“The best way to manage risk is to spend less money.” – Dave Sanderson (Tweet)
Welcome to Part 2 of our conversation with Dave Sanderson. In this episode we explore his trading program in detail, from the markets the firm trades to how they describe their program to investors. We also explore the challenges that he goes through as a business leader, dealing with drawdowns, and why optimism is so important to Dave.
Thanks for listening and enjoy the second part of our conversation with Dave Sanderson.
In This Episode, You’ll Learn:
- Why Dave believes that QIM is a unique firm in a reasonably similar category as KFL Capital Management.
- How to describe Krystal to investors and why it can be a challenge.
- About the length of time it takes for Krystal to compute the data and make a decision.
“It used to take six weeks to run that [data computation]. Now it takes about a minute.” – Dave Sanderson (Tweet)
- Why the breakthroughs in computational power are supporting KFL Capital Management to make their systems faster each year.
“The mathematical space is so big, it’s like searching the size of the Internet every time we make a prediction.” – Dave Sanderson (Tweet)
- The markets KFL trades.
- Expected drawdown and volatility Krystal expects.
“What gives me comfort is the fact that we’re agnostic in terms of direction.” – Dave Sanderson (Tweet)
- How Dave expects to deal with drawdown environments.
- Research cycles within KFL Capital Management and the potential for a second Krystal.
“Research is non-linear.” – Dave Sanderson (Tweet)
- About the 99.3% match rate between their live trading and back testing results.
- The biggest challenge for KFL Capital Management in today’s market.
- About the challenge of attracting AUM in the modern financial landscape.
- Asymmetry of agency and understanding how to focus on who you’re talking to.
- Regarding the difficulties of explaining machine learning and big data ideas.
- Commonalities in the highest level due diligence explorers.
- Entrepreneurial perspective, great books and an open minded perspective on failures.
- Why optimism is such a powerful force in today’s world.
Resources & Links Mentioned in this Episode:
- The Medallion Fund – Jim Simon’s Fund.
- The Innovators by Water Isaacson.
- Zero to One by Peter Thiel.
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
“I think the great experiences go to the optimists because in life you probably find what you’re looking for and the optimists tend to look for great things and expect great things.” – Dave Sanderson (Tweet)
Dave: I think that's what I would pass along: I think this world we live in is quite magical. I think the great deals go to the optimists because they're prone to paying more for them. The great experiences go to the optimists because in life you probably find what you're looking for, and the optimists tend to look for great things and expect great things.
Niels: If someone came to you and said that they had found a way to predict the future with more than 50% accuracy would you believe them? Would you be intrigued to learn more? What would it take for you to be convinced that this could indeed be done? Big questions in a world of big data, but in reality the answer should be, "tell me more." Since just because no other investment firm has claimed victory when it comes to predictive power using machine intelligence to successfully trade the financial markets, we can't really exclude the possibility of it having been done by a team of scientists in Ontario, Canada, and that's what we're talking about in today's episode of Top Traders Unplugged.
Introduction: Welcome back to Top Traders Unplugged. Where the best traders in the world come to share their experiences, their successes, and their failures. Let's rejoin the conversation with your host, veteran hedge fund manager Niels Kaastrup-Larsen. Niels: ... 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 Simons' Fund, are those the kind of people you would look at?
Dave: Yes, for sure. So QIM is a great example because they are still telling their story to the world. They haven't just closed down; they don't need to talk about it. QIM talks about it, they do certainly sound like they're in the same category. Jaffray Woodriff, the principal of that firm, is a self-studied computer science data guy. And he's obviously done very, very well. He also hired, at one point, he spoke with Jack Schwager about hiring John Elder. And Jaffray wrote the preface to John Elder's book. So my science team is very aware of John and the models that he's built and the ensemble models that they wrote about. So we're familiar with that, and that is a firm that is quite reasonable to talk about when you talk about what we're doing. They manage a lot more money than we do, so we have to give them credit for that!
On the Renaissance front, you know, it's obviously very difficult to figure out what they're doing. I think it's really cool to think about Mercer and Brown who have taken over the firm, and where they came from. Back in 1992, Mercer and Brown wrote a paper on language translation. It was quite heretical, it was quite a kerfuffle in that business, in that academic space, because they took the stance of not having any syntax knowledge, and no knowledge about two languages: French and English. It's really close to home, because the data set they fed their computer was the Hansford texts from the Canadian legislature. As you may know, we have two official languages in our Country: French and English. So we have these reams and reams of legislature notes that are in both French and English. So Mercer and Brown took those, they poured it into the computer and they said, you tell us what these words mean in English, or you tell us what these words mean in French, and we're not going to tell the computer anything about the languages. They absolutely crushed previous language translation accuracy. So if you look at that now, the one thing that is different though, is that's not time series financial data. If le is the then le is always the. So I don't know, if I was able to sit down with Mercer and Brown, I don't know whether we would be as akin as I hope we would be. But certainly a very interesting firm, and they've always stayed away from the Wall Street folks by hiring physicists and cosmologists and all those things. So that's something we emulate as well.
Niels: I'm curious about one thing before we move on to the trading program itself, and you can explain how it works. And I don't know... I'm not sure exactly how to phrase the question... but if you think back to the time where the feedback coming back to you... you know it can't be done. And then suddenly, Eureka, in 2012, it can be done. What do you think it was that you discovered that allowed you to do this? And obviously you we're not asking for any proprietary details but is there some kind of concept you can visualize to say that this is why we were able to do it. Because it's such a big step, it seems that you've accomplished here.
Dave: Mmm...yeah, you know, I think that most great innovations have a massive dose of serendipity. And I don't think we're any different. Dr. Gary Li is a unique human being, and he was just not letting this go. He will tell you that one night he had a dream, and as he often does, he keeps a notebook by his bed, and he woke up and made some notes. And that was one of the insights that allowed him to make progress. I think the things that really frustrated him became his focus area. So the concept that correlations change over time really informed the way he was going to look at the modeling.
The second thing is the idea that if a pattern was obvious, it wasn't going to be repeating itself. I think that was a big sort of philosophical part of his discovery. But beyond that I think... you know we've often talked about why... one of the questions we got met with early on, before the longer track record. Now, I say longer because it's only a year in the partner's fund that's out in the world. And people say well you're just a year along. We would say, well, a year might be forty-eight trades to a trend follower or somebody else, but we've done twenty-seven hundred. So I think that twenty-seven hundred may be a lifetime for most managers.
But anyway, before we had a long track record, people would say to us: So you've done something that huge firms with legions of Ph.D.'s have been unable to do? They're sort of incredulous. I don't know how to fully respond to that. I would say one thing though, if this project were being incubated at a major Wall Street firm; it certainly wouldn't have been given five years to come to fruition. And there may not have been the level of collaboration supported that is necessary for this kind of breakthrough. That's not directly responsive to your question, but that's all I got.
Niels: Sure that's fair enough. Let's move on to the heart of the strategy, namely the program or Krystal itself. When you go out, and you talk to potential investors, how do you explain Krystal?
Dave: We explain it by trying to be... well first of all, we declare that it's in many ways not explainable. And by that I want to be very careful the way I say that. People say are you a black box? And we say well it's very difficult to answer that question. If you mean that we're 100% reliant on our model to make a trade, then you're right. If you mean we have no idea why a trade gets made, well now I need to be more particular with you, so let me just take you through this: so at eight o' clock in the morning (because we've already used this example) we trade the S&P 500 and the two other equities indices we trade. It may be helpful to talk about what's going on at that moment.
What's going on at that moment is the predictor in terms of input data has the last three sessions of the S&P 500. And it has the last three sessions in terms of price movement of forty-nine other independent variables. And it looks across all those variables, and it's asking a question - have I ever seen THIS before? It then turns around to the historical data set and goes backwards through all that what we call mathematical space, back to 1996, and it looks for subspaces where those relationships have shown themselves before. And if it finds a significant number of them, it says okay, what happens next? And if what happens next is the S&P 500 goes up, then we go long. So that's the process that's happening twice a day in every asset. And that goes part of the way to explain to somebody what we do and how we do it.
But let me just finish the thought about black box, because it means that when we go long in the S&P 500, we can tell you with great specificity, if we tear back the model findings. We can tell you how many times we found those subspaces where this event occurred and why we think it's statistically significant. So we can tell you any day why we're going long with the S&P 500. But we can't tell you about the causality underneath that. We can't tell you why oil, gold, and the ten-year note are doing this seems to have this effect on the S&P 500. I just... I can't do that. I know if you turn on CNN somebody's going to proffer some answer to why those things are happening, but we declare sort of agnosticity there. We just have no idea what the underlying causality is, but we can tell you why we think the pattern is significant.
Niels: It makes a calculation you say, where it goes back many years looking for these patterns. I mean, with all these variables, I'm imagining that it takes quite a lot of computer power in order to do that. But I'm more interested in how long of a time does it take for Krystal to look at all this data and come up with an answer about what to do now?
Dave: Well that's a very interesting part of the story I left out. It used to take six weeks to run that. It now takes less than a minute. It takes about forty seconds. And if you think about the size of mathematical space, let me try to do this some justice, but I may confuse things here. So, those fifty variables, those forty-nine independent variables, and the one target, they're not just one variable of course, they're many, many variables over time. And if you think about a regression analysis let's say, and you have an X and a Y axis, you have a line. If you add a third variable, zed, you've gone from a line to a cube. So you've just multiplied mathematical space enormously. Well, what if you go to fifty? One can't even imagine the size of that space. And what if all those fifty variables were a thousand discrete variables, each depending on how they're acting at that time? So we like to say that the mathematical space is so big, it's like searching the size of the internet every time we make a prediction. Then what we're finding amongst that enormous space are these subspaces of relevance: pockets of time when these relationships exist. So you're actually, when you do it algorithmically you have to have a very, very efficient way of searching an enormous amount of space and coming back with an answer.
Niels: Do you have to use cloud computing to do all of this calculation, or is that something you can actually do in your office so to speak?
Dave: We do it in our own office. The computational power is just enormous these days. We don't have any particularly fancy boxes. The boxes we have, have some number of cores, and anybody can buy them with a reasonable budget actually. That's not where the uniqueness comes from. The root to going from six weeks to forty seconds... there's a lot of people helping with that. One of the presenters at the Battle of the Quants was a guy from Intel. Intel is doing a lot of work on parallelization. So it's the computations happening at the same time as opposed to serially. And that's the way we used to do it, was serially, and that's why it took so long. But as soon as you parallelize and as soon as you take advantage of what quite frankly so many people... that's one of the areas of big data that's helping us... these breakthroughs are never made in single disciplines. You have to have a bunch of things coming together. So storage and access, and expense of data has to come down to where it's come. Computational power has to increase. I think Moor's Law is still happening, so even if you look back at five years ago when we started this project, if computational power is doubling every eighteen months or so then we've had a few doubles since we've started this project.
Niels: If I'm summarizing correctly, you trade twelve different markets and you have... did you say forty-nine independent targets that you check every time, of which I believe a eleven of them are the other markets that you're trading. Can you give examples of what other targets, sort of the non-market targets are? Just for me to really visualize what Krystal is doing in the forty seconds.
Dave: Yes, absolutely. We have a broad representation of the financial markets in that list of independent variables. You can imagine; you put some currencies in there; you put some single stocks in there, some ETF's, some Indices, just a smattering of products all of which have data back to 1996. What we have found is that pouring more independent variables in doesn't help. Once you have a broad representation of what's happening we're able to pick up these subtle patterns.
Niels: Okay, and the twelve markets that you trade today, do they represent both financial markets and commodity markets?
Dave: Yes. We trade the S&P, the Dow, and the Nasdaq, the ten year note, gold, copper, silver, three soys, corn, and oil.
Niels: Okay, so quite a diverse portfolio actually?
Dave: Yeah, once again when Gary asked me originally... the real question was we want to validate this technology, how do we do it? And the answer was let's do a thousand trades or thousands of trades, and let's do it across as many sectors as are reasonably accessible.
Niels: Do you get 54% pretty much on all these different markets? Despite the fact that they are quite different?
Dave: Yeah, very, very close. There's no... in over a large sample size; there're no really dramatic outliers. I think our lowest is 47%, in terms of what we call the coin toss ratio. That's the combination of both win/loss and win multiplier. And the highest is 64%. But again, that's live trading over a reasonably short period of time. But it's amazing how that tightens up just the more trades you have, the more tight the dispersion is between the coin toss ratio of all assets.
Niels: Sure, you mentioned earlier on that your average trade length, I think you said something around thirty hours or so, but you also talk about doing something twice a day which obviously is less than thirty hours. Can you explain about how that works from when Krystal gets a resolve from all the voting going on, and it decides to do whatever it decides to do, what happens next with this decision?
Dave: Right, so the two things that are happening twice a day are a prediction. So we will always make a prediction. It's just that a prediction may not result in a trade. So if we make a prediction at eight in the morning and the S&P is predicted to go up, we don't touch that trade until two in the afternoon. We don't have a profit stop, stop loss, nothing. We have a timed new prediction. If the new prediction is consistent, i.e. the S&P is going up again, then we just remain in the trade. If the new prediction is the other direction then, we short two units in order to get short one unit. And if the new prediction is not sure, then we flatten the position. So you can imagine you're always going to have, by definition, more predictions than you're going to have trades.
Niels: Absolutely, very interesting indeed. Let's jump to a related topic which is the risk management. You mentioned a little bit about it just now in terms of you take on the trade, and you don't touch it, so clearly you're relying on signals to change in order to change your position. Tell me about how you look at risk and how you define the risk framework that you let Krystal operate inside.
Dave: Right, so our definition of risk is: the best way to manage risk is to spend less money. Our average margin to equity is 15%, which means we're 85% in cash on average. That, in our view, is the best way to "manage risk." Managed risk to us means you've got this wonderful technology; you know or your convinced that over time you're going to get this accuracy result of 54%; so going back to the roulette table metaphor, the only way you can really mess this up is to bet too much, or allow a customer to bet too much. So if a customer was able to come in and put a million dollars on red, he could put you out of business. So that's what we're managing. We're managing the bet size, and we have found that's the way to keep the real key metrics most positive.
So we like the Sortino ratio, obviously we're looking at downside risk per unit of return. What we have found is anytime we try to use other kinds of trade management: trade management that is something beyond just spending less money, we take risk-return attributes out of the results. So the Sortino goes down, the sharp goes down, the things people care about when they watch a portfolio every day get less attractive if we try to build in some of those more normal risk management characteristics. We run into a messaging problem around that because people will say to us, well it's going to be very hard for us to sell something, or to give it to our clients that don’t have some language in it that they're normally comfortable with. It's a challenge. We want to remain pure in terms of what we believe, which is we're delivering the best unit of return per unit of downside risk, and yet we want to characterize it in a way that helps people sell it.
Niels: If someone came to you and asked, what should I expect in terms of volatility, let's just measure it as standard deviation on an annual basis, and then my next line of question which is about drawdown; from all you've seen in terms of Krystal's ability, what kind of volatility do you expect it to produce and what kind of drawdowns would you expect from Krystal?
Dave: So we have a consistent relationship among those things. Given that we feel we can deliver this 54%, the question for each client becomes: what do they want to spend in terms of drawdown, or what do they want to spend in terms of standard deviation, what do they want to achieve in terms of return? So the relationship that we have found is that there's 1% of annual standard deviation, per 1% max drawdown, per 2% net annual returns. It's a 1/1/2 relationship. If we can continue to deliver that. It's been slightly better than that in live trading. Our net return at the end of the first year is 30%, and the max drawdown on a monthly basis is only 5, but peak to trough it's 9. So the 9 to 30, that's more than we are promising, but we expect that it will level out to a 1/1/2 scenario going forward.
Niels: So with that, just to clarify, the 1/1/2 means a 15% annual standard deviation should produce a 30% return with a 15% max drawdown, which is of course, if you can deliver that you probably won't find a shortage of investors who want to part with their money, so send them to Ontario.
Dave: The question is when, when are they going to believe?
Niels: Yes, that is true. Now in one sense you're obviously extremely experienced, and on the other sense you can say that you're still relatively new to this in terms of managing the money real-time and so on and so forth. So what I tend to do at this stage is I tend to try and ask the managers that I've spoken to who may have been around for 20, 30 years and how they deal with drawdowns? The problem with you is that you haven't really had any drawdowns, and even the drawdown you've had is not really severe, so have you thought about this? I asked it because my last conversation, my last guest actually, they're in a little bit similar situation like you, relatively new, have a very interesting strategy. They had been trading their own money for a while, and they actually found that taking on client assets, particularly in the drawdown periods put an unexpected pressure on them. I'm just thinking here. You're dealing with a model; you're dealing with Krystal and you're not entirely sure what Krystal will do other than it should produce 54% accuracy. But have you thought about how you will deal with drawdowns? Because we can almost be sure that they will come, because that's part of your statistical set. You know they're going to come. Have you thought about that? In particular, if it starts becoming a little bit more than what you expected.
Dave: Yeah, so that's a great area to talk about. You can have lots of plans for it but until you're sharing your experience with a drawdown it's something apart, I think. There are a couple of things that we are getting ready for. So first of all we agree with you, they're coming, those drawdowns are coming. This is not a business for the faint of heart. That's why the spoils are so big, and so we're going to earn our keep during those times. One of the things that we're going to do is we watch very carefully a comparison to the statistical record of the out of sample test and of the trading thus far. If we see something in live markets that is aberrant, that we cannot find in the long out of sample test and in the live trading to date, then we will flag that and say that's an indication that something has changed.
Let me give you a more specific example. There is no occurrence of three months of negative returns in our out of sample test; if that occurs, we will reassess. So that's an example where if we find something that's occurring in live trading that has never occurred before, then we need to reassess what's going on. We either need to explain it or turn it off, and that's something that's not been done in our industry. It's almost like having a service level agreement. So if you think about it we could have a service level agreement with a client that says, look, if any one of these things happens we will phone you. As another example, we can track the 54%, and we can do a rolling average of predictive power. If that rolling average of predictive power was to fall below a previous level, a previous minimum, then that's something we should tell our clients. So we're trying to do things a little bit differently, and nobody's taken us up on that yet, but I think they probably will.
Niels: It's really interesting to me what you're saying here. I think, in some ways, it makes perfect sense what you're saying that, OK if we haven't seen this before we call our clients or we switch it off. But I want to bring something to you to think about and that is, over the last decades you and I and many others have watched trend followers, those classical trend followers, we've also watched them struggle in the last few years with drawdowns that became bigger than usual, longer than usual, but what is really striking is that as you and I speak, on November 30, 2014, many of these great firms in my opinion are hitting all-time highs.
They're putting back some extraordinarily strong numbers for 2014. To me it actually makes what you just said really difficult because they did not see the longevity of the recent drawdown. Many of them did not see anything like this in terms of depth, but still their models have adapted, and when the market environment was ripe, they've been coming roaring back. The worst thing you could have done at the time of the valley of this drawdown was to do anything other than keep trading. So I think it puts a big responsibility on the managers and in a sense you as well, to have an opinion about these things and say, OK but even if we get four months of drawdowns, we may not have seen this before, but it might not be a problem. So it's that balance between picking up something unusual, and then deciding to do anything or not to do anything. Clearly the classical trend followers take the approach saying, well actually we're not trying to predict anything, so why should we try and predict exactly what kind of drawdown and exactly how long it's going to be, but we know in the long run these things work. So we keep trading as long as nothing is fundamentally wrong. Again, the death of trend following has been written so many times, yet again, it's just proven that that's a little bit too early to write it off. You know what I mean?
Dave: I know what you mean, and it's well said, and shockingly, I'm going to tell you we're saying the same thing. The reason I'm going to say that is this, the big distinction between us and the great trend followers you're talking about. So I admire those trend followers. I admire them even more when their death is prognosticated, and it was premature and so staying with it during those times; absolutely that's the right thing for a trend follower to do. The reason that we have a different decision to make is because we're looking at an experience of thousands and thousands of trades. If our out of sample test is 10,000 trades, and our live test is 2,700 trades, and something changes amongst that data set, we have an obligation that a trend follower I don't think has. If you were able to go back 13,700 trades in a trend following system and your prehistoric in order to do that, you might say the same thing as we're saying. I love the question. I think you're absolutely right about the great trend followers and anybody who says it's not working, I mean it's been working for 30 years, so you ought to stay with it during those drawdowns. I think our drawdown metrics have a different statistical substance to them, and we have the obligation to treat that statistical substance differently.
Niels: Well put. Speaking of risk, speaking of drawdowns, just as sort of a general question, is there anything that keeps you awake at night when it comes to risk? Something where you feel if that happened I would be a little bit worried about Krystal. I have to this word Krystal and your name, it's great, but it keeps reminding me about this series Dallas with Bobby and JR, wasn't she called Krystal?
Dave: (laugh) I don't know. I was too young.
Niels: You're too young. (laugh)
Dave: Maybe I'm talking about the first version of Dallas and not the second.
Niels: Anyways, back to the question, is there anything that keeps you awake that you think Krystal might not be able to handle so well?
Dave: What is on my mind these days are business execution issues as opposed to technology issues. I'm excited to see what the technology will do during times that we think are several standard deviations, if there is such a thing. So we have looked back at some anomalous times and once again, I have to declare that there's not a statistical significance to this narrative I'm about to give you. For example, during the Boston bombing, gold went down $60 and all of the mayhem of that week. We actually made 5% that week. We didn't make money in gold. We actually lost a little bit of money in gold. We looked at the Flash Crash; we looked at many days in 2008 and 2009. It certainly cannot be the case; you can never say in this business that I'm worried of nothing. I suppose I'm as worried as I am excited about the next really crazy event.
Oil's gone through some really interesting times in the last couple of weeks. We'll often be short oil; we'll often be long oil during these times. When you break it down into a micro component like that, where it's two session a day, you get a very different result. So just picking up on that Boston bombing example, I was asked the question, what happened when gold went down $60 overnight, and I thought to myself, gee, is that what happened? The truth is that's not what happened. It went down $70 over the course of four days. So there's eight trading sessions in those four days and eight different predictions from our model during those days. What gives me comfort is the fact that we're agnostic as to direction. So in a crazy period of time, we're going to be long some assets, short some assets, and we're only going to be holding, or making predictions on average every twelve hours. So, again, I'm not suggesting at all that there's nothing in terms of market behavior that won't have me turning on the machine saying, gee I wonder what happened? But if I were allocating money to strategies, I would say I'm more comfortable trading a multi-asset, multi-directional, short time frame strategy than I am say building a large position in one particular asset.
Niels: Sure, sure, let's shift gears a bit and move on to another very important topic which is research. In a sense, you are very research focused. In fact, research was the only thing you did for a while before you started trading, so that certainly is a very topical area for you to talk about. I wanted to ask you what kind of research are you doing today? You did the initial research to see if you could prove that it couldn't be done, but you realized it could be done, so what kind of research do you do today for Krystal to evolve?
Dave: Right, well that's a great area for us and what I love to say at the beginning of discussing this is that research is nonlinear. By that I mean, our team can go a very long time without expressing any modifications to Krystal.
Niels: Sorry to interrupt, Dave, but Krystal is learning by itself. Is Krystal doing the research for you?
Dave: (laugh) Well I wish I could look at my budget and declare that, but I can't. So yes, there's evolutionary computing going on. Really I'm not so sure as it's doing research as it's paying attention to the fact that the correlations that we talked about change over time. So it's finding these new correlations. So in a way it's really not doing anything dramatically different, it's just reacting to different relationships in its data set. But thankfully it's doing that on an on-going basis, so that we don't try to find one inefficiency in the market and just build a static model and let that trade until that inefficiency is gone and therefore the P&L goes away from that strategy. So I can't go that far with you, but what research does go on?
One of the things that we're close to now is coming up with a second version of Krystal. Right now, we've moved it along to the point where our accuracy is 53%. So it's not as good as the one that's in the market today, but the interesting attribute about this model is that it is not very correlated with the model we have in the market now. So it may be that if we launch a 54% with a 53% we can have a risk-return result that's even better than just the 54% alone. So that's the kind of thing that's going on.
Gary Li is the driver of that research, and he is somebody that feels like we have not accomplished... well he would say we've accomplished 10% of what we're capable of over a lifetime. So that's the kind or rigor, or pressure he puts on himself to come up with new ideas. He's always reading. Certainly deep knowledge and deep learning is something that a lot of academic folks are talking about these days, and how that applies to financial data, if it does at all. Certainly the folks that are working on Google Mind have some interesting ways of attacking this problem. If you think about one of the challenges of that technology, it's very easy for a three-year-old to recognize a cat in an image. It's very difficult for a computer to recognize a cat in an image. So Google is making some progress there. So we certainly... and actually some of the people that were hired to do that were out of the University of Toronto, and familiar to our folks. So he's watching all the time what's happening in various pods of excellence and applying it to what we do to seeing if there's any way we can keep improving our technology.
Niels: Now, I get a question from our listeners very often who are very interested in me remembering to ask this question, so I'll try and rephrase it to today's conversation, so let's see how I'm doing on that. Traditionally when you have a... and again I use a standard CTA strategy as a starting point. Let's say that you start losing money, and you track it down to a specific model. There's just a model that's not producing any return, in fact, it's losing money. So you can identify what the problem is and then you can decide from there what to do about it. But inside Krystal, if you start losing money and the 54% generally goes down, how do you located where the problem is, because Krystal is looking at all sorts of things every single time it makes a decision, so how would you even figure out why the 54% was not holding up should it happen?
Dave: That is a great question. I can tell you we faced it, not fully, but we faced it in this sense: if you look at our trading to date in what we call the KFL partners fund, we have a loss in the silver contracts of about 11% of our total profits is a loss in silver. There are 99 trades in that asset course of the life of the fund. So we did field the question that was something like, well why trade silver? The answer is because over a very, very lengthy out of sample test it proved to be profitable. So I don't know at what point, in terms of the number of trades, we would ever throw out an asset, but it would go to the core I talked earlier about, the two attributes of an asset that mean we can't trade it. It has to do with data, and it has to do with spread. If those two things are not the problem with, for example, silver, and we can figure that out, then there's something in the core of the technology that is the problem. That's an interesting conversation that I hope never comes up, but I don't have an easy answer for it.
Niels: Another question I think, again I'm trying to rephrase it in my mind as we go along because you do things differently from so many of the people that I've had the pleasure and privilege to talk to. It's the point about the backtest. Sometimes I would argue that looking at a traditional CTA that's been around for twenty years, you would argue that a backtest is probably more meaningful to look at if you could get a backtest of the current configuration rather than the live track record because it has changed so much. But in your case, I wonder, if you could have a choice to look at a live track record going back five years of Krystal, or a backtest of Krystal going back twenty years, which one do you think would be more meaningful? Or would they be the same in terms of value to you looking at it?
Dave: Hmm... let me answer this two ways. First I will tell you that Gary Li takes 80% of his comfort in the veracity of our model and the accuracy of our model. From the mathematical proof, he takes 10% from the backtest and 10% from the live record. So that might put it in context. Now I'm not as evolved as he is, and I would answer the question this way. We have made every effort to ensure that our backtest is what really would have happened. So when we launched in January of 2013, you can imagine it was impossible to go back over the previous five years, and trade live because you can't go backwards in time. So we could only launch with the benefit of the out of sample backtest. But as we've traded live, you do have the ability to do both live trading and backtest trading over that period.
We've worked very hard to the point where our live trade record for the next six months, or let's look at the last six months because it's accurate, so the live track record over the last six months is 99.3% the same as the backtest over that same period. So that's something that we track on a real time basis, even though we're able to trade live in a period, we backtest that same period. Our question is this, not what's the P&L, although we do have that question; we do have that measurement, but are the trades... S&P up November 30th, 8AM, did we go long? Did we go long in the backtest? Did we go long in the live trade? What we have found is we've got it to the point where we have a 99.3% match rate, which we're very thrilled with because that gives us some comfort to say, OK, if that's what's happening then the five year backtest, well, we're 99.3% sure that's what would happen.
Niels: You described, I think, earlier on that what you're trying to do is to predict human behavior to some extent expressed through financial markets. Now some of the guests that I've had on, and I'm thinking here of Roy Niederhoffer recently, and Scott Foster, previously, and I'm sure there's a few more of them actually. As far as I recall, one of their findings is that, generally speaking, that human behavior is more predictable through stressful times. Do you see that as well in Krystal? Is it more accurate making predictions when markets become in a sort of state of stress?
Dave: First of all I would agree with the conclusion that behavior is more predictable during stressful times. We also, in order to substantiate that from evidentiary point of view, we certainly in the out of sample test made more money during what we would call stressful times then less stressful times. What I don't have for you, and I'm a bit ashamed to tell you because I can actually discover it and I will as a result of this conversation, I could track our predictive capability through what is loosely labeled stressful periods versus non-stressful periods. We talked earlier in this conversation about the win multiplier getting better during those times, but I suspect the trade accuracy might also get better during those times, but I don't have the number at my fingertips.
Niels: That's fine. What's the biggest challenge that you feel that you face as a firm right now, when you look at it?
Dave: I think it's very clear, convincing people that they ought to allocate to an emerging manager. I believe that allocations have been made by-in-large to great infrastructures. There was a period of time, and you're familiar with it, when small assets were not really an indicia of anything except being nimble. Now small asset bases carry with them the fear that there isn't a robust infrastructure and that you're not going to ultimately get what you buy. So that's, I believe, our biggest challenge is telling our story in a way that's differentiated, but also helping people get over the chasm of allocating to somebody that is not easily categorized, and that is not currently providing the comfort that comes from managing hundreds of millions of dollars.
Niels: I agree on the challenge, of course. Have you made any progress, or maybe that's not the right word, but have you come to any conclusion as to how you think you do that?
Dave: Sure, pure force of will (laugh). No, I think you tell your story passionately as much as you can, and I appreciate you taking the time to help us do it through this medium. We have made progress. We have really our first account that has come through the very traditional due diligence process by one of the largest FCMs and they put in front of us a client that they have a high regard for and who has a very large asset base, so that account has been one we will fund it this month, and so hopefully, I think under the broad category of assets beget assets, maybe we're on the way to doing that. I hope we are.
Niels: But interestingly, if we can just spend a couple of more minutes on this, because I think it's so helpful for the majority of managers listening to this, because the majority is small relative to the minority who are very big. What kind of pushback... When you explained this story to me, and you explained the caliber of the people that you have working with you. You explain the computer power you have in order to do these things. In one sense it's pretty difficult to say that you don't have a good infrastructure, you don't have talented people, so what kind of pushback are people actually giving you? Or is it, in truth, pushback that comes from a different angle but is maybe not expressed in a direct form to you, saying OK, your infrastructure's not big enough, but that's not really why they might say no, or they might say it's too early? Do you have a sense for what the real issue is?
Dave: We talk about this all the time. It's so fascinating to think of the psychology of these meetings, so there're a few things that have come up that I think are reasonable assessments to make. One is there's an asymmetry of agency. By that I mean if you're talking to somebody who's an agent of money, whether it's an agent of the family office, or whether it's an employee of a large firm and that employee is tasked with doing due diligence on managers, the asymmetry of saying yes or saying no is this, if they say yes to us, meaning they recommend that allocation to us or any emerging manager, then they might get fired. If they say no to us, there's no chance of them getting fired. In fact, there's more chance of them deemed to be prudent. So guess what happens when you have an asymmetrical payoff at a very personal level? You saw what happened in 2008 and 2009 on the opposite side of life: make the bet, hit your number, get paid a lot, don't hit your number, nothing happened.
So there's a bit of an asymmetry of agency there. So you have to be careful who you're talking to, I would say to the managers listening. If you're talking to a decision maker, then you're talking to the right person. The decision makers in front of an emerging manager tend to be either high net worth individuals or family offices that are very close to the money. If you're somewhat disassociated with the decision maker because you're talking to a larger institution and asset allocator or pension fund, then just realize that you're probably... it's inappropriate to conduct that interview in a way that forces somebody into a yes/no decision. So what I think the mistake I made early on was talking passionately about this scientific breakthrough and really what the person felt was they were being asked to say, do you believe or don't you believe?
Of course, it's very prudent and very reasonable for somebody to say; I'm sorry, at the end of a 45-minute meeting I can't declare that I believe. So we stopped that, and we started saying, look we're not asking you to believe or not believe. We're asking you a very simple question, would you like to follow our story? If they say yes to that, if they say no to that then they're just I think not intellectually curious enough and it's not harm done, but if they say yes to that, then we put them on the monthly information update and we watch how they metabolize that information. So if somebody is metabolizing that monthly information, meaning they're going in, they're opening up the emails, they're looking at... we did a series of videos that we're very fond of. So if they're watching the videos and reading our material then they are metabolizing that information and at some point we earn the right to discuss with them again. Really the question at that point is not yes or no, but it's what do you think? So we've taken this very... we've changed our approach to that conversation.
Niels: You know, Dave, that's a great answer. I have never heard that explained that way, but I have to commend you for that because what you're saying, and I think this is so critical, and that is we shouldn't really market ourselves as we did twenty years ago, or twenty-five years ago. What you're saying is we need to do it in a way where we can monitor the dialogue and we can slowly make it more and more meaningful to a point where you are able to have the conversation about do we fit in, and so on and so forth. I think that's really interesting, and I think most managers today don't necessarily use that kind of technology, which is another kind of technology. You mentioned the simplest thing which is monitor do they open my emails, monitor do they actually pay attention to the material we give them via video or something else. That's a scientific approach to marketing, which I actually completely agree with, but I don't think many people are doing it today. So I love that answer, actually.
Dave: Thank you very much. It's born out of some frustration or some reflection on the early part of our marketing and sales efforts. Once again, I have to thank Meyler Capital for really giving me those ideas and changing the pace at which we thought we would bring on new folks.
Niels: Sure. As a still emerging manager as you say, if you could ask a question of a big manager, someone who came onto my podcast and 20, 30 years of experience, is there anything that you would like to ask them? I'm just curious what you think you could really learn from someone who has been doing it for a long time?
Dave: Yeah, I don't think there's been a meeting I've been in where I haven't learned something. I hope that always continues. I think this industry attracts really great people, really intellectually curious people. I was just reading some material from Northfield; I don't know how to pronounce the guy's name. Wonderfully smart people who write about our business and who opine on our business, so there would be something I can definitely learn from everybody. I think my questions today would be in the category of asset gathering. I would just want to know what they felt they did right, what they felt they would improve on over the course of time? I think probably from a philosophical standpoint I would also ask them have they enjoyed the journey, because I'm surrounded by partners who have had big economic payoffs and they assure me that it is not the destination, but the journey that matters. I can tell you we've had an enormous amount of fun and enjoyment in this journey even though we aren't yet at the destination, meaning managing a billion dollars.
Niels: Sure. Now that's sort of the question about... in talking to another manager, but when we talk about potential investors, what do you think they find the hardest to understand or get their head around when it comes to machine learning, a strategy that is based on that aspect rather than sort of the traditional research, historical data, come up with some rules, test them thoroughly and implement them if successful? What's the difficult, do you think, part that investors face when they come and talk to you about machine learning?
Dave: I think the first difficulty they face is the language we use. So when I try to figure out what I think about something, I'll write about it. So I wrote a twenty-page document recently on this very issue: what is the challenge of the recipient of all of this information? I ended up calling the document "Euphemisms, Metaphors, and Loose Associations in a Particle Physics World" (laugh).
Niels: A nice, easy to digest title (laugh).
Dave: So the point is this; we try to make things sound bites for everybody. So we'll use the sound bite of big data, or we'll use it of machine learning. What that person hears when we say machine learning, we have no idea. So I think it really serves all of our purposes to help the recipient really delineate the nuances and to really categorize things quite quickly. We started this conversation with this distinction between reacting to prices versus predicting prices. I will take that into all of our discussions just right up front, to say there is a distinction in the world of people you've probably seen to date; and I want to help you with that distinction. The great job I can do of helping them understand where everybody fits in, the more chance we have of them, again, metabolizing a very tough story to metabolize.
Everybody brings their own experiences to the meeting, and if you think about what we're saying to people, is we're saying here is an innovation of enormous proportions. It's just fields, metal, type stuff here and yet we're sitting in front of you, Mr. Investor, encouraging you to invest in us. They’re probably thinking one of two things: one - if this is... it's so good it's too good to be true, and secondly - if it is that good, what are they doing talking to me? If it is that good surely they would have two billion dollars to manage. So I think... maybe that's an analysis of a less sophisticated listener than the one you were thinking of, but those are the things that are on my mind.
Niels: It's fine. When you do meet, and this is going to be my last question before we jump to the final section, but just staying on that point. In these due diligence meetings is there something you feel investors today should be focusing on more, and asking you about compared to what they do? Is there something where they should focus their attention when it comes to talking to you?
Dave: Yes, I feel like the best due diligence conversation we have are with the most sophisticated people and the people who have had the most experience with statistical modeling and backtesting and the pitfalls of backtesting. You can tell pretty quickly when you're talking to someone whether their experience is such that they're asking you these really crisp questions about the pitfalls of backtesting: things like selection bias, and things like sampling bias, and things like symmetry of live data and historical data and the challenges therein. If somebody is in that track, immediately I know the information we give them will be encouraging to them.
The tougher challenge is when they haven't asked questions about the quality of the testing that has gone on at each firm they've talked to. So if those challenging due diligence questions aren't put to the various managers they're interviewing, and then they allocate capital to them, and then the backtest doesn't turn out to be reality, they come with that experience to our meeting and we want to say, well here's our out of sample test, and here's our live track record. They want to say, well, gee I can't put any emphasis on out of sample tests, because I've never had a backtest that's worked, and secondly your track record is too short. Meaning they don't give any weight to the number of trades versus the number of months. To respond to your question, I would encourage them to do a stronger level of due diligence on the statistical underpinnings of the information they're being shown.
Niels: Let's jump to the last section which I, I just call it general and fun, so it's a little bit of everything. You've clearly had to learn quite a few new terms from being a commercial litigator to now talking about statistical measures in a very fluent way, but what about the entrepreneurial gene, has that always been inside of you?
Dave: (laugh) Yes it has. It's a genetic predisposition in my family. My father was a serial entrepreneur, so I think I was masquerading and happily so, as a commercial litigator because it was a very interesting time. The one thing I took away from that, I think, was the confidence that I could be at any meeting and contribute some decent value. The entrepreneurial gene has always been there, and I can tell you that I've never been more comfortable on a day-to-day basis and more fully employed than launching this business. I suspect there's no going back.
Niels: No, true. You mentioned that your team reads a lot of books, and you've obviously done your share as well in what you've done. But, are there any books that, from your perspective, had a bigger impact on you as a person in all of this? It could be related to trading, but it could also be related to other things, just something where you felt, by reading this book, I really learned something.
Dave: I think you make a great distinction between trading books and non-trading books. I was recently asked by one of our partners, whose son has declared that he wants to be a hedge fund manager. And so this son asked me what he should study in school, and he gave me some choices. He asked if he should do a business degree, and economics degree or an economics degree and then a business degree? Or a business degree and then an MBA? And I said, the best traders I know went to study English, and if you're going to read a book, read Shakespeare because nobody since, has captured the frailties of human nature like he has. If trading is about anything, it's about the frailties of human nature. So that sort of what comes to my mind when you talk about books. Recently though, I've read "The Innovators" by Walter Isaacson, and I think that's a wonderful story about innovation and about how challenging innovation is even when it works. You can look at the steam engine; you can look at the first computer. These things took decades to come to fruition and be adopted, and so that gives us some patience. Yeah, that's what comes to my mind. Peter Thiel's new book is quite good, "Zero to One." I think that's an interesting read, especially anyone who wants to be in the venture space.
Niels: I asked you before about the entrepreneurial journey, and you obviously said that it was part of your genes. As entrepreneurs, we have our failures along the way, and... what about you when you look back on your journey? Are there any things that spring to mind in terms of where you say, yeah, I failed at this stage, I could have done it differently, or maybe you don't look at it in that respect?
Dave: I think at the deepest philosophical level, it's not a great idea to go back and say what I would have done differently. But it's also a great idea to be open minded about your failings, and certainly the self-examined life is the one worth living. So we as a group spend a lot of time on that. What I would say about my personal journey is that I waited too long. I would say that I was an imposter working for the big law firm, and the big bank and the big mutual fund company. And I really, if I was to do it all again, I would take a lot more risk a lot sooner. And what I mean by risk isn't reckless risk, it's just... if you're going to get to a result that puts you in the top quintile of anything, you can't get there by being comfortable all the time.
The risk you want to take, now I sound like I'm pontificating, I don't mean too, I'm teaching myself the risks I would take in life are the risks that, yes, have downside. But the amount of upside is commensurate or even greater than the downside. So what we look for going through life or what we call pod odds: if you know for example, that the chance of making an outside straight are say eight in forty-two, so let's call it one in six, or one in five. And you know that you're being asked to make a bet that has a ten to one payoff, well then you should probably make that bet. It's true that you could lose, but if you get your money in good, each time you get the chance throughout your life, then the rewards are amazing. There're far too many people, certainly I've seen it, I was a victim of it, who were not prepared to take that risk.
Niels: You mentioned earlier on in our conversation that you have three wonderful children. If you could take just one of your own skills and pass it on to your children, which one would it be and why?
Dave: Is optimism a skill?
Niels: Could be, I'm not sure. Could be.
Dave: I think that's what I would pass along: I think this world we live in is quite magical. I think the great deals go to the optimists because they're prone to paying more for them. The great experiences go to the optimists because in life you probably find what you're looking for, and the optimists tend to look for great things and expect great things. So that's what I would bequeath to my kids if I could.
Niels: Great answer. Just a couple of more questions Dave. Is there a fun fact about yourself that you could share? Something that even people who know you reasonably well may not know about you? Hidden talent? I don't know what it could be?
Dave: Well the people who know me know that I'm six foot nine, so that's kind of a fun fact. But I think perhaps... I referred to Peter Thiel's book recently, and apparently he asks a question in his interview process that I love. And the question is this: what philosophy do you have that when you express it out loud tends to get rejected by the masses? What a great question. And I think that my philosophy that tends to get rejected by the masses is that human beings are very creative people, very creative entities. We create our own realities; we create our external circumstances. And we do that, probably, in the quiet times opposed to the busy times. So, I know Ray Dalio at Bridgewater is a big fan of Transcendental Meditation, as am I. And I think that if I trace back the great fortune that I've had to meet the people on this project, and to achieve what we've achieved, I can actually trace it to a time when the meditation was working in my life and my own vision of my own future was changing dramatically. So I don't think those things are coincidences. But if you go out into the world and say hey, if you think it, it will come, people dismiss it. So there you go, there's my revelation, character revelation to you.
Niels: Great stuff! And now, I talked earlier about investors possibly not asking all of the questions that you would like them to ask you. So I'm going to turn the camera so to speak, or the microphone on myself and just ask you if we have been covering the right questions today? Is there anything that I've missed? I want to be sure that I do you and your firm justice. Is there anything that we've missed today?
Dave: Niels, I don't think so. It maybe... we've taken... I haven't looked at the clock lately, but I'm sure it's a great amount of time, so I would say that you set yourself apart from the crowd by virtue of when you and I chatted at the Battle of the Quants, you weren't... you didn't have the predisposition of "this can't be true." When we go around the world these days, we like to say the phrase, "tell me more." Even when somebody says something audacious, instead of saying, "Oh that can't be done!" We tend to use the phrase, "Tell me more!" And that's what you said at the Battle of the Quants after the panel. "Tell me more." So I think you set yourself apart in the terms of the way you interact with the world, and the list of questions you've asked today are wonderful for lots of different constituents. And I certainly feel that we've had the chance to express our story in a very fulsome way.
Niels: Great stuff. Now before we finish completely Dave, what's the best place for people to find out more and reach out to you?
Dave: Our website is a wonderful place to go to and see a lot of the things we've talked about. There're some videos on there that are great. In addition, there's a place to subscribe to our monthly information. So if you just put your email in there, we can send lots of stuff to you. And that website is: kflcapital.com
Niels: Fantastic, great stuff, and of course we will put on our web page in the show notes, we will, of course, put links and some other great stuff about your firm as well. So all I've got left to say Dave, is thank you so much! You took time out to share your insights and your vision and your story, which I thoroughly appreciate. And I hope people have learned a lot about a new way of trading which we don't come across so often. And I think it's exciting to hear about these things and I wish you all the best in the future with this. So, thank you very much Dave.
Dave: Thank you Niels, that was wonderful.
Niels: Take care. Bye bye.
Ending: Thanks for listening to Top Traders Unplugged. If you feel you learned something of value from today's episode, the best way to stay updated is to go on over to iTunes and subscribe to the show so that you'll be sure to get all the new episodes as they're released. We have some amazing guests lined up for you, and to ensure our show continues to grow, please leave us an honest rating and review on iTunes. It only takes a minute, and it's the best way to show us you love the podcast. We'll see you next time on Top Traders Unplugged.
Become An Insider
Subscribe for free and be the first to receive new and exclusive interviews with the world's top traders. As an insider we'll also send insightful bonus content direct to your inbox.Free Instant Access »
You might also like:
Date posted: 04 Dec 20144 comments