In partnership with:
CME Group

88 The Importance of Explaining Why with Andrew Baxter of Cambridge Capital Management – 2of2

"I believe track record is a very blunt instrument." - Andrew Baxter (Tweet)

In this episode, we dive into the models and processes that Cambridge Capital and Andrew Baxter use to manage risk, research new ideas, and grow their business. We also talk about the importance for managers to explain why they do what they do, rather than just what they do.

Listen in to learn more about the future of the managed futures industry and how to be successful in it.

Thanks for listening and please welcome back our guest, Andrew Baxter.

Subscribe on:

Subscribe in iTunes Subscribe on Stitcher Subscribe on Soundcloud

In This Episode, You'll Learn:

  • How his forecast method works
  • How he develops his basic models
  • How they come up with new ideas for models at Cambridge Capital

    "One can ask the data to suggest areas for interesting research." - Andrew Baxter (Tweet)

  • The various short term and long term models that they employ and how they use them
  • When they make interventions to the system
  • The risk management they use and why they have a risk budget

    "We actively control risk on a daily basis." - Andrew Baxter (Tweet)

  • Why they are continuously adjusting their risk in the market
  • The drawdowns that they expect to see in their program and the ones that they have been through
  • Why he thinks CTA firms had significantly higher drawdowns in 2013 then they had ever had before

    "I think the increased flow of assets into the larger managers is potentially increasing market impact problems." - Andrew Baxter (Tweet)

  • How their research cycles work
  • How to compare a life track record with a simulation of the current program
  • Why we should look at why managers do things more than what they do
  • Why raising assets is the biggest challenge his firm faces today
  • What it takes to become a great fund manager

    "We need to continue moving and increasing the sophistication of what we do." - Andrew Baxter (Tweet)

  • The skill that he would pass onto his children
  • The challenges he sees for the CTA space

    "We should consider an audience outside the traditional CTA space." - Andrew Baxter (Tweet)

Resources & Links Mentioned in this Episode:

"One should not underestimate how long things take in this business." - Andrew Baxter (Tweet)

This episode was sponsored by Eurex Exchange:


Connect with Cambridge Capital Managment:

Visit the Website:

Call Cambridge Capital: +44 (0) 1223 851 001

E-Mail Cambridge Capital:

Follow Andrew Baxter on Linkedin

"We have a structure to control risk without discretionarily turning models on or off." - Andrew Baxter (Tweet)

Full Transcript

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!


 … with your explanation of the forecast method and how it all evolves in the process. 


So we have some 5,000 data points a day, and what we’re trying to do is identify how well that model is doing. So if we asked a person what his forecast was for the S&P for the end of the year and he said, 10,000, then we probably suspect his judgement was quite poor. If his forecast would be the same price as we closed on Friday, we might think his judgement was better. So when we ask somebody a question, we make a decision about the quality of the answer. Sometimes that quality of the answer is based on how high quality of hit that individual’s answers have been previously. So really what we want, when we ask a model for a forecast is, “What is your forecast?” and “What level of confidence do I have in the answer you gave me?” 

So we don’t ask models for forecasts, we ask them for both: their answer, their forecast plus the uncertainty or confidence that they have in that. Knowing that, we know how to combine all these forecasts together to create a single combined forecast for the asset. So let’s think of a practical real world example. 

You are the manager of a proprietary trading desk. You have ten people on the desk. They all trade 10 markets each. You have capital you need to allocate to members of that desk. Your most experience trader trading the most profitable market this year - US equities, gets the biggest allocation; and the least experienced trader, the new graduate on the desk - trading Japan, gets a very difficult market, gets the least allocation. You make a judgement based on the underlying competence of that model and the market on which it is trading, when you decide allocate capital. 

In a similar way we can do a similar thing here. So when we look at the forecast coming out of our model, we look at the performance of the model itself. We also look at the performance of that model applied to that market. So how might the system… what are the failure modes of the system? 

Well we know that carry models which we talked about before, carry models work, and then they don’t, for a while, and then they’ll work again. When they stop working, they tend to stop working across a wide variety of assets. When they stop working, different types of carry model stop working across a wide variety of assets. So you’ve got quite a lot of potential information there that you can use. There’s some groupings and structure, some knowledge of how markets behave that you can take into account when trying to learn whether this forecast is going to perform well tomorrow. 

If a carry mode on asset A, one day goes by and you have one more data point. You can’t do a lot with one data point. But if you look across the cross-section of carry models and the cross-section of markets on which you apply them, you potentially have hundreds of data points every day that you can use to help learn whether carry models on those assets is going to perform well tomorrow. 

It’s that approach of understanding the drivers of the performance of the models, the drivers of the market, that is the main knowledge. That knowledge of the investment process using that as a starting point to apply dynamic or machine learning techniques is a key. It’s transformed the problem from being his 5,000 data points and ten years of history go and work that out, to let me give you some strong hints of what is the system underneath that’s driving this data. 

We talked before about the importance of, in finance, understanding what the system is that is driving the data because without that it’s just a very noisy data problem that machine learning techniques don’t work well on. 

So to answer your question, ensemble techniques is a dynamic mechanism for allocating across a committee or a pool of forecast supply to assets to generate a single combined forecast for each of the assets. 


Sure. So I’m going to try and make it even more simple so that at least I understand some of the challenges and how you deal with that. Maybe I’m going in completely the wrong interpretation of what you said, but I’m going to go for it anyway. So I’m thinking that you’re trying to forecast, to a certain extent, how the model should be overall positioned today, relative to how it was positioned yesterday. If there is a slight difference, no doubt I’m sure your model will adjust for that position wise. Am I OK so far? 




So what I see often, when talking to people is that essentially when you put all your techniques together and all your models and so on and so forth, at the end of the day they will give you what the position should be in each of the markets that we trade, what should it be based on all the information and processes that we have. If it’s different from yesterday, if we need to buy more Euros, we buy more Euros. If we need to do something different we’ll make the change. So that’s how we get to the current position size. 

Here’s my challenge, I’m thinking that your forecasting, and I could be wrong here, that you have all these different models trying to give some kind of input and if you look at the environment, for example for long term… I don’t even know if you are… I imagine you have some long term momentum or trend following models in there, and you might have some very short term momentum models in there. 

I’m thinking, it must be very difficult to know how to weight these two models because one might have a very strong conviction today, could be a short term trading system; but a long term trading system, at least in the trend following space, it might not look very attractive today. If you don’t stay with the position unless the signal changes, I agree with that, but sometimes you just have to ride out the pain – the correction, because often if you don’t it’s too late to reap the benefit. 

So where I can’t connect the dots in what you’re trying to explain is how do you align those different points if you’re asking your models every day where should I be positioned? When some models are looking much longer into the future and maybe very profitable in the long run, but may not be very confident today, but you don’t want to miss when the big move happens. I don’t know whether that makes sense. I don’t know whether you can address that – how to combine so many different models. 


I think, in the answer to your last question, we talked about the importance of not seeing it as an abstract data problem, but thinking about some structure for that data. What you’ve just said to me is I believe that there’s some structure in the world of models. There are short term models. There are long term models. There are models that have a certain behavior dynamic and there are other models which have a different behavior dynamic. 

Trying to find one approach and apply it to all of them, universally sounds like you think that that’s probably not a very sensible thing to do, and I agree with you. If we saw this as a data problem, it would be very hard for a lot of techniques to identify these different categories of models that are creating data of different dynamics: slow moving, trend following type; fast moving, reversion types; or whatever the catch regime might imagine. 

So What you’ve just said, I understand this investment space, and I understand that there are models of various types. In actual fact, I kind of know that because I wrote them.Therefore, I expect them to behave differently, so maybe we should treat these separately when thinking about how to combine them together. That’s an important observation, very important. 

Bringing the knowledge of the problem, the domain, simplifies the data challenge significantly. We need to do that. So yes, it is necessary to take that into account when one’s combining different techniques together. The fact that there are fast moving, slow moving, and different categorizations. 


 Ok, ok, that makes sense. Now in terms of developing the basic models, which I assume is you and Ian who are behind that, how do you come up with new ideas? I’m jumping a little bit into the research side as well. I wanted to talk to you about that as well, but I’m just curious, how do you come up with new ideas? I imagine it’s… the idea comes first for a model, or are you looking at the data and saying, “Well, actually here, if we did this then that might look attractive.” How do you do that? 


The idea of an understanding of the dynamics and behavior of markets is often the genesis for ideas. One thing we have taken advantage of in the last few years and I think that will be becoming increasingly important is the fact that one can ask the data to suggest areas for interesting research. So a panacea might be that we would send a machine over there to go off and build trading models and it will come back and say, “I’ve got the ultimate trading model.” We know that’s very difficult, because often one might find something that works in backtests, but has no… it can’t be explained. 


 Chance of survival. 


 Well, yeah, and it can’t be explained, maybe it doesn’t work out to some point. I think we can think of so many examples of that. Certainly many years ago we looked at genetic evolution of models. Millions upon millions of models and iterations to build the fittest and finest models. It’s a really hard problem. 

The looking for statistically significant patterns in data where that can be very useful is to hint at areas for further research. For us for a model to be acceptable it has to be based on sound investment process and understanding how markets behave. Once we have that foundation, then we can use a variety of different tools to implement that. 

One of the great advantages for us of linking into the guys of the engineering department here is there are tools which, maybe I’ll be very well understood in other fields other than finance that we can borrow and bring across – we can create a better mousetrap, a better method of implementing what we do, which gives us an edge and maybe generates a different return path which is important because we’re not with the crowd running for the door at the same time as prices are moving. 

So different techniques and more sophisticated techniques are very helpful. It has to be based on sound understanding of how markets behave, but increasingly we can use data, I really don’t like the term data mining, but data mining exploration techniques to hint at interesting areas for further consideration. 


 Sure, sure, I think that’s fair enough. Now, if I’m right in my understanding that essentially all of these different models on a particular market will each give their vote, both confidence and profitability I think, although I’m not using the same terms as you are, but I think that that’s sort of in the direction you were explaining. When you look at that, this is just to get a feel for the program itself, so since you’re not looking on a trade by trade basis, what would you expect to be the average length or duration from when the combined signal, if we call it that, starts getting long, or short. It doesn’t really matter. Until it changes direction. What I’m trying to get at here is how slow, how fast, when you combine all of these things, is the overall approach? 


 We have models that trade and hold as short as one day, and some will hold nine months. At the portfolio level it’s a blend and, as you say, it depends whether the market conditions are such that shorter term models are more effective at delivering a high quality of return, or actually optimizing transaction costs and moving much more slowly is more effective. That is a system determined weighting. It’s a continuous adjustment of weight in the portfolio every day. If we were to think of it in terms, as you said, from speed of turnover, it’s in the order of two to four weeks. But that does vary over economic cycle and varying market conditions. 


 Sure, of course. Have you ever had to… Obviously this has been a long process for you in terms of developing it from even before CCM, but have you ever had an incident where you felt, “We’re not sure what’s going on in the market and how it impacts the portfolio so we need to override what we’re doing,” or are you one of those who would say, “Absolutely not, we’re never going to make any intervention into the system?” 


 We’ve tried to identify the failure modes of what we do. We’ve tried to diversify away risk where we can across a variety of models, a variety of time scales – adaptive weighting of those. We bring all our forecasts up to a portfolio construction level and risk management level. We build a portfolio using a market risk model. 

We build a portfolio of holdings that we rebalance to and that portfolio contains within it the risks that we believe are the largest marginal return to that portfolio, but constrained by concentrationally risk limits, by asset class. The maximum standard deviation of our variance Var and our expected shortfall of the portfolio as well. 

So we recognize the risk model in stable markets will be a very good forecast of tomorrow’s risk in the portfolio. When markets are not stable, when we have events in the market, we need to recognize that risk portfolio, that risk model may fail and we need to put hard limits on the position – that we can manage the positions we take across the different assets. We impose constraints there as well. 

So, do we intervene? No, if market conditions evolve such that we become concerned about events, then we can revisit the risk limits that we apply at the portfolio level, or the individual holdings level. We operate under an institutional style risk management process. We have daily risk meetings of the risk committee and within there are considered whether there’s anything that we’re aware of that would warrant intervention. 

Having said that, over the period that we’ve been running this portfolio, that’s barely ever warranted – adjustment of those risk limits. There’s a structure in place there to control risk without discretionarily turning models on or off, or so forth. So it’s a fully automatic process. Risk management is built into the whole process all the way through. 


 Sure, and actually, your foresight is great because the next area I really do want to talk about is the risk management. You’ve already touched upon some of the things. What would you say, if you picked one and maybe there is more than one, but what is the main risk framework or definition that you target? Is that a Var or standard deviation, or is there one overriding of all of the ones you mentioned? 


 Risk for us is a number of things. I think what we’re trying to deliver is a high quality risk adjusted return to the investor so the volatility of that return should be low relative to the return we deliver. We actively control risk on a daily basis and that’s controlled at multiple levels as we talk about the errors in the models diversified away, models allocated dynamically across the ones that work, and then we build portfolios. 

The portfolio we build is a set of holdings in the markets we trade that maximizes what we expect is the return for tomorrow for a fixed quantity of risk. We have a risk budget and we build a portfolio to that budget. We forecast the future of variances and covariances and how those markets are going to move in the near future with respect to each other. Obviously some markets diversify each other and we take that into account. 

Our risk model, won’t be right, we don’t have a crystal ball. It will have errors in it and those errors will be greatest when there are events in the market where markets become volatile where markets become more correlated so we recognize that and we put limits, concentration limits on the different holdings that we can hold. 

As a backstop to that the typical risk model has a view that the world is normally distributed but it’s not. So we need to think about that left tail. We need to think about the possibility of extreme events that will result in large losses and we look at our portfolio and look at the historic performance of markets and how much pain there might have been in the past if we’d held this proposed portfolio in a variety of different scenarios. What’s the worse outcome we would have seen, so we look at Var, and all the tail risk measures as well. 

Then on a final, once the portfolio has been constructed we also run a wide variety of market risk scenarios through there as well. Whether that be September 11th or whatever, to try and explore if things really did go wrong, is this a portfolio that we want to hold? So there’s multiple levels by construction of safe risk, checks and balances as well, to deal with where things may be different then we expect. 


Sure, sure. Just to simplify things, for me to be able to visualize what it is that you’re doing, is your risk budget fixed on a daily basis or does it actually vary? I know you have a limit and that’s probably fixed. I imagine that doesn’t change, but the actual… I’m just curious as to how the data… because I think this is one of the most interesting things I’ve come across recently and that is that many managers have historically always said, “I target a fixed volatility,” so the portfolio has a 15% volatility – annualized volatility portfolio, and that’s how we sized the positions, etc. etc. 

In my own experience, at least, I don’t know… I don’t think that’s a particularly efficient way of doing it because we know, and I come from the trend following world, but we know there are periods of time where the environment for trend following is not very conducive, so why would you target the same level of risk or volatility in the periods where there are fewer trends? 

So, again, for those managers who have mastered the technique of, in a simple way, describing it being able to measure the environment for their strategy, be it trend following, be it something else, it doesn’t matter, and then set the risk budget accordingly. I’ve only come across very few and even fewer of those who have been able to demonstrate success in doing so. How do you… Is there any of these things that I’m talking about here that you do, or how does it work inside your approach? 


I think there’s a few traps. If we go back to where we were talking many years ago when people were typically thinking about the world as a trade. They’d place a dollar long or a dollar short, or maybe recognize that a bond market has a different level of risk than a natural gas market, so actually we’ll place a risk adjusted dollar long, or dollar short. 

An evolution of that is to say, well, actually natural gas and all the energy contracts have correlated in some form, so maybe if I’m a dollar long on one I probably shouldn’t be a dollar long on the others as well, maybe a bit less than that. We take that correlation into account, but maybe natural gas isn’t correlated with the cheese future. 

So people have always talked about the world in an expected risk and correlation sense. What we’re doing is doing that in a more formal basis. So we build a risk model which encapsulates what do we expect the future volatility of each asset to be and how correlated do we expect them to be in turn? 

Now that means we can build a portfolio where asset A maybe has a smaller position than it would have done because we’re also long B, C, and D, or maybe asset A hedges B, so that’s all great, but there’s some traps. One of the traps is sometimes markets are quite benign and have low volatility and, as we all know, risk builds up. 

You can take larger and larger leverage and markets are not moving. So that’s impacting these returns and no risk is coming through and then risk does come through. There’s an event or something and the prices move dramatically and everybody is limit long on risk on all positions and that can be very painful. 

Well, can we deal with that by looking at the volatility of those assets? Well not really because we’ve had a period of 3, 6, 12 months when prices have not moved, and we only look back 3, 6, 12 months to assess the volatility of that asset, then it didn’t move, so the volatility is zero. We’ll have a large position in that asset because it’s not very risky until it is. 

So the risk model needs to incorporate the fact that assets can behave for relatively long periods in a way that suggests that they’re benign when in practice they’re not. They are capable of jumps. Their distributions are not normally distributed. So they need to take that into account and deal with that.  So you need to have quite a sophisticated risk model, first of all, to constrain the risk that you will take because you’ve been misled by your assessed risk. 

Secondly, one thing that we do at a portfolio level – at the very top level, is we have a risk budget as you say. We’ve been running a 10% annualized standard deviation of daily returns risk budget. So when we build our portfolio, if we’re at high-water mark, we try and take 10% volatility risk. 


If we have an active return and we drawdown and then we continue to have an active return and we continue to drawdown, then the portfolio is telling us something. In spite of all our efforts to build models and dynamically weight them, and build risk, and control, whatever we’re doing across our whole system, it just isn’t working, if we’re continuing to publish negative returns. 

So our ability to forecast and build robust portfolios in the market is not working. We probably should do less of that in the future. If we believe that markets tomorrow will have similar characteristics to markets yesterday, if they’re serially correlated, we probably should do less tomorrow of the thing that wasn’t working yesterday. So therefore we reduce our risk budget. 

So as we drawdown we stop building full risk budget portfolios and we build portfolios with lower risk. Then when the market starts performing again, and our approach starts performing again we start building higher risk portfolios. 

The reality therefore is that we are continuously adjusting our risk in the market, both across assets, maximizing the sources of return by model, but also at a portfolio level and just take less market risk in difficult to forecast markets. So we talk to some managers and they say, “Well I gone long or short a dollar. I have a take profit stop loss, and I apply the same approach at my portfolio level. If I have a 20% drawdown, I’ll half my risk.” 

For those managers that are trading trend type strategies which can have rapid drawdowns and bounce back, keeping risk on so one has the full risk allocated for the bounce can be a risk management approach. What happens when it doesn’t bounce? When it goes down 10%, down 20%, half the risk is taken off and then it bounces; you have to make a lot more than 40% to get back to where you started because you’ve only got half the risk on and now you’re at 80% trying to get back to 100% with half risk. 

So actively managing risk on the down side is what we do. It’s akin to having put protection at the fund level. So we could go to the market, we could buy a put option to protect the return of our portfolio. Now, no one will sell us a put option because no one knows what we have in our portfolio and it’s changing every day so you can’t go and buy that, but as a risk manager you could construct your own. 

We came from a derivatives risk management background where we were continually hedging options in our book. We dealt with hedging those options, so we’re doing a similar thing here. We are taking risk off as in our force or putting risk back on as in our varisons. That delivers a much smoother risk profile and return profile at the fund level. It has a cost. You’re taking risk off, you’re putting it back on, and there’s transaction costs there like there would be with hedging any… synthetically creating or hedging any option position. We believe that drag on performance is more than justified by the high quality control of risk and downside risk control. 


Sure. Speaking of drawdowns, which is the next topic I wanted to ask you about, in a strategy like yours, what kinds of drawdowns do you expect the program will suffer from time to time and how does that align with what you’ve seen so far? 


 I think, in the long run, a manager that can deliver a net of fees return of a sharp ratio of 1 and a drawdown to vol ratio of 1 I think is doing a good job in the long run. We think the active management of drawdowns should deliver better than that. It does depend on the nature of the way the drawdown has occurred, clearly if we’re taking risk… adjusting risk on a daily basis, then if the drawdown occurs gradually and slowly over many days, then we’re slowly taking risk off. If the drawdown occurs over relatively few days, then we’re not taking risk off minute by minute, but we’re taking it off day by day. If a drawdown occurs over five days, then we probably held more risk, on average, than if it occurred over a hundred days. 

A 10% drawdown on a 10% volatility portfolio I think would be a good result for many managers. 


Yeah, I think that’s an excellent result. A sharp of 1 for a CTA is significantly higher than what the industry has delivered which is about .4, so I think that’s an excellent result if you can deliver that. Obviously you can’t know what other people are doing, but I’m just curious to your point of view. A lot of firms are doing some really sophisticated stuff out there and there are firms with a lot of bright people yet most of them suffered significantly higher drawdowns in 2013 that they have ever experienced before despite all the brainpower, despite all the computer power that we have now-a-days. Why do you think that is? 


 I guess we will, without perfect ability to look inside every firm, that’s hard to say. I think that there are perhaps two things to think about though. One is, and it doesn’t make sense as it’s the same thing, it’s the concentration of approach. We talked about before, some of these risk premier are easy to access. I can buy a book on Amazon. I can implement a trend following strategy. If we all use the same model and then we all decide to get out, we’re all rushing for the door at the same time, which causes prices to overshoot. 

We have seen a structural change in the business in the market CTA managed futures market in the last, I think across hedge funds in general, in the last ten years. As smaller investors coming through the private banking channel, coming through the fund the funds channel have decreased in importance post Madoff and concern about operational risk and due diligence and so forth. The assets that are flowing into the industry tend to be flowing into the larger funds. 

Given the choice of being able to invest with a hundred managers, managing a hundred million, or one manager managing ten billion which is the better choice? Well if there was no operational risk, there’s no business risk, then a hundred managers managing a hundred million is likely to give some diversification because the techniques will be slightly different, the timing will be slightly different, there will be smaller market impact when those managers trade. 

Going to the ten-billion-dollar manager, he’s unlikely to have the same level of diversification that a hundred smaller managers would bring. So I think the increased flow of assets into the larger managers is potentially increasing market impact problems. When we have some… some of the events we had in ’13, we saw a dramatic decrease in the dispersion across asset markets, a correlation between them. We had political risk factors, and a lot of event risk, and I think that’s what we saw coming through. 


 It’s interesting, isn’t it, because I think a lot of investors probably favor the big managers because they feel that that is less risky, but what you’re saying, and I agree with that, inadvertently in doing so they’re actually creating a lot more risk in the system. Similarly, as we’ve seen from central banks trying to lower the risk in the system and talk about systemic risk and all of that, but in fact with all the increase regulation and changes, we have a lot fewer but much bigger financial institutions today and if we’re going to use the same rule, in a sense the system may actually hold much bigger risk today than it did before the crisis that everybody got burned in. 


 I think that’s true, but I think the decision to allocate to the large manager is the right decision given the criteria those large allocators use. Investment performance is one of many criteria. I think fund the funds used to insulate the smaller managers from some of the risk or diversified away the single manager risk so larger allocators could access those pools. It is harder today than it was and that increases some of the market impact challenges as we said. 


 Sure, sure. Now drawdowns are clearly very emotional, both for the managers and for the investors and you probably, through experience, have found ways of dealing with those emotions, but I’m happy for you to talk about that. My real question is, looking at the world today, looking at your approach, your systems, everything you’ve done, is there anything that keeps you awake at night; some kind of risk that you just know that you can’t model? 


 We try very hard, as I think you hinted on that in the question, to try and find the risks and address them. It’s the risks we don’t know, it’s the unknown unknowns which is the famous – I think it originally came from NASA, originally before Rumsfeld. It’s a well-known long standing phrase, but it’s the unknown unknowns. The known unknowns we can build, we can diversify and manage the risk we know. We can build circuit breakers for the risks – the known unknowns, but it’s the things that we don’t know about that we didn’t know they existed so we couldn’t build some safety measures in place. That’s the risk to any business and it’s a business both an investment risk and an IT risk is a business risk. 

Does it keep us awake at night? We have tried systematically through the way we built our business and through process to make everything we do repeatable. We’ve learned a great deal over the years and at HSBC also. So we believe there are very few unknown unknowns. By definition, no one knows. 


Sure. I want to just touch upon another topic as we go into the final part of our conversation and it’s, of course, research. We’ve already talked about some of this, so I’m not going to dwell on it too much, but in terms of the research cycle, from idea generation to model implementation, etc. etc. Different people have different ways of doing this, what does your research cycle look like? 


For us we have an idea. We use a variety of software tools to test the idea. They could be rapid prototyping tools, data visualization tools, tools which explore the search space of different models that might help us extract a premier in a certain way. One of the things that we spent a lot of effort on when we started CCM was thinking about the research tools that we need to use. 

I think it’s a lot easier and it’s very convenient for people in a research environment to live in a cozy world where they don’t have to face up to implementing the thing that they researched. What that means is one can cut corners, one can use scripting languages, one can make assumptions, one can write much shorter code and simplify the approach if one doesn’t have to think how would I implement this. One thing we spent a great deal of time doing was to ensure that our trading system is our backtesting and research system. 

So it’s the same code base. It uses exactly the same step through time. The only difference, really, is the system’s told it’s the 1st of January, go and trade; it’s the 2nd of January, go and trade, it doesn’t go to an exchange. It goes to a synthetic exchange with a transaction cost model and it gives representative fills of what you would have had at the time, and it does all the end of day processes, post margin, funds cash, etc., etc. It’s a fully faithful backtesting system. So that when we move something from research to production, it represents reality. We know that we can trust what we’ve tried in the lab. 

So we go through an idea generation. It’s got to have a solid idea, a solid driver of return; it can’t be purely a data driven approach, rigorous backtesting. We look at the marginal contribution of that idea to our portfolio. One of the interesting benefits of our approach is that we can put a model into our framework and its influence on the position we take in the market is a function of its marginal contribution. How much better is it than we’ve already got in there? If it’s not really any better, it doesn’t have a large influence in our forecast and therefore in our portfolio. If it performs well for a while and then stops performing? It was a badly specified model. It gets deselected, confidence goes down, and it gets turned off. 

So for us, it’s not like we’re carving out a block of capital from all the models and allocation across to all the new model when you put it in the portfolio, and if it underperforms we’ve not only had a loss, but missed an opportunity to use that capital elsewhere because it’s a self-adjusting system. 


Sure, sure. If you were going to put yourself in the investor’s shoes, looking at either your own firm, or other similar systematic firms, what would you rather look at: a three or five-year live track record, or a twenty-year simulation of the current configuration of the model? 


Has that team ever traded live or do they…? 


 Yeah, yeah. Again, going back to one of the early questions we talked about, for all new managers, length of track is difficult to deliver, but actually even with people with a twenty-year track, I would argue that it’s probably not worth a lot because the model will usually have changed so many times that whatever investors are looking at is going to be very different. 

So my question is really, are live track records… how do you compare a live track record with a simulation of the current configuration of the model? Could investors in theory, at least, be better off with a simulation, understanding all the obviously that people could do a simulation badly, and it’s a dangerous thing to look at, but let’s just assume the people that we’re talking about things that are done the proper way and not over-optimized. 


 So I think a twenty-year track record is useful. If a team has continued to deliver strong performance over twenty years, it has demonstrated that it has a skill to adapt to markets, and that’s what investors are buying – a skill. As to whether a track record or a simulated track record and what are the relative merits? 

I think track record, unfortunately my comment isn't going to change anything, but I believe a track record is a very blunt instrument. Investors shouldn’t be looking at track record. What they should be doing is understanding what the manager does, but more importantly why he does it. Why did you make the choice to do this thing that you say that you do as opposed to other things you could do? Just define that, do you have an understanding of what you’re doing? Do you have a process? Do you have the skill to deliver that repeatedly in the future as market conditions change? 

A track record, as we’ve said, is evidence of performance which is skill plus luck so they need to go much deeper than that. A backtest? Yeah, it’s useful. It shows me that something that you did produced some numbers. The problem with the backtest is anybody can produce a great backtest. So you need to get far deeper to understand why did you do the thing you did when you constructed your models which produced this backtest? 

Why is live track record more important? Because live track record you can’t hide. There may be assumptions made in the backtest, but also market conditions change and a track record catches that. What investors should be asking is I see a three-year track record, what did you do as you passed through those three years? What did you change? How did you evolve as market conditions changed? Why did you do that? Show me evidence that you have adapted in the past and you will continue to adapt for me, as your investor, in the future. I’m not buying a point in time system. We’re not going to take your system and put it on a USB stick and take it away and trade it. I want you to be continually improving and adjusting and adapting and questioning the thing that you do. Not tinkering, not interfering when not necessary, but we don’t want a fifteen-year-old system. 

So I think the investor should be looking through and having detailed conversations about why to assess whether that team has the competence to deliver good returns in the future. Backtest alive is just some evidence of that. 


Sure. Sort of a slightly different question is as we leave the research, I’m just curious, if you won the lottery tomorrow and you had, suddenly, an extra million dollars to spend on research in 2016, what would you spend that money on? 


 If I look forward, if it had to be in research, there are many things as an investment manager one can invest in outside of just the research, but if I look at research, if I look forward where markets might be 15 years from now, and when I look where they were 15 years ago, we were voice trading, then we were screen chat, then screen trading, and then we’ve got machines making markets now. I’m reminded of the equity start-up work that we did at HSBC and the markets we traded then. 

Equity startup was fast moving from being something which is traded in the late 90s on perhaps a monthly basis towards being traded daily and then intra-day. It looks more like a short term market making strategy. I think, within the CTA space, we’re trading risk premier. We have a systematic approach where IT sophistication levels are high. 

I think we’re going to move in that direction where we’re increasingly trading a blend of forecast models from short term, minute, hourly, way out to one year. People will need to embrace this to stay ahead of the complexity curve, otherwise we risk, with those managers that are trading using today's techniques in ten or fifteen years’ time, I’m sure today’s techniques are going to be available on Amazon, in ten years’ time. 

We need to continue moving and increasing the sophistication of what we do. So I would be looking into research in those areas. 


Sure, sure. I want to just touch very briefly on what I call the business side of the firm, but not much, and then jump to the last section that I want to cover today. It’s to get an understanding. What do you think are the biggest challenges, from a business point of view, where you are and where many firms are in our industry? What are the key challenges you face today? 


 Raising assets. 


Yeah. How do you… because I think that’s something that most people would say, yeah, that’s really tough, even the bigger ones. So how do you overcome that challenge? 


 I think it’s… we talked before about the move towards large allocations from institutional space and the very largest managers. I think it’s difficult for an emerging manager, even a manager with strong evidence of skill in the past, to grow rapidly because there is a process. One needs to show evidence of skill in the current organization. One then needs to initiate conversation with institution allocators. 

Those conversations, as you well know, move at their own speed and are relatively slow. I think at the present time one of the challenges is that the allocations into the CTA space are… new money coming into the space is coming in slowly, and the money that is already allocated in the space is turning over relatively slowly as well. So I think what we need to do, as an industry, is to perhaps look at our investor base and ask ourselves whether we are delivering what they want, rather than considering just the investors that traditionally allocated to CTAs and managed futures, whether we should be talking to a wider investment audience. 

The fact that we implement our approach through futures doesn’t prevent us from saying, “This is an asset management business and we should be talking the language of the broader asset management investor community and look to those investors to bring us into this space.” I think that’s a big challenge for the industry. It’s a hard thing for a smaller manager to do, but one can think much more broadly than just the traditional allocators. It’s easier for a smaller manager to be more nimble and more creative in how he delivers his product the perhaps some of the larger managers who have an established product and distribution channel. 


 Sure, sure, absolutely. It’s a big topic. It’s a big issue, and a big challenge. The next question is, I think is the last one in this section and it’s not directly relevant to you, but I will ask it anyway because of your long experience and because it’s a question that I get from people listening. Today, my understanding is that you don’t have an offshore fund that you run. So you don’t necessarily have to deal with this issue, but at some point I’m sure that you will. 

In our strategies we don’t use very much of the cash, 5%, 10%, 15%, 20% maybe goes to putting on the positions as margins. So there’s a lot of cash sitting around not doing anything. If you had a fund today, how would you, in the world we live in today: the zero interest rate environment and all the talk about rate increases we’re hearing. How would you allocate, or what would you do or have done to all the spare cash in a fund? Because I think people sometimes forget that that can actually be a big risk in the strategy itself if that cash is not managed properly, so do you have any thoughts of how you would do it if you were in a situation like that? 


You’re right, and there’s been increasing focus on that in recent years, where we can look at the standards that have been laid down under UCITs and the expectations there from cash management and counterparty credit risk management. It is tempting for, perhaps, a manager in these markets with yields at these levels to think more creatively as cash management as an opportunity to improve returns available, but we’re in the business of delivering returns from the managed futures asset class, and cash is just a product of the fund investment, so that should be invested across sovereign and a ranged cash counterparties to minimize that counterparty risk. 


Sure. Before we jump to the last section I want to ask you a question that I try to ask everyone and in particular someone like you who has been in, I’m sure not doubt, many, many due diligence meetings or due diligence calls with investors. I wanted to ask you what do you think is the question that investors should be focusing on when they talk to someone like you, but may not be focusing on? What are the things that they’re not asking you that you think they really should be asking? 


 I think I’m probably going to answer this the same way that I’ve answered on of the questions earlier, Niels, I apologize. I think investors should try to understand there’s more than one right way to deliver a good investment return from the markets that we trade. So asking what a manager does and writing that down on a form, I don’t think really satisfies the aim of that investor. They should understand why the manager made the choice he made, and assess whether the reason given was a good reason. He could choose to do A or B or C, all very good reasons, but can he explain why he chose A and what would cause him to do things differently in the future? 

So to try and understand the full process behind the decisions that were made when choosing the investment process – how he chose to implement that. To try and understand how he chose to address risks within his business, within his investment process rather than perhaps what some first stage filtering of allocators, where they’re collecting data, and collecting data and asking whether you do A or whether you do B or whether you do C. I think this doesn’t really help you understand what is the driver of those returns of the investment. 


Now let’s jump to the last section today, which is about getting to know you better, Andrew. I call it General and Fun and we’ll see how we go. The first one is probably more in the general thing. In your definition, in your mind, what does it take to become a great trader or great fund manager? 


 Perseverance. It takes a very long time to build a business, an investment management business. One clearly needs a wide range of skills within the team, a lot of luck, a lot of perseverance and one should not overestimate quite how long things take, and things outside one’s control. What is the quotation? “People underestimate what they can achieve in 10 years and overestimate what they can achieve in 2.” 

In this business there are so many event risks outside one’s control that affect the rate at which one can be successful in the business. A basic grounding in statistics, in computer science, in quantitative theory is a given, but it’s in launching an investment management business could take three years longer just because of something that happens tomorrow in the market because markets are frozen for three years. One needs to keep going and maintaining the team and the business and show that one can adapt and evolve and still deliver good returns through those changing market conditions. 


Sure. Earlier today a number of times, in fact, you have made it very clear what you think people should ask and there’s one particular word that you always mention so I’m going to make sure that I don’t miss that, so I’m going to ask you why do you do what you do today? I don’t mean the technical part; I mean you as a person. 


 It’s a fantastic challenge. As I said when I started, at heart I’m an engineer. I enjoy building things, continuing to improve and adapt the thing that we’ve built as we get more feedback on how it performs and we learn more about the problem itself and then we can improve the machine, if you will. So it’s a fantastic challenge. One of the things that makes finance so challenging is that it’s continually evolving and changing as the global economic market evolves but it’s also a competition. 

If I extract that risk premier from the market today, can I do it more efficiently than you can, and if I do, then my returns are going to be of a much higher quality. So it’s a game in competition with other smart people in the industry and I’ve worked with some fantastic people over the last 20 years. It’s a great industry in which to work and it’s a lot of fun. It’s hard, it is hard, and for people who see the world as black and white, as this is the right answer, that is the wrong answer, it’s a particularly hard business because the logic is fuzzy and one needs to live in a world of this isn’t a proper answer or solution to the problem today, and will iterate and improve. If one can deal with that, then it’s a fantastically satisfying business. 


Sure, great. If you’re going to recommend a book to someone about trading, let’s start with that, something where you think by reading this people will improve and learn, is there any book that you can think about that you would recommend? 


Well I thought you might ask this, Niels, and I thought long and hard to way back when I moved into finance, and I though what book was relevant then to me and what book is still relevant today and actually I came up with two. One was the extraordinarily popular Delusions and the Madness of Crowds, written by Charles Mackay in 1841, a Scottish journalist. He covers the bubbles from the 1600s and 1700s associated bubbles, mania, and so forth. The irrationality of human behavior that when one is innovate of the time, for most people they don’t see it. That continues to appear throughout human behavior and financial markets, and we’ve seen that relatively recently. 

So that was an observation of. The second book was Winner’s Curse by Richard Thaler. He looks at the fact that there are biases everywhere and people’s behavior and the fact that in horse betting the long odds are more favored at the end of the day as people are trying to recoup their losses before they go home. You’d think that’s the irrational unsophisticated world. 

At Deutsche Bank in New York, we launched something called economic derivatives. They were derivative bets on non-farm payrolls and other numbers, where people could, rather than using futures markets or similar, could actually take discreet bets with odds on non-farm payroll outcome. It might not surprise you to know that the most popular bets were the extreme bets. The number comes in very different from expected therefore the best risk adjusted return was to buy consensus because consensus is underpriced just like buying the favorite in a horse race is often a very good strategy. So I think those two books, they were relevant in the early 90s and I think they’re still relevant today and that human behavior persists through all markets. 


 Sure. Is there another book, but not so much to do with trading and stuff like that, but just a book in general that you read that really had a big impact on maybe the way you see the world of an entrepreneur, or personal development, or something that just had a big impact on you? You seem to have read a lot of books. 


Well I picked a climbing book. This was written by guy called Reinhold Messner, he was the first mountaineer to climb all 14 8,000 meter peaks. This is back in the mid-80s. Climbing a mountain, the guy who got to the top gets his name on the record books, but he had a team of people behind him. He had a team, whether they were carrying the oxygen, the porters on the tents, he had to manage and hire and coordinate that team. Mountain climbing is management of risk: risk you know about, management of resources and even risks that are things that could happen and things you’re not expecting and there’s an element of looking out as well. In my early days in finance I was still able to spend time climbing mountains and that was a book that I remember as being quite influential for me. 


 Sure, sure. May I ask you, Andrew, do you have children? 


 I do, I have a son and a daughter. 


Ok. If you could just pick one of your own skills to pass on to your children, what would that be and why? 


 A skill? That’s hard, isn’t it? A single skill? 


A single skill, maybe two, if you have two. 


Andrew:  Maybe it’s not a skill. I would say more of an outlook on life: optimism, and persistence. It never fails to amaze me what can be achieved by a man who sees possibilities and just keeps going and doesn’t gives up. I guess the lists goes on. I think they’re two pretty good attitudes and attributes for outlook on life. 


Sure, absolutely. Now, just a couple of questions left, Andrew. I wanted to ask you about a fun fact about yourself, something that people who know you might not even know about you, is there any secret talent or something else that you wish to share? 


 I think it’s already come through, my passion for flying. I many years ago I got my private pilot’s license down at Biggin Hill, the little airfield on the outskirts of London and I’ve flown widely when we lived in the United States. I few around many parts of the US and North America and Canada and the United States and also around Europe. I have an instrument rating, which means I can fly in cloud and a great trick for me is to jump in the plane and off down to Geneva and go and hit the Alps. Within Europe, increasingly with the security system we have it’s fantastic to be under one’s own steam and from door to door. It’s far, far faster than commercial. 


Sure, what kind of plane to you fly? 


 I fly just a single engine Piper, so it’s like a Ford – a standard family car with wings. 


Fair enough, fair enough. I said earlier or I asked you earlier about what are the questions that investors fail to ask, so I’m going to turn it on myself at the very end here and just ask you is there anything that I’ve missed today, and if so what should I have asked you that I didn’t? 


I think you’ve been very thorough, Niels. I can’t think of anything. It’s been a pleasure to talk to you. Thank you for your questions. I hope… that was certainly interesting for me and I hope that it was interesting to your listeners, and useful to people. I think you’ve been very fair, thank you. 


I appreciate that and thanks again, very much, for your time. Before we end our conversation, just sort of looking into the future a little bit, what do you see for Cambridge Capital Management and the managed futures industry as a whole? You’ve obviously looked at it for a number of years. If you’re just looking a little bit forward, what do you see for both yourself and the industry? 


 For us, as you said, we’ve just passed our three-year track record which is important for many people. The markets have been challenging for the space and we’ve certainly found that so. We fortunately have had good numbers this last year in this. We’ve got, since we started talking to the institution allocators in earnest early this year, we have some very productive conversations ongoing and we hope to be launching an on-shore with on-shore US, off-shore fund just in the new year. 

In terms of the space, something I said before, the challenges for this space to embrace the need to explain what they do, to an audience which can have a wide variety of experience, some which are familiar with more sophisticated and some with less sophisticated techniques, but embrace the audience, educate, teach people, explain what they do, why they do, demonstrate skill and consider also an audience perhaps outside the traditional CTA space. Managed futures is a process that happens to trade futures, but is an investment process and I think that is applicable to a wider range of investors and that is the challenge for us to bring new investors into this space. 


Sure, absolutely. As we wrap up I want to make sure that I thank the Eurex Exchange, who actually sponsored today’s episode and as many of you listeners know, with all the talk about central banks starting to increase short term rates, the Eurex Exchange would certainly be one place to go and hedge your portfolio risk. Before we finish completely, tell me, Andrew, what’s the best place to go and find out more about Cambridge Capital if people want to dig a little bit deeper? 


 Well I’d encourage everybody to go and look at our website. The website address is or you can type that into Google and it will come up at the top there. I hope everybody has enjoyed the call today and find the information useful. If you wish to contact us, we’d be delighted to talk to you. Contact us on the website. 


 Great, Andrew, thanks very much for this. This has indeed been a great conversation. I really appreciate your transparency and your willingness to share your insights and views and your strategy and your firm, and also your journey, which was very different the way we started out today. Of course the listeners can also go to the usual place which is the show notes for the TOPTRADERSUNPLUGGED.COM website and find much more out about today’s conversation. I hope that we will be able to connect at a later date, Andrew, and find out more about the great work you’ve done. In the meantime, and all I want to do at the end is wish you and your team the very best. 


Thank you Niels. 


 Excellent, take care. 


Thank you, Bye. 


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.