We’ve lost a lot of money in the last month, and the Boss calls me in for a meeting. I’m afraid he might ask me to cut my position, but he wants me to do a scenario analysis instead.
This is great news. I love doing scenario analysis. You can use scenario analysis to prove absolutely anything.
A scenario analysis spreadsheet, as the name implies, is a spreadsheet that analyzes various economic scenarios. For each scenario listed, there’s a qualitative description (for example, “the economy picks up steam, measures of inflation rise, and the Fed hikes interest rates”) followed by impressively precise predictions for various market variables (“10-year bond yields will sell off to 3.59%, the S&P will climb to 1182 and oil will rise to $72.35 a barrel”). Combine these predictions with your current portfolio positioning, plus any changes that you expect to make in the portfolio, and you get P&L outcomes for each scenario. Assigning probabilities to each scenario then gives you the “expected value” of your portfolio, which in turn leads to the most important statistic of all: how much you expect to get paid.
A well-crafted scenario analysis spreadsheet can be extremely impressive. The sheer quantity of detail crammed into the spreadsheet can easily overwhelm the unwary reader: examinations of every possible state of the world, with associated probabilities, market predictions and P&L projections all to three decimal places or more. But where do all these numbers come from? Easy: from a thorough examination of current and future economic conditions, coupled with a deep understanding of market dynamics.
In other words, I pull them out of a hat.
Let’s be honest. I might put hours or days or even weeks of research into building my scenario analysis spreadsheet, but even with the best intentions in the world, there is no way I or anyone else can predict the market. What’s the probability of a rebound – 30 per cent? 60 per cent? 90 per cent? Who knows? Even if I knew for certain that a rebound was on its way, what would that mean for oil prices? Would they go up? Down? Sideways? Who knows? And even if I knew for certain that oil prices were going up, what would that mean for my portfolio? Will I still be long oil at the time? Will I have cut my position in half? Doubled it? Gone the other way and sold oil short?
It’s the tyranny of efficient markets: anything you know, everybody else knows. So if you’re certain a bounce is on its way, then there’s a fair chance everybody else is equally certain, and current market prices reflect that information. Trying to predict the future is a mug’s game; building a complicated spreadsheet based on these predictions is beyond stupid.
So I fall back on the cardinal rule. I build a spreadsheet that will help me keep my position.
Building a good scenario analysis spreadsheet is an art. A common beginners’ mistake is to paint too rosy a picture: assigning too high a probability to the “good” scenarios, or assuming that the market will react to unexpected news in just the “right” way to help you make money. A moderately competent desk head will see right through that, so you have to be a bit subtler in your approach. One good technique is to shovel all the bad news into a scenario that you know your boss does not believe in. For example, if he’s a peak-oil fanatic, then make sure that your worst P&L occurs in the scenario where oil drops back to $20 per barrel. Another tactic is to choose your market outcomes such that your boss’s favorite trade appears mediocre: your overall spreadsheet will then seem suitably conservative, but you won’t be asked to change your position. And of course you can always tweak the probabilities, fudge the predictions and alter the positions (“expected strategic/opportunistic reallocations”) to ensure your portfolio works like a charm in all states of the world.
Armed with these tricks of the trade, I create my spreadsheet. It takes about an hour to build (15 minutes to generate the numbers, 45 minutes to format them), and it’s a masterpiece. It has every single scenario imaginable, including several that are ludicrous in the extreme. Amazingly, my portfolio seems to make money in almost all the listed scenarios, yet there are no clear instances of bias or manipulation. All the numbers are given to five significant figures, and there’s an abundance of extraneous detail. Important cells (profits) are carefully highlighted, while irrelevant ones (losses) are equally carefully downplayed.
The spreadsheet works like a charm. Confronted by rows and rows of densely packed figures, the Boss’s eyes glaze over. He focuses on a cell that implies that we’ll make a ton of money while taking minimal risk, and that seems to please him; he says to me, “Good work!” Then he does what any good manager does when out of his depth. He changes the subject.
Wednesday, September 30, 2009
Tuesday, September 22, 2009
Blinking Lights
Last month’s investor meeting merely confirmed what I have long known: financial decisions are rarely made based purely on (or even primarily on) financial merit.
I remember learning this lesson at my very first investor meeting, when I was barely six months into the industry. But we have to rewind a bit. What was I, fresh out of university and ignorant as the beasts that perish, doing at an investor meeting? Surely no rational investor would trust an obviously clueless 22-year-old with their hard-earned moolah?
Ah, but investors are not rational. And this particular 22-year-old had a secret weapon up his sleeve: Das Blinkenlights!
It all began when Professor Q, our head quant, came to me with a new model he wanted implemented. The model was of the type called ‘Monte Carlo simulation’ in the finance jargon. You simply did a large number of trials, and took the average value (of some variable of interest) over all these trials. Nothing particularly original or revolutionary, and a piece of cake to implement.
I programmed the model in about 20 minutes, put it into a friendly-looking spreadsheet, and was all set to show it to the powers that be, when inspiration struck. I added a cell that showed the trial number. As the program cycled through the trials, this cell ticked upward rapidly. It was a trivial little addition, and it slowed down the computation considerably, but it had this advantage: it made the spreadsheet look very NASA-as-imagined-by-Hollywood-esque. Positively cool, as a matter of fact.
I demoed the spreadsheet to Prof Q. He liked it. He called the Big Boss over; the Big Boss liked it as well. So did assorted mid-level bosses. None of them would admit it, but I could tell that the thing that impressed them the most was the rapidly flashing trial counter, which fostered the illusion that they could actually see the program crunching its numbers. All bow to the power of Das Blinkenlights!
I got more plaudits and “good jobs” for my 20-minute spreadsheet with its 2-minute trial counter addition, than I did for the many hours of complex coding and tedious debugging that went into creating our core analytics system.
And that’s not all. I was asked to make my Monte Carlo spreadsheet the centerpiece of a presentation to some potential investors. Yes, this same spreadsheet that was hacked together in 20 minutes by a rookie programmer with zero financial expertise, was now being touted as the super-secret, ultra-splendid, and utterly unique core of our trading strategy.
I was a bit hesitant initially, but that only goes to show how young and naïve I was. In retrospect, I don’t know what I was worrying about. In the years following that original meeting, I attended other presentations by other fund managers which made my spreadsheet look like rocket science. Incompetence, it seems, was not confined to Sopwith. The funny thing is, many of these other funds raised billions of dollars; some of them are still going strong.
But I digress. Let’s return to my first investor meeting. I’m afraid it was a bit of an anticlimax. Oh, the investors said all the right things, oohed and aahed at all the right places, asked a few perfunctory questions, and were unfailingly polite. But I got the impression the entire process was just an elaborately choreographed dance: the true investment decision had already been made. Looking back, I can’t say I’m surprised. At the end of the day, you have to trust your fund manager. If you don’t trust him then all the models in the world are no good; if you do trust him then that trust will remain even if his trading tools are a wind-vane and two fridge magnets.
The lesson for fund managers, of course, is that their single most important task is to establish trust. I was to learn this lesson again and again in the ensuing years, in the context of investor relations, in the context of trading desk politics, in the context of risk management, and especially in the context of position sizing. Once you have people’s trust, you can get away with anything.
That wasn’t the only thing I learned that day. The second lesson, and perhaps the reason why trust is so important, is that nobody knows what they’re doing. Sure, there are folks who can talk the talk: academics who spout jargon, traders who flourish track records, investors who brandish due diligence questionnaires. But nobody really has a clue. That’s why they’re willing (and indeed, eager) to be taken in by a dodgy spreadsheet and a few blinking lights. All you have to do is say the magic word – money! – and their cupidity will do the rest.
The third lesson was the simplest but in some ways the most useful: I learned that, no matter what the task, the important thing is to look good while doing it. From that day on, all my programs and spreadsheets have had flashing lights, blinking digit counters, sexy graphs and slick production values. I’ve never looked back.
I remember learning this lesson at my very first investor meeting, when I was barely six months into the industry. But we have to rewind a bit. What was I, fresh out of university and ignorant as the beasts that perish, doing at an investor meeting? Surely no rational investor would trust an obviously clueless 22-year-old with their hard-earned moolah?
Ah, but investors are not rational. And this particular 22-year-old had a secret weapon up his sleeve: Das Blinkenlights!
It all began when Professor Q, our head quant, came to me with a new model he wanted implemented. The model was of the type called ‘Monte Carlo simulation’ in the finance jargon. You simply did a large number of trials, and took the average value (of some variable of interest) over all these trials. Nothing particularly original or revolutionary, and a piece of cake to implement.
I programmed the model in about 20 minutes, put it into a friendly-looking spreadsheet, and was all set to show it to the powers that be, when inspiration struck. I added a cell that showed the trial number. As the program cycled through the trials, this cell ticked upward rapidly. It was a trivial little addition, and it slowed down the computation considerably, but it had this advantage: it made the spreadsheet look very NASA-as-imagined-by-Hollywood-esque. Positively cool, as a matter of fact.
I demoed the spreadsheet to Prof Q. He liked it. He called the Big Boss over; the Big Boss liked it as well. So did assorted mid-level bosses. None of them would admit it, but I could tell that the thing that impressed them the most was the rapidly flashing trial counter, which fostered the illusion that they could actually see the program crunching its numbers. All bow to the power of Das Blinkenlights!
I got more plaudits and “good jobs” for my 20-minute spreadsheet with its 2-minute trial counter addition, than I did for the many hours of complex coding and tedious debugging that went into creating our core analytics system.
And that’s not all. I was asked to make my Monte Carlo spreadsheet the centerpiece of a presentation to some potential investors. Yes, this same spreadsheet that was hacked together in 20 minutes by a rookie programmer with zero financial expertise, was now being touted as the super-secret, ultra-splendid, and utterly unique core of our trading strategy.
I was a bit hesitant initially, but that only goes to show how young and naïve I was. In retrospect, I don’t know what I was worrying about. In the years following that original meeting, I attended other presentations by other fund managers which made my spreadsheet look like rocket science. Incompetence, it seems, was not confined to Sopwith. The funny thing is, many of these other funds raised billions of dollars; some of them are still going strong.
But I digress. Let’s return to my first investor meeting. I’m afraid it was a bit of an anticlimax. Oh, the investors said all the right things, oohed and aahed at all the right places, asked a few perfunctory questions, and were unfailingly polite. But I got the impression the entire process was just an elaborately choreographed dance: the true investment decision had already been made. Looking back, I can’t say I’m surprised. At the end of the day, you have to trust your fund manager. If you don’t trust him then all the models in the world are no good; if you do trust him then that trust will remain even if his trading tools are a wind-vane and two fridge magnets.
The lesson for fund managers, of course, is that their single most important task is to establish trust. I was to learn this lesson again and again in the ensuing years, in the context of investor relations, in the context of trading desk politics, in the context of risk management, and especially in the context of position sizing. Once you have people’s trust, you can get away with anything.
That wasn’t the only thing I learned that day. The second lesson, and perhaps the reason why trust is so important, is that nobody knows what they’re doing. Sure, there are folks who can talk the talk: academics who spout jargon, traders who flourish track records, investors who brandish due diligence questionnaires. But nobody really has a clue. That’s why they’re willing (and indeed, eager) to be taken in by a dodgy spreadsheet and a few blinking lights. All you have to do is say the magic word – money! – and their cupidity will do the rest.
The third lesson was the simplest but in some ways the most useful: I learned that, no matter what the task, the important thing is to look good while doing it. From that day on, all my programs and spreadsheets have had flashing lights, blinking digit counters, sexy graphs and slick production values. I’ve never looked back.
Subscribe to:
Posts (Atom)
Posts
