"We set out to solve an enormous technology question. The technology risk on day one was very high. To some extent, it wasn’t very rational..." - Dave Sanderson (Tweet)
KFL Capital Management is setting out to do something every financial team on earth dreams about...
The ability to predict the future.
Or at least get it right fifty four percent of the time.
Our next guest on Top Traders Unplugged is the CEO and Co-Founder of KFL. In this episode we explore their trading strategy and uncover the fundamental differences between what they are doing that makes them so different from traditional alternative investment organizations.
Thank you for listening in on our conversation with, Dave Sanderson.
In This Episode, You'll Learn:
- The value of Battle of the Quants - Hosted by Bartt C. Kellermann
- Transitioning from commercial litigation to wholesale mutual fund vending
"There is something intuitively intriguing about the investment business." - Dave Sanderson (Tweet)
- How Dave Sanderson was exposed to alternative investments in the first place
"We’re really a data science firm." - Dave Sanderson (Tweet)
- The convincing required to get top big data scientists to work on financial challenges
- The comical story of how carefully big data scientist come to conclusions
- What Dave Sanderson loves to do when he isn’t working directly on KFL Capital Management
- How Dave Sanderson sees the deviation between machine learning and systematic trading
- What it means to exist in a deluge of big data
- How Dave Sanderson and KFL perceive themselves and the usefulness of labeling
- Is machine learning a superior method than conventional approaches to trading?
- About the choice of Krystal as a name for their fund
- The structuring challenges behind KFL Capital Management and why they are more like a tech firm
"We’ve had a level of communication with our shareholders that’s probably very unique." - Dave Sanderson (Tweet)
- The focus for expanding KFL Capital Management
- About evolutionary computing and how KFL Capital Management grows with the markets
"We retrain the model every single day. So the model is using more and more information as its training data." - Dave Sanderson (Tweet)
- Is there an environment which would be optimal for Krystal?
- Is there an environment which would be severely challenging for Krystal?
- The use of non-parametric modeling and why this type of prediction takes KFL out of most conventional finance sector buckets
- Plus much more...
Resources & Links Mentioned in this Episode:
- Attend Next Year’s 10th Anniversary of Battle of the Quants.
- Learn more about the team behind KFL Capital Management.
- Man AHL - Dave Sanderson’s initial exposure to alternative investment.
- Thomas K. Hunter - “He’s probably done more tech deals than anybody in Canada.”
This episode was sponsored by Swiss Financial Services:
Connect with KFL Capital Management:
Visit the Website: www.kflcapital.com
Call KFL Capital Management: +1 (416) 849-1925 x212
E-Mail KFL Capital Management: Info@KFLCapital.com
Follow KFL Capital Management on Linkedin
"Unless he could prove it, from mathematical first principles, he wasn’t going to believe it. It’s like the physicist who says, ‘Sure it works in reality, but can you prove it in theory." - Dave Sanderson (Tweet)
The following is a full detailed transcript of this conversion. Click here to subscribe to our mailing list, and get full access to our library of downloadable eBook transcripts!
Welcome to another episode of Top Traders Unplugged, where my goal is to give you the clarity, confidence and courage you need to invest like, or invest with one of the top traders in the world. It is the stories you never get to hear, set out as the most honest and transparent account that I can make of what goes on inside the minds of some of the best investors in the world delivered to you via a one-on-one conversation. Today you're listening to episode 49. If this is your first episode, I suggest you go back and listen to all of the previous conversations? Before we go any further let's find out who is on today's show.
My name is Dave Sanderson, and I'm the President and CEO and one of the founders of KFL Capital Management. We're a machine learning firm that trades the futures markets, and you're listening to Top Traders Unplugged.
Thanks for doing that, Dave. By the way, if you want to read the full transcript of today's episode just visit the TOPTRADERSUNPLUGGED.COM website where you can find lots of great details about today's guest. Now let's get started with part one of my conversation. I hope you will enjoy it.
Dave, thank you so much for being with us today, I really appreciate your time.
My pleasure, Niels.
Now Dave, we met only a few weeks ago at the Battle of the Quant, in London, which for those who are not familiar with it, it is a series of one-day events that has been hosted by Bart Kellerman for the past nine years, I believe. I want to just say that I think it was a really valuable event with some excellent speakers and panels and highly recommend that people attend one of next years events, as I know that Bart has something really cool planned for the 10th anniversary. How did you find the conference by the way?
It was wonderful, Niels. Bart does such a great job, and he's connected. He's like the connective tissue to everybody quant. So whether you're a service provider, an allocator, a manager, I would recommend it highly.
Absolutely, and of course the bonus was that you and I ran into eachother, and there we are, a couple of weeks later you're on my podcast, so that's a great outcome.
Now Dave, you and your partners have come up with a very different way to trade the market compared to my previous guests. It's intriguing; it's cutting edge, and I think the audience today will really have their eyes opened as to the possibilities we have today when applying the latest technology to financial data. But I want to ask you a questions before we get into that which is completely different and just something that I find interesting to hear the answers, and that is if you meet someone that you haven't met before in a social context and they don't know what you do, how do you explain in simple laymen's terms what you do for a living Dave?
Well it's a great question, Niels. I must tell you, leaving the Battle of the Quants; I felt like I wasn't doing a good job getting our message through. So it's both difficult in a sophisticated audience and at the cocktail party as you've described it there. In the cocktail party I say things like, " we're really a data science firm. You've heard of big data? That's what we do, and we search huge mathematical space and we predict where prices are going in the next few hours of futures and commodities," and usually that gets a few questions coming back.
(laugh) Yeah, I'm sure of that. Now, we're going to stay with you for awhile longer because to me numbers when people look at a firm, a manager, don't really mean a great deal without the story, the context. So what I'd like to do is for you to go back pretty much as far as you want and tell me your story and how you got from where you were as a young man or young boy growing up to where you are today.
Well, that's... I just want to pick up one more thing on the introductory question because it was actually you that gave me the clarity on this. So you and I were chatting about trend followers, and I have a great affinity for trend following. In fact when I was a discretionary trader that was my style, so we used to say around my shop, "if it's working do more of it and if it isn't, stop" and that's sort of the underlying thesis around trend following. When something starts to move, get on it, and it becomes an issue of entry and exit, but really everyone is reacting to price moves, and that makes a lot of sense. That's the big distinction, is that we're predicting price moves as opposed to reacting to them. So I just wanted to put that out there as a contextual point and then maybe it helps everyone understand as we go through the next number of minutes.
Absolutely. So Dave, back to your young days, what were you like as a kid and where did you grow up?
I actually was born in Westchester County, New York, and had the pleasure of flying back there recently with the KFL business cards. So that was a great moment of reflection as the plane banked over Westchester County and landed at the private little airport there. We came back to Canada when I was less than two years old and I've grown up just outside of Toronto, Ontario, Canada, which is a wonderful place to raise children, but a challenging place to raise money. We have a wonderful city. I now live in London, Ontario, and my career started in Toronto. I was trained as a lawyer and licensed as a lawyer in the early 1990s and practiced commercial litigation for the better part of the 1990s.
That was a wonderful journey. It was great to be trained in that way, but the parts of that business that weren't consistent with my personality were that it's all about conflict. I prefer to build things and agree with people than I do to fight with them. Secondly, we're usually fighting over things that have already happened in the past, and I find myself much more future focused than past based. I guess, in retrospect, it was no surprise that that career was not going to be my final landing place. I was then brought into the investment management business, first wholesaling mutual funds to stock brokers in the greater Toronto area, and subsequently managing a couple of retail brokerage offices in Toronto and London.
Can I just stop you there for a little bit, because that's an interesting transition from commercial litigation to mutual funds. What actually happened back then that sort of led you to take that jump.
That's a funny story. I was sitting at my desk on the 54th floor of Commerce Court, in Toronto on Bay Street, lawyering away and I get a cold call from what turns out to be a headhunter who's establishing a sales force in Toronto of Invesco. I responded to that call somewhat surprisingly to myself and so I find myself shortly thereafter in an interview and started the interview with a wonderful guy named Dave White, who's now a shareholder of KFL and has stayed in my life ever since. Dave came into this boardroom that was not nearly as well equipped as the Stike and Elliot boardroom where I practiced. He shook my hand and said, "why?"
Because I wasn't in interview mode, I said, "why what?"
He said, "why would you ever leave the practice of law?"
I thought for a moment, and I said, "have you ever practiced law?"
He laughed, and we became very good pals. You know what? There's something intuitively intriguing about the investment business. I think it touches so many people that I was just intellectually curious and it was the right time, and the right moment, and I was probably overconfident as well, Niels, thinking I could make it whatever I chose to do. Life is funny, and serendipity happens and certainly that's part of my story. I feel very grateful to have ultimately met the partners I have met and to be working on such an important project, but I'm jumping ahead, so I'm happy to stay with any part of the story you wish.
Yeah, so for how long did you work with mutual funds before you moved on?
I started in 1997 and between wholesale and mutual funds and being a retail stock broker - I bounced back and forth until 2005 and then I began managing. In Canada there are basically five bank owned brokerage firms that have the lion's share of the market, so one of those firms I began managing the retail office for in Waterloo, Ontario and that's where I ultimately met the people who've become part of this KFL story because, as you'll learn later on, Waterloo is a very interesting geography for engineers, data scientists and lots of tech innovation.
So I was managing a retail brokerage office in Waterloo. I met the M&A tech lawyer who has become, really, the connector for all parts of this story. I started that in 2005. I then got hired to manage a different office in London, Ontario. I moved here in 2006, so I've been in London, Ontario since late 2005. I began managing quite a large retail brokerage office - one of the 10th largest in Canada. I had the great fortune to meet my partners Bob Siskind and John Drake, who are about a generation older than me and have had wonderful experiences, and we just clicked as soon as I met them. You kind of have to take these geographic stops before you're lucky enough to accumulate all of the parts of a really, really interesting story.
Sure, that's true. Now as part of that part of the journey - the retail brokerage business, was there any time during that where you got exposed to the alternative investment side? Because to me, I guess I'm talking from ignorance here, you know retail brokerage is usually about stocks and bonds and not much else, but how did you initially... or did you have any exposure to the alternative side hedge funds, etc., etc., before you got introduced further to your coming partners?
That's a great distinction and a great question. Not only is it only stocks and bonds, but it tends to be one direction, and that's long. So that business, I have a great affinity for the retail brokerage business, but it is very much about asset gathering and service providing/fee charging, so it's a very, very different business. If I knew anything about alternatives it was because, as I made the transition into retail brokerage, one of the firms that I interviewed with, and that I was considering working with was Man AHL, and I flew to Chicago a couple of times to see the operations. I was blown away by the operations of Man AHL and the people there. That was a wonderful story. So that was my first real introduction to alternatives. As you've rightly characterized, the retail brokerage is virtually absent that kind of product and in Canada, at least, they make it very difficult to sell. So I had an enormous to learn as we started the KFL project. But as you'll hear, as we go along, we don't put a lot of emphasis on domain knowledge. We're a very scientific firm, so by domain knowledge I mean the knowledge of the space of alternative assets and trade management, trade execution, these kinds of things. We try to keep it a purely data science firm.
Sure. So you meet, I think you mentioned, a lawyer and he introduces you further to someone that he felt you had to meet, is that right?
That's right. The lawyer's name is Tom Hunter. He's a wonderful guy, and he's probably done more tech deals than anybody in Canada. He always wanted me to meet a client of his named Stephen Bacso, and Steve had quite a reputation in Kitchener and Waterloo for building a company called PixStream and selling it to Cisco for 540 million dollars in 2000. So clearly as a retail broker I wanted to meet Steve. So we met, and I was trying to pitch Steve on what I thought was a quantitative hedge fund. It was a fund that looked at four factors mostly around earnings, earnings momentum, earnings surprises. As I pitched Steve, I saw him respond quite positively to the idea that a hedge fund could be quantitatively based, but in the end he had a bigger idea for a quant hedge fund.
So he was working with Dr. Andrew Wong at the time, and they were using Dr. Wong's predictive technology in the genomic space. As I was pitching Steve on the hedge fund, he was further convincing himself (I know he always had these ideas), but further convincing himself that now was the time to use a very sophisticated predictive modeling approach to prices rather than the things that he was seeing in the alternative space for his own money.
Did that come about at the same time he came to you and said, "you know, instead of you pitching me I'm actually going to pitch you." This takes place in 2005, 2006 there about?
So in 2006 you can imagine, I'm a retail broker having a nice dinner with a quantitative hedge fund manager and pitching this perspective client. So the next day I phoned Tom Hunter and I said, "Tom, how did we do?" thinking I might get a big account from Steve. He said, "well Steve really liked you Dave, but in terms of hedge funds, he's actually going to start his own," and I said, "tell me more." When he talked about using genomic algorithms and making predictions on price data in the short term, I guess I intuitively felt it could be done and now was the time somebody was going to do it so I said, "keep me posted."
That was 2006. Steve went away with Dr. Wong and another player you will hear much of in this story, Dr. Gary Li, and they worked for three years until I got another phone call that said, come and see this. I think we've got something.
And this is 2009 you said.
Yes, so June of 2009 I traveled an hour down the highway from London to Kitchener Waterloo and sat in a boardroom with Steve Bacso and Tom Hunter and they showed me the first PowerPoint slides on, what the time was call Knowledge Funds Limited. We've since shortened it KFL. While I didn't, because of my background or my absence of background in statistics and predictive modeling and machine learning, I certainly couldn't verify what I was hearing throughout the presentation. I knew a couple of things for sure: One was that Tom was a very good friend of mine, and I wasn't being misled and there were not misrepresentations. Secondly, Steve could do a lot of things with his time and with his money, and so this was obviously an important project that had some merit.
Absolutely. Tell me a bit more. What happens next in your story?
So I left that meeting slightly confused about temporal suffix trees and other parts of the statistical world that I had heard about. I knew a couple of things and that was if this technology worked it could be exploited for great economic benefit and we could find out if it worked in fairly short order, or at least we thought at the time. Even though they didn't ask me to raise money for the entity at that time, they were really reaching out to a person is the financial services industry and they knew they would need a partner in the financial services industry. So they were sharing with me the story as it had evolved to at that point.
I drove back to London, Ontario and I met with Bob Siskind, who's a very good friend of mine - a very successful real estate business person in London, and I told him this story and about three sentences in, he said have you met my best friend John Drake, who I knew owned a golf course in town - a very beautiful, beautiful golf course, very exclusive, he and his partner owned it. I knew that he had a lot of success in the junk bond era and was just an intellectually curious guy. I hadn't had a deep relationship with John at that point but Bob and I and John sat down and I went through the story with him, and they said, "let's go see it."
So one week later we traveled back down the highway and we saw the same presentation from Steve and Tom, and it was about three hours long. I thought it was going to be fifteen minutes because at about the fifteen minute mark John looked up and said, "I gotta stop you guys, it's one thing to make predictions on scientific data, it's wholly another to predict human behavior and I don't think it can be done. I was feeling at the time that that may be the quickest due diligence ever done.
We kicked it around for awhile and I think the thing that kept us at the table was the idea that if you show human beings the same set of facts they will react very predictably particularly when they're speculating in financial markets. They almost can't help themselves. So three hours later we get back in the truck to travel home, and Bob is sitting in the back seat and he says, "John, what do you think?" And John very profoundly said, "I don't know. I don't think it's possible, but I just can't miss it." In other words, it's all about pot odds. You're going to make a bet, in this case the bet was two million dollars to fund the technology development, and if it doesn't work your two million is zero, but if it does work it's worth an enormous amount of money. So that's how it all started in July of 2009.
Wow, wow, and it took a little while before you joined, or did you join straight after in terms of building the infrastructure?
Well, it's never a straight line, is it? The best plans never seem to go... So what happened was we spent a good portion of the summer into the fall getting a better understanding of the project before the money went in. We even had them produce a set of test trades for us that we said needed to be better than random. They met that in October, a month's worth of trading had been quite a bit better than random. So we negotiated a deal, and we put the two million dollars in, in January of 2010 and for that we got half the company. So the technology development started. I stayed in the retail brokerage world at that point because there was a lot of building to do. Obviously it was very exciting but we needed to get further down the road.
So, what happens next is the really interesting part of this story. Dr. Gary Li becomes really the person on whose shoulders we were relying or standing, and Dr. Lee was one of the students that had been supervised by Dr. Andrew Wong. Dr. Wong, whose a shareholder of our company and one of the founders of our company was the founder of the Pattern Analysis and Machine Intelligence Lab at the University of Waterloo, Ontario and he was trained at Carnegie Melon. He has spent his lifetime in data science before data science ever became a degree; before big data was ever a term, Dr. Wong spent his lifetime in the predictive modeling space. Dr. Wong will tell you, to this day that, Gary Li is the most brightest student he's ever supervised, and he's supervised I think it's 160 Ph.D. students at the University of Waterloo.
So he tapped Gary on the shoulder and said come and work on the financial data set. Gary was very resistant at the time because his experience in predictive modeling was that the financial data set was untouchable. Certainly Gary Li and Dr. Wong are very, very cognizant of what's happening around the world at all the great institutions in the category or predictive modeling and lots of progress was being made on scientific data sets. In fact, Gary had commercialized the algorithm in the Oil Sands. In the Oil Sands in Canada, one of the things you can do with predictive technology is maximize the recovery out of the bitumen in the tar sands. In your models are things like flow rate and temperature and all the sensors that you place around the plant.
So Gary's resistance came from the idea that he was in the prime of his career and he certainly didn't want to waste a number of years trying to do something that there was no evidence that it could be done. It took a little convincing to get Gary on the project. He agreed to do it on a part-time basis originally until some progress could be made. Dr. Wong had some Ph.D. students that had to give up on the thesis because it was too challenging to make predictions on financial data.
So Gary Li started and he made a little bit of progress, a little bit of setback, and really, two and half years goes by and the better part of that two million dollars gets spent and there is no progress to show for it. Around Christmas of 2011 Gary comes forward and says, "it's too hard, it can't be done." We encouraged him to keep going. Gary, actually wanted to keep going. He knew it wasn't a contribution to the academic file on predictive modeling if he just says it's too hard. If he comes back and proves the thesis that you can't do it, now that's a contribution to science.
So that's what he actually started out to do in late 2011. He started out to prove that you can't do it and the two things that he focused on, these two characteristics of financial data that he believed were the reason that it cannot be done and it hasn't been done are these: number one, the correlations among variables in financial data changes over time. So if you think about that Tar Sands example, if flow rate is column A in your data set, and temperature is column B, when you look at the relationship between flow rate and temperature, it's going to be the same on Friday afternoon, as it is on Monday morning, as it is three months down the road. But in the financial data set, if column A is the S&P 500 price, and column B is the price of gold, you can imagine that there is sometimes a relationship, strong negative on Friday afternoon; and sometimes on Monday morning there's no apparent correlation; and three months down the road it can be positively correlated. So that's a very unique attribute of financial data sets. The second attribute is that when you see a relationship in science, perhaps in the genome, you can call everyone over and point to it and nothing happens to it. When you see, or when everybody sees a relationship in the trading world, the world of financial data, it gets arbed out. People trade that relationship away.
So those two things: the fact that correlations change, and the fact that patterns can appear and disappear make that data set profoundly more difficult than other data sets when you're trying to use historical data to make predictions about the future. So Gary started, and it was his proof, he thought, that he could prove that those two things made the data set untouchable. But to his dismay, I suppose, or to his surprise, in the spring of 2012 he actually gets a model that works on the S&P 500. He picked the S&P 500 because, in his view, that's the most efficient market in the world if there is such a thing.
He saw that his model could achieve 52.8% accuracy across about a year and half of daily price data at that time. So he was interested in that, but he quite frankly thought that after he tested it he would find some look-ahead bias in the data, or that he would find another reason why it wasn't true. He expanded the backtest, the baseline test or out of sample test, from a year, and a half to five years, and the 52.8% remained. Then he expanded it across not just the S&P 500 but fourteen other assets and low-and-behold it remained.
So now he was being more convinced but still, in his words, and it's great to hear him tell the story, unless he could prove it from mathematical first principles he wasn't going to believe it. It's like the physicist that says, "sure, it works in reality, but can you prove it in theory?" So that was Gary's view of the world. He literally sat down with pen and paper and built the mathematical proof. The mathematical proof indicated to him that something was wrong. But what was wrong was that it shouldn't produce 52.8% accuracy, it actually should produce 54%.
So he went back to the models, and he optimized the three parameters that he felt were contributing to overcoming those two characteristics that we talked about. When he did that the backtest was consistent with his mathematical proof and ever since we've been both telling the world and trying to validate to ourselves that we have an edge, and the edge is 54%. So that was May of 2012 when Gary comes forward and we amass a meeting of the partnership and Gary shares his findings and what I've learned about scientists is that their tonality doesn't change, so they can say "you're out of business" and "Eureka!" with the same tonality. Quite frankly we missed the headline that meeting because the headline was, "Eureka! We've done it."
We weren't as excited as he expected and hoped us to be, but shortly thereafter we called him back and said OK if you believe, then let's continue. We raise some more money. We hook the model up by way of API - Application Programming Interface no-touch trading to interactive brokers. So over the course of 2012 and we refined the infrastructure such that the data feed would come in consistently and we could treat the live data symmetrically with the historical data. We can get into all that because that's been a huge challenge. Anyway, we get to January of 2013 and Gary has tightened up the backtest and we say nobody in our industry, Gary, is going to believe the backtest. He was a little bit incredulous to that because he had used such scientific discipline to produce it, but he took our word for the reaction of the industry and he asked us what we wanted. The answer was, well let's get 1000 trades. If we can get a 1000 trade sample and you can show this 54% edge, then we will launch a fund.
So my partners and I put $100,000 into that interactive brokers account and the machine in Kitchener Waterloo, Ontario started shooting trades into that account. Over the course of eleven months, over 2013, by November, we had 1,000 trades. The accuracy number was 54.02% which, I don't know about you Niels, but I have never seen a backtest become reality with such consistency. As we move forward through this whole discussion, I will share with you we're closing our one year of our partner fund, which is the publicly distributed fund, and we now have in addition to those 1,000 trades we have 1,700 more trades and the accuracy is 54.04%. So it's amazing, it's just fabulous, but I'm jumping ahead.
That's a great story. It really puts things in perspective, so thanks very much for sharing that. Before we jump into the nitty gritty of things I just want to take a step back and ask you a question and that's, you're obviously busy building KFL and that's a big task, we all know that. Being entrepreneurial in this day and age is not a nine to five job. What do you like to do when you're not spending your time on the business?
I'm a lover of this business, and it bleeds into every part of my life as you've just hinted at there. So I'm going to tell you that one thing that I love to do when I'm not working directly on the business is read, think, and write about the business. We've had a level of communication to our shareholders that's probably very unique. For a while during the 2013 period I talked about, I would write weekly reports on the portfolio and not just numbers but narrative as well. This year we've gone to once a month, and I'm just finishing the twelve-month report now. I have three beautiful children, 18, 15, and 11 and so if I'm not working I'm spending time with them or my wife Debi of twenty-two years. So we have a nice life here in London, Ontario. I play some hockey; I play some golf, and I have wonderful business partners, so that's my life, Niels.
That sounds pretty good. Now, before we jump into the organization and how you've set that up, I wanted to ask a couple of broader questions. Now, you come from this unique world of machine learning as part of the investment process. And in fact you could say that you're replacing the human brain when it comes to making forecasts. Tell me why you think this is different? And why it's important to understand the difference compared to what we normally refer to as systematic trading. What's the distinction?
Well it's a very timely question, and I must say as I left the Battle of the Quants, as happy as I was to be on the new quants panel and to have met all the people there, I was frustrated by my inability to communicate the difference. So you and I chatted post the Battle of the Quants and you made the point to me that makes the most difference, which is mostly, systematic traders are waiting for prices and reacting to those prices. So you can imagine, as you well know, the trend followers of the world are waiting for trends to occur. And sometimes you get on those prices and it turns out not to be a trend, so you get off. But you're really reacting to prices, and that's the key verb, is reacting. What makes us different is that we are predicting where the prices are going, and we're predicting where those prices are going over the next few hours. Our average hold period, we'll learn later in this, is thirty hours.
So, really that's the difference. And these terms, "machine learning, and predictive modeling, and big data," they all get kind of lumped in together and they can be very confusing. And I'm only starting to get clearer and clearer on it. But what I would say is this: The concept of big data is that data is now everywhere. There's a deluge of it. Now what does that mean? Well, it means a couple of things. It means that not only are things like Twitter feeds becoming zeros and ones. Meaning the Twitter feed can be interpreted, stored, and accessed at a reasonably low cost. So there's this expansion of the kinds of data that are available for interpretation. And then secondly, there are bigger and bigger computers that are able to crunch more and more numbers. So to me those are the two parts of big data.
It doesn't really apply to us, and I say that because we're using a data source that's been around forever. All we use is price data. And so that data has been flashing on people's screens for years, for decades, and it's being stored and cataloged, and it's fairly accessible. So really, there's nothing new in terms of our data set that makes us different. What I think makes us different is the number of people who've tried to do this: to just take in historical data and make a prediction about the future movement of an asset. There's very many, but from our review of literature and anybody who's talking, we can't find anyone who's done it robustly or consistently.
Interesting. The other thing I picked up on, just preparing for our conversation, is that you see yourself being a little bit outside of the CTA space, despite the fact that you actually trade futures like most CTA's and you use models, and you're in fact registered as a CTA. Is the label you wear, is that important or not? And maybe you should explain how you see yourself as more important?
Right, that's a great question. When Gary and the team developed the predictive technology, the question was what are we going to use it on? And there are a number of reasons why the futures market makes a lot of sense to point this technology to. So obviously, it's incredibly liquid. There is an immense amount of price data available; it's electronically traded, there's embedded, costless leverage in it. And for all of these reasons, it made it the place to point the technology. But really the underlying technology is predictive modeling, it's machine learning. So we could point it to any data set, and we could point it for instance to cash equities. But when Gary asked for the first list of assets to trade, it made so much sense to say we'll trade the S&P, the Dow, and the Nasdaq, and then trade some other asset classes, like metals, agriculture, or interest rate products. And if we can prove the technology across all those asset classes, then we will validate the technology.
So that's how we began. And then when we changed from a prop account only into a fund, we as you say, had to register as a CTA. We're happy to be registered as a CTA, there's absolutely nothing wrong with it, it just gets challenging when we try to differentiate ourselves from all of the other CTA's and all of the systematic traders, or anyone using a computer-based model to trade markets. So, is that responsive to the question?
Yeah, that's absolutely fine, Dave. Now, we have the classical way of developing models where you hire a bunch of researchers, they look at the historical data, and they come up with rules that we can apply to the markets and make a profit over time. But, I guess I have a feeling that you think it's so, but let me ask you this way: Do you think that machine learning is superior as a method to analyze and trade the markets? And if so, why? Clearly the traditional way has worked for decades for a number of firms.
Yes. So, I would never use the word superior because I think that there are some amazing traders out there. And I think it's a wonderful skill set to have, and it's very unique. The gray traders have created enormous economic value. So I think that all we're saying, Niels, is that we have "a way" to create some economic value, and that's all we're saying. So there's no monopoly on how to do it, and there's no sense of superiority, or if there is it's by mistake. And that's the first point I want to make.
I certainly tried my own hand at trading and I'm keenly aware of how difficult it is. I'm keenly aware that human beings have biases that we're not set up to trade the markets well. While trading discretionarily my partners and I came up with some core beliefs, and one of the first lines in the core beliefs was fundamental and technical analysis. And other forms of hard work will just exacerbate human tendencies that are harmful to trading. So that's my own personal view, but there are certainly wonderful traders out there that create great value.
I think that once we became familiar with what the machine can do, it's hard for us to trade any other way. And so, maybe there's a way of describing it that I've just recently come on to that I think helps and maybe will help you understand why we're so enamored with the machine as opposed to the ways we used to trade. And that is the metaphor of the roulette wheel. If you think about a roulette wheel in a casino and the expected value of a bet by a patron. The house advantage is approximately 5.26%. It depends on where you play, and how many spaces are on the table. But let's just use that 5.26%. If you look at our system, over twenty-seven hundred trades, the expected value of the bet, the house advantage, is 13%. So, if you had a roulette wheel, where you could spin it, and you had a 13% house advantage, it would be really difficult to go back to any other way of trading. Because all that you need is a high frequency of spins, and you're assured of where you're going to end up.
Sure, good way of explaining it. Now, let's jump to the first topic that is sort of more specific about your business. And maybe I should just say that, I guess your model has a name, and I believe that it's Krystal. Am I right in that?
Yes. We were actually encouraged by our marketing firm, who's done a great job for us, Meyler Capital. They've done a wonderful job for us. They said that you should personify the machine. And quite frankly, we hadn't had a name for it, but Gary was fond of talking about the entire infrastructure as being a crystal ball. And he didn't mean that so much in terms of foretelling the future as he did the clarity that we have throughout our system. So we have great clarity on things like slippage and real time. And we have a comparison of our historical trades and our simulated trades over the same period as we're live trading. And we watch that in real time as well. So he felt like the whole system was a crystal ball. So we just took the word crystal and changed the C for a K to pay some homage to KFL that we have the name.
Fantastic, okay great! Now when it comes to your organization, there's obviously a couple of interesting things that come to mind, for me at least. That is that I kind of in part see you as an investment manager, and to me at least, I kind of see you also as a tech firm if I can use that word. How have you structured your business so far in terms of that perspective? Because machine learning is a very new area for me, so I don't even know how labor intensive it is, and so on and so forth, or how computer intensive it might be. So how do you structure a business when you've entered into this type of trading?
That is a great question. So, we've built it like it's a technology firm. A lot of the people on our project have experience building companies, start-ups. And the normal way to do that is very different from the normal start up in the hedge fund world. The normal start up in the hedge fund world, as you know, is a limited number of people with some expertise, and they're proving that expertise in a very low cost way. You need an account; you need some licensing, etc. We set out to solve an enormous technology question, so the technology risk on day one was very, very high. And to some extent it wasn't very rational. If you think about putting two million dollars in to solve a question that virtually opened check books that had been trying to do for decades. It was a bit audacious to think that we could get there on a couple of million dollars.
Well, we were wrong; we have raised just over five million dollars to fund this start up. And we're raising more money. Because in our view, if you get this right, the ultimate value of the business is enormous. So, starting out on day one, we tried to solve the technology problem. We felt that at that time it was a great business because if you solve the technology risk, the business execution risk, we thought, was very low. So here we are five years later and five million dollars later, and we're now telling everybody, "Hey, we've solved the technology risk!" Because you can't do what we've done over twenty-seven hundred trades by luck. It's just virtually impossible. And so the technology risk, we feel, has been reduced to something near zero. And now we're in business execution risk. It's a lot harder to raise assets than we thought five years ago. But this too shall pass!
So your question was about structure and how we view this business. We view it as a very, very exciting intellectual property start-up. And we've come a long way in a relatively short period of time, with a relatively small investment given what we're trying to do. And then secondly in terms of the manpower that you talked about, it does take a lot of manpower to run this business. So we have a science team of six people. These are both data scientists and what we call data engineers. Once you've figured out the really big question about how to make predictions on time series, financial data, you then have to build an infrastructure around it that is as subtle and elegant and sensitive as the models themselves.
And by that I mean taking in real-time data, processing it and working with it in a way that makes it one hundred percent symmetrical with your historical data set which is required for a model as sensitive as ours. You have to have a perfect symmetry between live data and historical data, or you'll get different trades in you live trading then you will in a simulated trading experience. So in order to do that, we had to do a lot of work on our data feed. So we have three full time people who watch the data feed, who work with historical data, who make sure we're connected to the places where we trade. So there's been a lot of what we call engineering challenges beyond the scientific challenge.
So back to your question about manpower. There are six people on the science and technology team. So you can imagine what that costs per month. And then the data feed, I mean we pay Thomson Reuters twenty thousand dollars a month for live and historical data. And then you have on top of that in this country, you need to be a regulated entity under the purview of the Ontario Securities Commission. So we have to have those licenses, and report to that body and keep all that infrastructure going. So you add all that up, and it's a high monthly cost. But once again, in our view, if you get this right, you can, in fact, make predictions on financial data, then you've created an entity worth an enormous amount of money. The real questions then are, well, is it sustainable, and is it scalable? And we can talk about those two things as you wish.
Sure. The model itself, I mean I can fully take on board the data feed and so on and so forth. But obviously that kind of work doesn't really change. In that sense, the data feed is the data feed. In terms of the model itself, does that just run by itself? Or do you have someone overlooking it at all times?
It is far more the former, it runs by itself. Now it runs by itself, so we do trade twenty-four hours a day five and a half days a week. We trade twice, or we make predictions twice in each of the twelve futures assets that we trade. Just to give you some kind of idea for frequency of trading. At various points around the clock, we are making trades. But there is no human being making those trades. Those predictions are produced by the models, and then when a prediction is made, the trade execution is by way of no touch trading directly to the FCM. So the only human supervision is a check and balance type supervision. We have all kinds of real-time checks and balances that make sure if a prediction is made, it gets sent to the FCM. If it gets filled at the FCM, we get a confirmation. If there's any discrepancy between all that, we get an alert. But it's very much an alert system, as it is opposed to a human overlay. Watching the trade get generated and putting any kind of discretion on whether that trade ought to be sent to the FCM or not.
I'm curious about one thing. Which is…I mean you're at a very interesting stage, I think, in your business - a very exciting stage. It's kind of the..sort of initial growth phase. You've got all the basics covered, and you know, you need that little break to get your AUM to a certain level. Where you can then start to even further expand. I want to just try and get into your expertise and your brain in a sense that from an entrepreneur point of view. Because many of the listeners today will also be people who are probably not fully established, or people who even are just thinking about starting their own business. How do you see expansion as a concept? Because it's really a risk every time, you expand. You know, are you expanding ahead of time, or are you expanding sort of reacting to increased AUM coming in. I know you're already ahead of the game here because you're looking at it as a start-up. You're raising money to fund your business, which as you said, is very different to how most, I think, managers start out. But how do you from that point of view, view your expansion, and let's assume that assets come in at a decent level..where do you want to focus that expansion, and what's critical for your success in terms of the areas of the business?
Hmm..I thought you were going to finish that question slightly differently, so let me respond to that. So, people who are listening and wondering about expanding and funding that expansion in their early days…there is one clear thing that I would suggest. And that is to get some patient shareholders. So we have the most wonderful, patient shareholders. All of whom have had a lot of success in their own lives, and all of whom are participating in this venture, not because it's just another start up, but because we're trying to do something really important. So getting the right shareholders is absolutely critical. And going back to them and saying, you know, we haven't gotten as far as we've wanted to. We need to fund the operation some more. That is clearly what has killed many great businesses. So I just want to give a nod to our current shareholdership and the latitude and trust they've shown us.
The second pieces are our own future and where we go from here. There is no shortage of ideas on our technology wish list. One of the things that we're doing now, and don't need any more funding to do is expanding the number of times around the clock that we can trade. We've tested the accuracy of our model at various times around the clock, and it just doesn't change. So we're getting ready for higher AUM. Right now at five million dollars, there is no problem with any kind of slippage in these markets...you can imagine. Even though there's no problem with it. We track it, every day and every trade. But we're getting ready for higher AUM, and we're multiplying the number of times that we can trade around the day.
Secondly, we're looking at expanding the list of assets that we trade. And there are only a couple of things that will stop us from trading an asset. Our view is that if this model works, it will work on any asset except those assets that have an absence of data.. of course. We go back to 1996, and we need hourly data back to 1996 on every target that we trade. So you'd be surprised at how many targets we can't put on our list. The second thing that might stop us is if there's too wide a spread in an asset, then it just erodes the P&L from that asset. So barring those two things, we can expand this list.
I don't know how far I read CTA's who say they trade a hundred and fifty markets..I just can't get my head around that right now, but there is some number that is certainly greater than twelve. The other thing that we'd love to get to work on is cash equities. And we really like it because clearly the data is there for the more traded cash equities, the spread is nice and tight. And we don't need a lot of leverage to make this P&L work very, very well. So we don't need the kind of leverage that's possible in the futures markets. So given all those things that's going to be a wonderful place for us to spend some time.
The R&D keeps going on...just because we've made a breakthrough, we feel, in predictive modeling doesn't stop the two primary data scientists, working on improvements to the model. And so there's a lot of things we can do in that category. For example, when we make a prediction…at the top of let's say an 8AM prediction in the S&P500...what we're really doing is looking at a decision forest. So you're familiar with a decision tree; a decision forest is many trees in a forest. So there's five hundred trees that ultimately make both a direction and a magnitude prediction for that asset. There may be a lot of information in the five hundred tree voting. In other words, right now we're keeping fairly constant bet sizing, fairly constant position sizing. But there may be information that we have in the distribution of those predictions from the forest that is exploitable. So I think if you asked the science team: A. They're thrilled with how far we've come; B. They're thrilled with who is surrounded them right now; and C. They're excited to keep making progress.
Interesting. Now the next topic I wanted to spend just a little bit of time on is something I think is quite important actually, and certainly to investors. Because that's really the starting point, and that's the track record. People look at it to get a feel for the manager, and to gather a level of interest so to speak. But here's my intellectual challenge when I think about what you do...because of course we know that track records evolve over time, and models change and are not constant. But, someone who starts to do trend following is obviously not going to change completely if they still call them a trend follower twenty years later. So there is some kind of consistency in maybe the way they approach the markets. But in your case we're talking about predictions being made by a machine. I think it can be more challenging to get the comfort of consistency. Because a decision at eight o' clock in the morning is based on one thing, but the same decision made at two o' clock the next day is going to be slightly different. So how do you best explain to an investor how they should read your track record and the likelihood of it being able to be repeated? I know we've talked about consistency…54%..and all of these things. But in more visual terms, maybe you could explain and how investors should look at your track record?
That's very well articulated, Niels. So we actually...you know I like the way you framed that question because you sort of framed it as a challenge to us, when we've been thinking of it as a unique advantage. So let me explain that to you. When I talk about having, in our view, shown predictive power over twenty-seven hundred trades. By that I mean, if you try to get 54% heads in flipping a metaphorical coin over twenty-seven hundred coin flips, there are just a whole bunch of zeros before the first integer. If you don't have predictive power, it's almost impossible to do that by luck.
So our view is, you know, sort of please agree with us that we've shown predictive power to date. And if somebody will come that far..they cannot be able to in the first instance..but if somebody will come that far, then the question becomes: are you going to be able to sustain that power? So, we look at the consistency you've talked about, we look at it as an advantage of ours. We retrain the model every single day. So the model is using more and more information as it's training data the more we trade. When we first launched our live trading account, in January of 2013, we only had training data from 1996 to January of 2013. But next month we're going to have training data from 1996 to January 2015. So as your training data set expands, your algorithm, your predictive model, has a chance to learn more relationships and more correlations. And as it learns, it may even get better.
About that, so the 54% could essentially get higher?
It could, but we think there's a limit on where it can go. And by that I mean this...so what we haven't talked about so far is that feel of what we're doing in the marketplace at 8AM…what are we really picking up? And one of the things Gary concluded early on was that there wasn't going to be any obvious patterns in financial data. The more obvious they get, they more they're going to be traded away by the hundreds of thousands of smart people trading these markets. So we're only going to find those patterns that are very, very subtle. They're subtle, but their non-random. So what we find is on any given trade..any given prediction, we have an enormously high confidence, 99.99% statistical confidence that the pattern we're seeing is going to repeat itself a slight majority of times.
So we have a very high confidence of a very slight repetitive pattern. And so when we find patterns it's because we're fifty-four or fifty-five percent...we're very sure it's going to repeat itself fifty-four or fifty-five percent of the time. And that's a kind of under the radar assessment that's going on at every trade, and that's part of the reason why we feel it's sustainable..is that there's this evolution going on, there's this retraining going on. They call it evolutionary computing. And so as opposed to being a static model that we know has worked in our back test period and has worked for two years, but committing not to changing it going forward at the risk of style drift or whatever, we're not saying that. We're saying we are going to evolve. We're going to evolve with the changing correlations in the market place, with the changing participants in the marketplace. But the one thing that won't change is that we always find subtle but non-random patterns.
Here's a question. I mean you said that you have data back to 1996 and every day Krystal gets another set of data and it learns a little bit on a daily basis. But data since 1996, to some extent, you could say that it's quite a short period of time, really...I mean financial markets have been around for a long time. And part of that period has also seen this unprecedented period...I would say, with a lot of intervention from central banks. The environment that Krystal operates in...what's the optimal environment, if you can talk about that so to speak? Or can you foresee an environment in which it becomes really difficult for Krystal? Because it may be an environment that it hasn't really been exposed to before. Do you know where I'm going with this?
Yes, definitely. So there's a lot in there. Let me pick up on the optimal environment, and also some of the challenges. So let's talk about the times when Krystal hasn't performed as well, because that's helpful for everybody. The optimal environment, if you look at...when we talk about 54%, it's really only meaningful because it's made up of two things. It's made up of trade accuracy, meaning simply, how many trades out of a hundred are we winning any money on? And then secondly, the win multiplier. So how much money do we win when we win versus how much we lose when we lose? Those are the two parts that get amalgamated into the 54%. Because of course, fifty-four by itself is meaningless. There's lots of great trend followers that win only 30% of the time, but they win so many multiples when they win that it makes a very profitable business.
So when you look at those two elements for us, to break it down, we only win about 50% of our trades. It's just slightly over 50%. But when we win, we consistently win 1.13 times what we lose. So, why do I raise that in the context of an optimal environment? Because we feel like volatility is good for the P&L. I need to disclaim immediately that there isn't enough evidence yet in our trading to statistically conclude this. I'm suggesting that volatility is a good thing, but I want to claim first that there isn't a statistical significance to this. Why there's a fundamental reason that would be the case is this..if you think about a low volatility time when you're winning half your trades, and you win $1.13 when you win, and you lose $1.00 when you lose, you have that same relationship of 1.13. But then if it comes a very volatile time, and when you win, you win, say $13 versus when you lose, you lose $10, you still have the same win multiplier 1.13. But your net win on an average trade is now $3 instead of thirty cents. So there's some evidence...there's some intuitive feeling that volatility will be a good thing.
The second piece is that as you look back our out of sample test, it did include 2008 and 2009. And those were fantastic years for the model. Now let's look at the challenging years. The biggest challenging moment we had in our out of sample test was actually 2010. Not something I expected going into this, in fact, I was encouraging the science team not to back test through 2008 and 2009 because it was so abnormal. Well, they don't look at the world that way, and if it works it works. So we tested through that period. What we found in late 2009 and 2010 was that was the least amount of predictive power in the model. And so we've speculated to why that's the case. And I think it goes something like this...there was a fundamental shift in the markets.
If you think about March 2009 when the equity markets bottomed and the change of players in the financial markets at that time...we had the United States government, was participating in a way that it had never done so far. So short selling was banned. And all these fundamental things changed. So then when you think of the predictor, looking back over that time frame, it's looking for relationships that have now disappeared. So it took a while…the decline was three months, and the rebound was seven months. In terms of being underwater, there was a ten-month period while Krystal sort of got its correlation feet underneath it again. But I actually quite like it because it's as if there was a hundred year flood type test and it only took that long for it to find those relationships again and begin to make a fair bit of money. So I don't think I answered every part, you had a lot of stuff in there.
No, that's fine that's fine. There's no right or wrong here. I have another challenge for you Dave. You know, when you go in, and you talk to investors now a days, they always want to put you in a bucket. So, you're a trend follower, or you are a counter trend manager, or you are a whatever they call themselves now-a-days. Where do you fit in, in all of this?
I'm now going to try and work with the questioner and say...how about this? You help us...put us in a category because clearly trend following doesn't fit, momentum doesn't fit; counter trend doesn't fit. In order for the predictive model to have those attributes, we think, you have to give it knowledge of the domains. We haven't talked about this yet; we use non-parametric modeling. It simply means we have no prior knowledge about the dataset. The only information the predictor has is price data. Two times a day price data for the target assets and all the independent variables we give it. That's all it knows. It doesn't know, for example, that there is a normal distribution. It doesn't know, for example, that a price point is some set of standard deviations away from the mean. It doesn't know that there's a mean to which it should revert. So if it doesn't know all of those things then low and behold it's probably not going to exhibit the characteristics of those others.
The other thing that I would say to the person wanting to categorize us in a bucket is to say that our correlation to the SMP500 is quite low, point zero nine, and that shouldn't surprise anybody. But our correlation for example to the new edge CTA index is even less, .06. Now, those correlation numbers are fraught with challenges, I know! But just directionally it certainly doesn't look like we have a high correlation to any other bucket.
Maybe the solution here Dave, is you need to invent your own. You know how it is, if you want to be first in that category and you know you can't be, then you invent your own category and then you are first.
Yeah, well I'm fighting...we have enough experience around our firm to fight hubris at every turn. So if you ever feel like I'm exhibiting hubris, please put me in my place. So I think you make a good point, it's very difficult to categorize us. I would say though that maybe what I should respond with, in the future is to say that..why don't we together go find entities, groups that are claiming predictive power over time series financial data? And we'll happily be in that bucket, and we can distinguish from among those folks. I think it's going to be a limited number of people that are put in that bucket.
But speaking of that, and that I think is quite valid and quite interesting. If you were to mention a couple of names...you know, people like QIM, The Medallion Fund, Jim Simon's Fund...are those the kind of people you would look at?
Yes, for sure. So QIM is a great example because...
Ready to learn more about the world's top traders? Go to TOPTRADERSUNPLUGGED.COM and sign up to receive the full transcripts of the first 10 episodes of the show and visit the show notes where you can find useful links to other amazing resources. Thanks for listening and we'll see you on the next episode of Top Traders Unplugged.