we tested Dennis May's system, now known as MicroHedge, back in 1991. Real quick, what was the worst story you saw in the crash? You were in the OEX pit during the— I was a clerk in the OEX pit. I saw—I guess just from—there's always some stories of groups that blew out, but I think what was most— I always remember, I looked over the top of the, where we sat was elevated about a second floor, two-story overview, because the way the pits would come from bottom all the way up to the top, we were on the

back, and so we could look down and look down about maybe 20 feet, but when you look to the open area before the S&P pit was built, there was a big open area. And I just remember looking over during the crash of 87 of traders just wandering around and looking like, you know, you could see the distress. I didn't realize until a day later, like what the devastation was, because I wasn't carrying positions. Our firm was carrying positions, but we were in good, good shape. But I didn't realize how many people were actually blowing out in their lives and just changed dramatically. And I always remember that sight of several traders walking around with their heads down, and I'll never forget that. And David, throw it over to you. So when

did you first meet Joe and you guys started thinking you could work together? So when I met Joe, it was 2003, and Joe was managing a stout arb at a subsidiary of TD Bank in Chicago. And I joined as an aspiring portfolio manager. And right away when I joined the team, Joe had me working on some pretty exciting projects with capital structure arbitrage modeling, convertible bond modeling, and using insights from fixed income markets and credit derivative markets to drive relative value signals and market neutral equity portfolios. And I stuck around. It was originally a rotational program that I joined with and joined Joe's team permanently. And I've never really looked back. I've been working with him for 16 years now.

Somewhere in there, you guys became partners, right, instead of employer-employee. How did that go down? Yeah, so kind of our model, and we carry that same model right now. We really have a flat model as it is. So it's easy to kind of convert that into a partnership and an actual LLC type of structure later on. But when we work as a team, as a trade team, now as a service team, we have a culture where everyone's voice is heard. And so early on in David's career, I think David and Larry Lai were partners on building out the capital structure arbitrage modeling, and they got a seat at the table ASAP. I mean, they were doing great work. And so when we sit down as a team, it didn't matter

if someone was there for two years or three years, if they're adding value, we had that type of partnership. But when we actually formed the LLC, it was 2008. And what was interesting was we ran a successful book at TD Bank, and we didn't really want to move up to Toronto to continue running our book. So I decided I think the best thing is a service technology business. David and I were following up on some leads over the years that had come to us, SAC and some other groups that said, hey, you guys run a great book. Why don't you come over by us? Well, 2008, during the middle of the crisis, a lot of groups were not handing capital

out anymore. So we were first trying to find groups that wanted to run our portfolios. In the meantime, I was trying to convince David to join and let's do this technology play for low latency trading. And it was until about three months of realizing we were not going to raise capital for our strategy. I remember we left JP Morgan's office and David turned to me and said, okay, what's your idea on the technology? And I said, okay, you're the COO. I'm going to run it. I'll put the capital in. And we got a space already over at Benji Schwartz's office. And that day we kicked it off. So that's pretty much how I knew David was the right partner. And the reason why was David covered technology trading. He has a legal background through his

family. He's not a lawyer, but he understands the structure of companies. And I knew David was the right partner to start this type of program. And you guys made that very tax advantageous decision to skip out on no income tax in Toronto and instead the pension issues here in Chicago. Chicago's home. But one of the other things that happened along the way as we were running strategies under the bank is our core portfolio was market neutral equity, which has kind of weekly to monthly forecasting horizons, portfolio rotations on a daily to weekly sort of scale. and built DMA market access, wrote our own basket execution algorithms, and started also running smaller, higher-frequency strategies and market-making and single-stock futures and all sorts of other stuff. As we got more exposure to the technology side and as high-frequency trading was kind of apexing in the 2008 volatility, we started looking more aggressively at high-frequency trading strategies

and realized we would definitely benefit from a more cohesive technology stack that could bring together that backtesting through production research and development pipeline on a consistent code base and be fast enough to do large back tests off of thousands of securities with full depth of book data. And the kind of need for that sort of technology drove the idea for the company we formed and how we thought about the product we wanted to create. And so you're, for lack of a better term, you're a quant yourself, right? More or less, yeah. Yeah, and so unpack that for me. Like, you get a quant degree. How do you become a quant, and what does that entail in your day-to-day? So for me, I only have an undergraduate degree. I was a math and physics major, sorry, math and economics major

with a physics minor. Only, only. I also had software development jobs during college on the side and background in C++. When I joined the bank, it wasn't to be a programmer. It was more to be a trader. And quant trading, as it got more and more high frequency, it required me to go back and brush up on more of my coding skills and get back into the programming world. So for me, the quant stuff was more a practical matter. When we joined the bank, there was an extensive training program. We learned a lot about derivatives modeling from some talented PhDs in the industry. And for me, I've always just like kind of picking up a book and self-learning

and figuring out what I need to do and what I need to learn and what I need to implement to make something work. So it just kind of was a matter of necessity. How do we try to find some edge in the market, implement it and deploy it? A little off topic here, but do you feel right? The banks have largely gotten rid of their trading operations and their desks. Do you feel like that path to where you've become is no longer? Like if you fast forwarded your life and you were just coming to the industry now, would you get those same opportunities, do you feel? Well, that program was rotational. So I got exposure to risk management, IT,

floor trading on the SIBO and quant equity. And I think IT, floor trading on the SIBO, and quant equity. And I think aside from floor trading on the SIBO, all those things are still pretty relevant. The nature of a trading desk at a bank has been evolving over time, but it's more algorithmic, more systematic, more based on math and programming. But there still are sales traders. It's just a different mix of what people are doing. You guys also got involved with the university program. So tell me a little bit about how that came to be and how that's working out. Yeah, so it's a big part of our growth of our team, and our approach to the universities has been practical and also kind of a give back. Back in the mid-'90s at Stafford, they basically used the O'Connor model of hiring, which was a very quantitative approach and hire the best people and have them go through, like David said, a rotation program.

So I inherited that program when I took over the stat arb team at Stafford Bank, at Stafford Trading. And it was basically a great way to hire. But when we moved into TD Bank, we modified that because our focus was really Waterloo University. And we did more co-chaired programs in high frequency modeling and blue gene technology with IBM for our market making and things like that. So it was more centric around University of Waterloo and when we moved out onto our own we did more local projects on the university to help us out with our programming of technology so why IT or DePaul or even University of Chicago for local hires we moved into Wedbush we expanded the program to be more of a practicum focused program to where we do projects at UCLA's Masters of Financial Engineering Illinois's Masters of Financial Engineering Notre Dame's International Business Club.

We've done stuff at University of Cincinnati and Northwestern and Carthage College, many different universities. I sit on the board of directors for the Masters for Anderson School of UCLA and also Illinois' Masters of Financial Engineering. So the main thing we do there with those projects are we're using our technology, we're using our skills and our team to help lead projects every semester, but we also have installed our software. So we're letting the universities use our software to help their students understand the microstructure of the market,

and we're getting students that are vetted and certified on our software. So when they go to firms to go work, they're certified with our technology. But the main reason we do it is really to give back. And it keeps us fresh, too, when you're dealing. And we have great hires. I think every one of our hires over the years have come through these university programs. David, I can't think of anyone that hasn't come through the program over the years that we have on.

So it's been very... And so the students are getting access to some next-level technology, software, and real-time, real-market access. Yeah. The groups involved that the students are helping do testing are getting really smart kids that are working on projects for them. So it's kind of a win-win. And you've brought in other groups now, right?

Other trading groups that can access these students? Yeah, so RCM right UCLA, from Illinois, from as far away as Basel University, Northwestern. And along with those professors, we also have some kind of a call to arms to the exchanges and the data providers to provide more for the universities, not just be there to say, hey, you know, this is great. We'll give you some sample data. I think in order to produce the next level of quant finance engineers to keep up with the growth in the business, the professors and the universities have to have better data and better systems and so rather than talking about it and asking others to do it, we've dedicated

and volunteered and donated our software to many of the universities currently. Did the exchanges listen? Yeah, actually they they did. Actually, it's interesting. Not only they listen, in all fairness to the exchanges, a lot of these universities actually have exchange university licenses for raw data and data, but they don't know how to use them. They didn't really have a system to incorporate it. So yeah, and I think with our software now, some of these groups are able to use some of these APIs and some of the data that's available. But yeah, in all fairness, the universe...

Essentially, the exchanges said, we already give them the raw data, and you're saying, hey, the raw data is really of no use to them. Let them use it through one of these platforms where they can actually do real things with it. Yeah, for the most part. And they're usually very willing because they do most of these, all these exchanges have research groups that are, and they're very good at donating and giving back. And they just need to know where they need to go. And I think we're helping, along with many other companies, we're a small company, there's bigger companies doing this on a much larger scale. That's great. Congrats on that. Thanks. scale. That's great. Congrats on that. Okay, switching gears a little bit and get a little

bit more into RCMX. So you're on an elevator at a hedge fund conference. You've got seven floors to tell the billion-dollar hedge fund manager across from you what you do at RCMX, why it matters to him. What's your elevator pitch? Mine? David, maybe you want to go first? Sure. So RCMX provides two core technology services. One is execution algorithms that help traders who have large orders or even moderately sized orders to try to execute those orders more efficiently, more systematically, get a better grasp on optimizing their fill quality, understanding their true transaction costs, whether they're managing a portfolio or executing on a desk on behalf of other clients.

The other service we have is our core software, which we use to build the execution algorithms. We license that software to systematic traders, black box traders, to help speed their process of automating new trading strategies. I'd probably go maybe instead of 12 floors, I probably can do it in two. black box traders to help speed their process of automating new trading strategies. I'd probably go maybe instead of 12 floors, I probably can do it in two. Let us be your quant extension. Let us do the lifting for your execution. We have a full on-demand process that you can access to improve your execution quality.

And then we have software if you want to get alpha quality for building out alpha algos. So let's talk a little bit more about that. Most of the people, as I said in the intro, most of the groups we have on this pod are talking about how they create alpha, how they're generating signals. So you guys are kind of the next step to that, right? Once the signal is generated, actually getting that order into the market. And you believe there's alpha in getting that order into the market a little more efficiently. I wouldn't call it alpha because I would say we try to reduce the slippage. Can we save them money versus what they're doing now is really the goal. If we can improve with our technology, they've made a decision what they

want to trade. And instead of just putting it out to the market or trying to use the manual process of putting the orders in the market, I think we have a proven technology using statistics data and so on and so forth. But to put it out to the market with a system. We want to reduce their slippage. We want to save them money over a series of trades. And with a lot of data, you could see the results of that in a transaction cost analysis. Going back into what motivated us to start developing expertise in this area, when we were portfolio managers ourselves, we would look at our turnover and our slippage. And if we're slipping, like, say, five basis points on average on the securities we were trading, and if our turnover was like 25% a week, it would add up to a few percent drag a

year on our portfolio returns. If you have a CTA or hedge fund-like product, that can be a big cut of your expected annual returns. If you're trying to have, like, let's say 10% volatility, 10% to 20% return profile, slipping a few percent a year definitely eats into your returns and so finding a new source of alpha versus improving your execution quality those can be kind of pari passu in terms of overall improvement to an investment strategy and so it depends on who you are and how you look at it but like whether you look at it at turnover and slippage in terms of basis points and returns, or if you just look at how many contracts of a future you trade per year, and if you can shave, say, a quarter or an eighth of a spread even on your average execution,

some couple dollars a contract, that adds up if you're doing a lot of contracts. And so well known that the futures markets especially are a zero-sum game. So if you're saving that client those eighths, who's giving that up? Who are you competing against to get that? I mean, I think it's fair to say that as markets have become faster and more electronic, volumes have picked up. The volume quoted at the spread might have decreased in some cases, but the spread on average across different product categories tends to be tighter. And so overall, everybody's getting tighter spreads, but it's like kind of an increasing game of more volumes for market makers at tighter costs to liquidity takers. And as traders adopt more efficient technology, it's kind of a game of leapfrog.

The people that they're trading against are the liquidity providers. They're giving up less to those counterparts. Yeah, and I kind of, the way I look at it is the way I think a lot of the big energy firms, for example, when they would put in a hedge against their physical or a hedge against something or an offensive play if they're getting ahead of the trade, they would place these orders into the broker, which would be matched up in the pit, usually against probably another energy firm where they had inventory or they were trying to build inventory. And then there might be a local trader that might take part of the trade. But usually they'd find a mid-price to trade.

And I think it's changed quite a bit where now you have a lot of quantitative groups that are very fast, more nimble, they might not understand and care to understand moving oil across the world or moving corn across the world or hedging against weather, maybe so in that. But what they've done now is they've come in as statisticians, math guys with very fast systems. So what happens now is the trade is now this big energy firm trading against a market maker. And so what we're doing is we're providing technology that's just as fast as those market makers or those quant hedge funds. And we're providing anti-gaming techniques and that to try to level the playing field for our clients. The market maker would be fair to say is typically some high-frequency trading firm that's moving into the futures markets? Yeah. I mean, I think there's some, you know, I think a lot of those groups, a lot of people would know these names. They're capitalized well, and they've got some great quantitative

approaches and very good technology. We're talking like citadels of the world. I'd say the citadels, the jumps, and akunas, groups like that. And the other side of that is they're providing, like David said, they've narrowed the spread down. So there's value to having market makers out there. And I think what we're trying to do is give our clients who are doing trades, who are experts at their business and trying to give them improved fills by trying to match up technologies with some of these quant firms. Now, say I'm your hypothetical big energy firm there, and I have whatever December 2024 oil risk, and I'm hedging something, right? Seems to me it doesn't matter as much if I have a super long timeframe. What's your argument to that?

So I think it's, you know, back to that characterization of thinking of things in terms of turnover, contracts per year. If you don't trade a lot per year, obviously, you can't save as much from an execution improvement because we can't save you money if you're not trading. But I think even on a long-term order, what people underappreciate is the importance of speed. If you are holding a position for a week and maybe you're working portfolio rotation over the course of a day, you might think that it doesn't matter if you just carve that order up naively, sign it over a screen with some slow technology.

But any automated execution system, at the core of what it's doing is it's taking in a data feed, looking at live prices, trying to figure out how much to send out and at what times, whether it should cross the spread at the moment, whether it should post an order passively. And as you're adjusting these orders and where you are in the marketplace, if you're slow to process market data, you're systematically adverse selected because you're slower to get into a passive queue and more traders ahead of you getting fills. It increases your odds that when you do get filled, you're adversely selected or that you don't get the fill in the first place. If you're trying to take liquidity and you miss your price, you have to chase the market to a worse price if you still

want to get filled by the end time of your horizon. If you're posting a query and you see the market turning against you and you want to adjust your limit to a more passive price, if you're slow, you might get filled before you're able to do that. So all of these things create a systematic disadvantage to being slow, even if it is a long order working over a good amount of time. Right. I think for sure people get confused with like, I'm only trading once a day or something. How can that help me? And it seems like everything you just said is a lot to work through. But the whole point is it's the computers doing all that work, right? Like, can you give me an example?

Say I'm that big energy firm I'm trading. I need to get a thousand contracts of oil to buy. From that simplistic example, how are the algorithms treating that order? Make a few assumptions on what they want. Yeah, so there's a few common categories of algos. One is an algo that's scheduled in the sense that it has a time horizon over which it works, tries to carve up the execution over that time horizon to achieve an objective like minimize slippage or minimize risk-adjusted slippage. And so those algorithms might try to blend in as much as they can to normal volume patterns to try to push the market less. They might trade a

little bit more aggressively up front to minimize price risk or to trade off price risk versus that market impact of trading quickly. Another category of algo tries to trade in line with the current market activity. So if you see, say, a thousand contracts print, you're allowed to trade, say, 30 contracts if you're trying to track at a 3% rate. And then another that more just kind of takes some of the burden off of a trader who would traditionally be working a limit by hand and giving a more enriched set of capabilities versus a traditional, say, iceberg-style algo. And those might not be guaranteed to fill to completion, but it's an alternative to a laborious manual process of adjusting limit orders throughout the day. And so fair to say the whole concept is don't do all 1,000 at once. Yes. Split it up depending on if you're more preferential to the now, to the later, to the average. Split it up in some way that you get. If you have to execute 1,000

contracts and you want to do it immediately without risk of not getting a fill on the market running away, your only option is to send an aggressive order, eat through the book. If there's maybe 500 at the top of the book, you're going to get filled. Crossing the spread, you already have lost half of the spread in slippage. You might have to walk another couple price levels to fill the rest of it. And of course, it could be an even less liquid product, multiple price levels. So the alternative is spreading it out you have more price risk but as you're taking liquidity from each price level you're giving more time for liquidity to replenish or alternatively you're trying to be passive and post and try to avoid crossing the spread in the first place which in a lot of futures products is important because many futures products

have relatively wide bid-ask spreads for the volatility of the product, which encourages a lot of queuing dynamics, which everyone wants to get a pass to fill, but it's harder to do so. This game's been around in the equity world for 20 years, 10 years? How long? Yeah, when we were running equity portfolio strategies around 2000, execution algorithms were already gaining a lot of popularity. At the time, we were using third party execution algorithms and the equities world decided to do it ourselves eventually to save on

both brokerage and slippage and get better control over the process. Those were different markets. Speed wasn't quite as competitive. And in equity markets, particularly, you don't have REG and MS yet. And the order routing was simpler, order types were simpler. The technology spend was less of a burden. Looking back, if we were to try to do that ourselves today, I think we would have been crazy to try to be both portfolio managers and execution specialists. And so why did it take futures longer to catch on? There's less granularity, contract, notional contract sizes are bigger. What was it? I think futures is a different sort of marketplace. Every product trades on one exchange. So the routing equation

isn't as big a part of the problem where you're co-locating for different exchanges being in different data centers isn't a part of the problem. And also a lot of futures traders seem to specialize in a particular niche, like you might be an energies trader or an ag trader. So that kind of cross asset class correlation matrix-based data that is typical in an equities trading environment is less present in the futures markets. And those sort of like intercorrelated marketplaces drive development of quantitative trading techniques and automated trading. If I got to buy a million Apple and sell a million Microsoft, I have a much bigger need for a... Spread it out over different products, over different markets, different geographies. Yeah, I think it also goes into that options were, especially with the SIBO and equity options were

more relevant to the box, reverse conversions kind of forced people to be a little more, I'm not going to say quantitative, but a little more systemized to look and say, okay, if the price of the option is here, what would be the box market? Or if the price market market's here on this option at any time with the underline, what's the reverse conversion market? So it forced people to kind of computerize things a little bit more and quickly. And like David said, the microstructure market being that there was competitiveness to where you can get that same product on the NYSE or Amex or something like that, that it created an arbitrage scenario. So it forced people or pushed people to be more systematic where the CME has its products, ICE has its products. Yeah, you have oil on CME and oil on ICE. So there's an arbitrage.

I just think it just made more sense, and there was a lot of liquidity. There was a lot of players, naturally more retail players fed into the system. So you didn't hear people now, you hear a little bit of retail talking about trading in the ETF golds or maybe getting into futures a little bit. I think the small exchange is coming up soon. That's supposed to be retail-focused. But most retail, even today, in their 401, they're trading U.S. equities, right, or global equities. They don't have exposure. And so, therefore, there was a pool for, I think, arbitrage people and systems to go after, hey, if we can partner up with Meritrade or Schwab or somebody and get their flow for the retail, we can then, you know, if we can systematically take that trade, we can make more money with this and provide liquidity for the retail groups.

And on the flip side of it, the futures markets arguably may have led the pack on the ultra-low latency trading in terms of trading firms taking news feeds and trying to transmit them wirelessly to a very liquid product like S&P futures or futures trading firms that were arbitraging in this index cash behavior in New York versus index futures in Chicago so there are aspects of that trading world where futures actually led from a technology speed basis. And for execution algorithms, while there's not the routing problem, there is still a lot of quant work to do that you don't have in the equities market to the same extent where different products trade with different matching engines, different products have exchange-determined rules in terms of tick sizes and

contract multipliers. And so the size of the product you're trading, the tick size, the matching engine, product-specific seasonalities are more pronounced and how products behave across the expiry cycle, how calendar spread volumes correlates to outright volumes. There's a lot more product-specific nuance and modeling to do to achieve optimal execution in the future space. I like it. Joe, we talked a little bit off mic about you're kind of viewing this as execution algos are kind of tying the whole value chain together between the trading firms, the FCMs, the ISVs. Walk me through that. Tell me what you're thinking about that. Sure.

And, in fact, I think putting out a blog today regarding enhancing the chain across for with using execution algos and i think dealing with channel selling with the fcms and with the isvs and working with the clients directly i think we execution algos are enhancing and adding value along the chain for example let's start with the exchanges the exchanges, their largest clients that are doing big volume and paying these exchanges, they need these execution algos because they're slipping and they need better fill quality. Well, exchanges don't provide execution algos for these clients. So we provide them and other third-party groups or banks provide the execution algos in general. So I think there's value there for the end client. I think the ISVs, which is the TTs, the FIDESAs, the CQGs, the Bloombergs.

What's ISVs? Independent software vendor. And I always get tripped up on that. I use it all the time, but it's the independent software vendor. And so the independent software vendors, I kind of look at- They have the connectivity to all the traders that connect into the exchanges, which is very detailed and a lot of technology that makes that happen for thousands and thousands of traders. But with our execution algos and us doing fixed connectivity into these ISVs, we can provide the algos, execution algos, on demand for these clients. So the ISVs have very little lift.

The clients have very little lift, but we can provide these execution algos through the ISV. So in other words, their clients, the ISV's clients can now access these algos without even having to build an algo out. So we've add value, I think, and it's something the clients need. And then obviously along the chain, you have the end client, the exchanges, and the ISVs, and then the FCMs. And the FCMs are basically the clearing merchant or the bank. At the end of the day, they're there to hold the capital for the client, to meet margins, be that face and the gateway to the exchanges for their clients. And at the same time, they're always trying to find better products and better quality service. So working with the FCMs who really normally don't have the capability of

building out execution algos, and it's not their focus, right? And so us partnering with these FCMs makes it a lot easier, and especially through the gateways of the ISVs. So I think along the chain, everyone is being enhanced, and at the same time, every one of these groups can get access, have access to our service on an on-demand process. So the exchanges get to help out their biggest clients, the biggest of the big firms. The ISVs get to enhance their product. Basically, they're just providing them a screen now. I mean, there's a lot of bells and whistles, but it's a screen enabling them to execute.

And now they get a little extra tool there. And then the FCMs, similarly, hey, we're more than just clearing. We got some value add here. We can provide you with these algorithms. Yeah, and we can commission the ISVs and the FCMs along the way for introductions. And listen, they have a lot of investment into these clients, and we respect that. So we work well with the FCMs and the ISVs and try to make it a partnership along the way. David, you touched on this.

In today's world with technology costs, you'd be crazy to build your own algos. I mean, if I'm a $20 billion hedge fund, is that a different story? Or like what's the sweet spot? Why don't people just take this on themselves? Well, part of it is division of labor and what you want to specialize your resources. And even if you're a $20 billion hedge fund, you might really want to develop your expertise on predicting where the market's going or predicting where a sector is going or where... Or asset raising.

Yeah, anything like that. And it just might not be what you want to focus your capital expenditures on becoming experts. But the other part is just the kind of fixed costs of getting into the space have increased over time, at least in several regards. Market data costs are a lot more expensive now than they were in the early 2000s. Market data volumes are higher and messaging rates are higher. Now everybody who's doing this professionally is consuming depth of book data versus, say, top of book data. So it's a lot more data to consume. The expertise to be able to take those streams of data and not fall behind and start queuing up and getting stale on the data you're looking at

takes more investment in software or hardware development. And having the colo presence, the fast connectivity, the staff to monitor all the systems. It adds up. But surely a bank like J.P. Morgan has the bank role to do all that? In terms of as an organization like that, could they take their revenues or their assets and invest in this if they chose to do so? Of course. But the question is, is that what they want to do? Is that what they're good at doing? Sometimes a large bank has also overhead in terms of compliance and procedure and a lot of regulatory risks that they tend to be very conservative about and have a lot of procedures in place around code review, code release, around regulatory review, regulatory release,

and that slows down their cycles in terms of being able to enhance and release algos in an agile sort of fashion. And so that opens up opportunity for third-party providers to still have high-quality control but get to market quicker with enhancements and research and improvements for algos? Yeah, I think even along that way, I think our stack of technology, which is C++, and I think we continuously focus on improving the stack technology, Strategy Studio, and our team focuses purely on execution algos also when a large client usually

has multiple clearers so if that's the case they want consistency with placing the order in the isv or through a fix that they may use whatever they might use to place the order they want it to be seamless and smooth going to all their clearers at once and we provide that for them where if you're clearing one group that has execution algos and the other clear doesn't have them it creates an inconsistency and once again i think you know david hit on it with regulation and compliance and updating and trying to maintain and keep quantitative programmers working with the bank sometimes is difficult in this in this time so our stack our

experience coming from the practical trading side and then our technology and who who wants to work at a bank no offense taller bank listeners out there but yeah i don't want to say that we work really nicely with a lot of the banks i think there's some great banks out there and but the top tech talent wants to have like a ping pong table and a right they want to be working somewhere else the biggest challenge these days is keeping talent in the industry and not going into like talent wants to have like a ping pong table and a right they want to be working somewhere else the biggest challenge these days is keeping talent in the industry and not going into like facebook or google or biomedical research or self-driving cars which they should look in the mirror right like google rolls out tons of new products and stuff they're still entrepreneurial

despite their size because they don't have all that regulation and maybe it's coming with a lot of rumblings but they can roll out all that stuff rather quickly um despite their size so i want to get into the algos a little bit um but before we do that can you take a quick second and just get a few definitions out there so we we've thrown around microstructure, maker, taker, liquidity provider. Can you just give us some base goalposts of what we're talking about with all that? Sure. So market microstructure just refers to the details of price formation and a market where some people are trying to buy something and other people are trying to sell it. So it can either refer to theoretical concepts as like what's the role of a market maker

and what's optimal market making strategies given things like inventory risk aversion and the need to kind of cover their costs. Or it can refer to practical microstructure in terms of the rules that the exchanges, the technology connections to the marketplace, the protocols that you transmit orders and receive market data via, rules for how orders get matched, which falls into both a little bit of practice and theory. So an example would be a large order comes in that's larger than the largest order on the book. Basically, who gets that next chunk of volume? That's a small example of microstructure. Or even, yeah, like what are the rules for what price points you can post at

if there are three people bidding at the same price and an order comes in that matches against those three people's orders. Who gets filled first? How are the fills divided up? Sometimes it's not first in, first out, but there's pro-rataata allocation so it's all that kind of down and dirty rules about how orders actually match is it fair to say there's groups out there with teams of brilliant people studying this microstructure in order to gain an edge absolutely yeah like right they they figure out hey if you post way up here the the exchange doesn't allow it it gets kicked out but the prices move up to it

whatever that's and with the granular granularity out there with the data now and the systems and the computing power it's amazing you know how things have even changed to where techniques that have been around for a long time are now relevant because of the computing power and the speed of these computers now. And you asked about liquidity providers. Liquidity provider is a more generic term than market maker. It's a trader who's out there posting orders at passive prices. They might be momentarily a liquidity provider because they're looking to get an order done

but looking to try to post it and not cross the spread. It's not really a true liquidity provider. A true liquidity provider is really serving the purpose that they're trying to add volume, not because they have a directional signal, but because they're looking to facilitate a continuous market. And so that could be a high-frequency trading firm that is fulfilling the equivalent of an official market maker who's agreed to always place a continuous two-sided market, but with a less official capacity, they're providing either bids or offers or both to add liquidity to what's available to trade against in the marketplace.

All right, so now let's get into the algos a little bit. So you've got all sorts of cool things, prowlers, icebergs, implementation shortfall, VWAP, TWAP. Walk me through, I don't know, your favorites. So we talked a bit about scheduled algos. So those are things like the TWAPs and VWAPs, implementation shortfall, close, those all spread out the parent order into smaller child orders, spread them out over time to minimize slippage or risk-adjusted slippage, typically employing some randomization and other anti-gaming techniques to

minimize their footprint into what they're revealing about your intention to the market. And real quick, parent, child, that's my theoretical. I've got 1,000 crude oil to buy. That's the parent order. That's right. The child order, I'm splitting it up into hundreds. Correct.

And so VWAP trades in a volume-weighted way to try to blend into typical volume patterns over the time you're trading. TWAP is time weighted, so it trades at a roughly constant rate over time. Implementation shortfall and close trade more rapidly at the start or end of an order, shifting more of the execution either upfront to balance the slippage of trading fast upfront with risk of spreading the order out over time. And effectively, it samples more of the prices that exist at the arrival of the order. So it's an arrival price algorithm. Implementation shortfall, the phrase comes from the idea of you have an idea of what you want to trade, maybe you run a back test and

assume a fill price of your decision time. Implementation shortfall is the difference between that decision time price and the price that you actually get when you execute. And so minimizing that is called minimizing implementation shortfall. They should rename Congress implementation shortfall. Start a petition. It closes kind of the mirror image of that. You might be a trader that's benchmarked to end-of-day prices, and you're looking to sample more of those end-of-day prices. You'll pay a little more in slippage to trade more actively towards the end of the day versus spreading it out over a longer horizon,

spreading out more evenly over a longer horizon. And then Prowler is RCMX-branded algo. What it really is is a hybrid of randomized iceberg and sniping and pegging with reference to the passive side of an order book. And depending on what parameters you set, it can have a variety of behaviors. Can we go through some of those? So iceberg, you see 10% of it at the top, 90% of it's below the surface. That's the concept there? Yeah, so a typical iceberg, and often exchanges have native iceberg functionality, means you have a total order size and a displayed order

size, often also called the max show. And as the displayed order size is executed, eventually it replenishes and posts that size again. And when an iceberg is implemented by an exchange, it has the advantage that that replenishment happens immediately and with no latency. But it also has disadvantages in that depending on the exchange, it might be more visible. For example, on CME and their order book feed, you can see the order ID of a native iceberg replenishing as that tip of the iceberg replenishes. And so to gain additional stealth, you can have a broker implemented iceberg or an iceberg implemented by someone like us, and by sending fresh child orders every time you don't have that order ID visibility issue. If it's implemented on fast technology, you have less of the disadvantage of not having that ability to replenish immediately by being at the exchange.

And then you also gain access to additional randomization features, like randomizing the size of how much you're showing, randomizing the timing of when you replenish. And then like in our Prowler algorithm, additional functionality, like not always posting at the same price or posting at multiple levels in parallel or outside of the market to snipe at it with a fill and kill or IOC order. And everything you just said is the computerization of a skilled broker back in the day in the trading pit? Or have we gone beyond what a human was capable of doing? No, I think David can get more technical on it, but I think the basis of most of these algos are really, the basis are what humans were doing. They're obviously intuitive of what's happening in the market. They're reading the information.

They're in that pit or they're watching from a desk for years, and they see patterns, recognizing those patterns and trading chunks at a time or whatever you might want to do. But I think that's the base of a lot of these algos. Then you get to the next level where you're putting fast technology to it and you're having learning tech around it. But I think the base of most of these Elgos are really from what the humans saw and how they traded it manually.

And we just enhanced it, and anyone who had built an execution Elgo would have enhanced it. That's my opinion. And even back then, right, if Jimmy was filling paper for Goldman or whatever and every time he came in with 100, you knew there was 900 back then, right, if Jimmy was filling paper for Goldman or whatever, and every time he came in with 100, you knew there was 900 behind it, right? The floor guys would catch on to that pretty quick. They would kind of get a feel for the actions of the trader, of the broker, of how he was positioning himself and who he was first going to get a trade done

with, knowing that, okay, if I can get it started here, I got more to do. There was a lot of techniques that were legal but also very identified. Listen, I think there's an iceberg behind that. But, you know, it wasn't on purpose. It was just I think now the computers, there's a lot of gaming that happens that detects icebergs, not just, you know, the order ID coming back up, but the size of the order coming out systematically, the timing of it, and other things. And then also just the fact that the time of the day or maybe what's happening in the market with high correlated type of technology, they can see, okay, this is happening in the market. All of a sudden, oh, this should be when there should be a whole series of icebergs coming out. David could probably enhance on that a little bit more. Yeah. So really is kind of product specific in terms of how much liquidity you might expect to find. And in different products, you have different incentives in terms of how you post liquidity

in pro rata markets. You're incentivized to show more size to get pro rata fills. And all these kind of lead to different behaviors in the order book. And certainly looking for patterns where traders are placing orders can gain insight into kind of their intention of the flow and is there someone like if in an illiquid market is the spread wide and does someone keep on coming in and inching up and cutting the spread and then chasing a fill cutting the spread chasing a fill so you can kind of look for these repeated patterns to sniff out the presence of traders looking to work a larger order

and making sure when you're writing algorithms to execute that you're thinking through what are the feedback loops that the behavior of the algo could have on the market, how do you minimize these things, minimize the detectability is important, not having the market move against you as you're filling your order. And do you guys use AI at all to make those decisions? Or is that your trading background helps you better make those decisions on what should be looked at with the algo parameters? So AI is a pretty broad term in the sense that you might take an AI class and the first thing your professor might teach you is still basic linear regression.

So what falls under AI and what doesn't is kind of a gray area. And there's a lot of ways to apply AI in a trading algorithm. Like you could have something like a deep learning routine that's looking at maybe optimizing the parameters that's governing something that's more rules-based and deterministic. You could use AI for forecasting volumes or liquidity or volatility in the market or whether the next tick is going to occur on the bid or the off or maybe even short-term price changes. So certainly part of our R&D is vetting which techniques are most helpful for which subsets of that problem. And you're currently working on a splicer? Yes.

Is that the same as a spreader or a legger? Legger, not necessarily a spreader. Yeah, so a splicer is our multi-leg algorithm, and multi-leg algorithms are something that is in high demand in the futures world where a lot of the volume is spreading between different products or different X-rays in the same product. There are exchange-listed spreads and even exchange-listed intercommodity spreads, which help make sure that the execution is atomic and that you don't pick up lagging risk. But sometimes those exchange listed products don't

have a lot of volume or they don't cover every sort of combination that a trader is looking to trade. And a multi-lag algorithm comes in and fills the gap and helps the convenience of allowing a trader to place one order for a combined sort of product and helps try to manage the risk that they get lagged on the fills. Often if you're trading something with multi-legs, some of those legs are negatively correlated pretty strongly to each other, like a calendar spread, a long one month, short the other. That calendar spread's price is going to be a lot less volatile than the price of the underlying. And so if you take just one leg of that fill, you're exposed to a lot of price

risk before you get the other leg of the fill. And being able to keep that in line while also working the objective of minimizing slippage is a tricky problem, especially as you're possibly working orders on products that are in different data centers on different exchanges and keeping those legs working both with the benefit of fast local market access, but also fast inner data center coordination on what prices are actually existing at the moment for the spread product. And maybe there's a limit price in place that should shut off the bidding or constrain the bidding. And where is that limit price? How do you get that knowledge? It's essentially trading a synthetic price, but you have to actually be able to trade it.

Yeah, and to do it well, you have to be able to calculate that price and act on it quickly in a distributed way. So let's shift gears a little bit. Coming up on the new year here, where do you see hedge fund industry in general, the futures industry headed? Are we getting ever more automated, way more AI coming into the market, way more volumes? Volumes have been increasing year over year over year, it seems like, for the last 15 or 20.

Joe, where do you see things going? Well, in the futures world, I think it is changing, where it's going to be much more quantitative. I do believe the equity worlds use a lot more quant techniques over the years, and I think a lot of those traders have come over to the futures industry and hence I think to be competitive you have to get more quantitative. I think there will be a lot more alternative data sources that will be coming in, not just daily snapshot of weather or inventory or government information that comes in. I think it's going to come in a lot more readily and it's going to come more by API. We're prepared for that. Our technology takes in, we could take in any data that comes through an API and normalize it. So we're excited about

the growth in that area. I do believe many of these big energy firms, I keep referring to energy firms because we deal with them on a regular basis. They have $180, $200 billion of gross revenue, and maybe they net maybe 1% or 1.5%. So they're still making and generating a lot of money. They have the resources to buy out and work more with data science and techniques and technology. So I think it's going to be an interesting year of quantitative trading and approach by these big firms. I think the hedge fund world, I think there'll still be some

consolidation. I think that the fee-based structure, I think you look at Magnetar and some of these other groups, I think they're being somewhat disruptive in some ways, but I think the fee-based structure is going to change and continue changing. I think there will be a consolidation. That's just my opinion. And I think overall, I don't know if you mentioned the markets, I think there's still a lot of cash on the sidelines. I think there will be volatility, but I think the volatility will happen in short waves. And in the equity markets, I think there's still a lot of reason for buyback. If you look at most of the filings of these corporations,

they have a lot of buyback still they can do, and they filed for, and I think any time there's dips, they're buying. So, yeah, and I don't know, you know, I'm not sure where it will go from there, but I think. And talk through that a little bit with me. This concept, these firms are huge, right? You're saying these energy firms, whatever, pick one. Billions and billions of dollars have data science teams, have data analytics teams.

But it seems to me they're missing that last mile, right? So they might have all these teams in there to do all this data analytics, instantly hedge in this market when X happens in our storage or when this tanker goes off course, whatever the case might be. But they're missing that last mile, as I call it, of the actual execution piece, right? Like those data scientists might be able to figure out exactly how and when to hedge, but then they're just going to come in and cross the spread. It seems like they could be missing a big opportunity there.

Yeah, I believe there's investment going on to support more of a quantitative approach by these firms. I think they're hiring more PhDs. They're finding better techniques. And I think there is still a gap, and that is you can take all these scientists and these approaches that work in the marketplace, but I think you need a centerpiece to bring it all together with traditional market data, alternative data. And what does that mean? When you hedge your physical, when you be more offensive if you have a predictive pricing that says, okay, I think oil is going to go up to this much, or I think corn is going to drop by this much. What do you do with it? And I think, you know, we're excited at RCMX because we've completely focused on our execution algo business, but we have deployed software. And we're looking forward to servicing our clients in the future with the

quantitative alpha side of the execution or the algos, I should say. And we're looking to support our clients long-term with our technology, and I think that's going to be a game-changer for us in two or three years from now. But I do believe quant trading is just at the beginning. It's just at the surface in the futures world, and I think it's going to keep scaling from here. Off topic, we got Star Wars Rise of Skywalker. What is it? I'm having a brain aneurysm here.

Yeah, the Rise of Skywalker coming out tomorrow. Dave and I are going tonight. Both have our Star Wars shirts on. Who's wearing it better? You got more like it looks like David Cassidy on your shirt I don't know which character and David's got more of a Star Wars look I don't know this is Luke Skywalker I don't even know who David Cassidy Luke Skywalker family yes I have Partridge family come on um cool and let's run through. I'll go.

We'll alternate between you. We like to do on the end of every pod favorite things. So I'll start with David, and we'll go back and forth. David, favorite book? Favorite book? In general? In general. Rapid Fire.

Virginia Woolf, The Waves. Virginia Woolf, The Waves. What was that about? I don't think I read that one. It's a stream of consciousness, modernist literature. Yeah, good stuff. All right. I'm thinking Virginia O'Keeffe, not the same person.

Virginia Woolf. She's like the cattle skull artist. Joe, favorite restaurant, non-Italian like i like uh i like joe's stone crab nice the the miami or no actually i like chicago i miami i don't care for that much they don't have steak there and i think they're a little pompous i like the chicago but i get really mad at someone i go to joe's stone crab and they don't order Stone Crab. Like you're missing the whole point. Yeah, well, we have Stone Crab here in Chicago and we have steak too, so it's great.

All right, favorite Italian restaurant? That's a tough one. Go to David on that because he's actually our foodie. Osteria Longe. Oh, good. Where's that? It's in Humboldt Park. All right.

David, favorite investing book? Favorite investing book. I'll loosen the rules to quant type book. I'll go with Tillio Mucci's asset allocation book. Okay. Joe, favorite place to vacation? I still love Orlando. I love going to orlando we our family goes to the dolphin and we feel like it's our place i just love it and we love going to the dolphin hotel down there i know it's you ever see book of mormon i have not seen

the book of mormon david? The Orlando scene? The Orlando scene. No. All right. We'll look at that afterwards. David, favorite Star Wars character, apropos for tonight? Favorite Star Wars character. I guess I'll say Luke Skywalker. I'll answer that.

I don't know. I think I've seen bits and pieces, believe it or not. I like that in the theme of this execution, I'll go in technology. I'm going to go with that vacuum cleaner guy, R2-D2, is that his name? Vacuum cleaner. Yeah, he looks like a car vacuum. All right, you're kicked off the podcast. Is that his name, R2-D2?

Yes. Okay. He doesn't do any vacuuming. Well, he looks like a car vacuum. He's an astromech droid. He fixes things in outer space. Okay. Yeah.

R2-D2. And lastly, David, there's a rumor you're an opera fan favorite opera i'll cheat a little bit and say the ring cycle three four operas the ring cycle is three separate operas four operas four david is an excellent piano player too he's got a baby grand at home nice all. All I know, I went to the German opera. What's the long one? That might be one of the rings. Several. But I equated it to being drugged, stuffed in the trunk of a car,

and driven to Mexico. I was hot. I was itchy. I was falling asleep. It was terrible. And that happened to you a couple years ago, didn't it? That happened? Yeah.

Oh, the Mexico thing? Yeah. All right. That's it, guys. Thanks so much for coming along. Thanks. I appreciate it. Yeah. All right.

You've been listening to The Derivative. Links from this episode will be in the episode description of this channel. Follow us on Twitter at RCM Alts and visit our website to read our blog or subscribe to our newsletter at rcmalts.com. If you liked our show, introduce a friend and show them how to subscribe. And be sure to leave comments. We'd love to hear from you.