Scoring.RmdTLDR: You win by making as much money as you possibly can – using all means available to you – but you’re penalized for risk.
Earning returns by managing risk with aplomb is part of the bedrock of the core values of the Duke FINTECH Program and the Gothic Hedge Society. We don’t mind taking on risk – so long as we’re appropriately paid for it. The scoring system used by this Competition was written to put that principle into practice.
To see the big picture, let’s start with an example. Suppose that at the end of the Competition we have two students: a reckless gambler named Wild Bill and a shrewd trader named Clever Susan.
Bill puts all of his money into a tiny, risky startup that has just been listed on the NASDAQ. He then forgets about it and lets it ride all semester until the end of the competition, at which time he’s earned an amazing 250% return!!!! He clearly should win the competition, right?
Well… when you look at the value of Wild Bill’s performance, it was all over the map during the competition. As news was released about the startup, Bill was down as much as -300% on some days, and up as high as 375% because the tiny company’s price was fluctuating so much.
That means, that if Wild Bill had had a cash emergency or needed to close his position and withdraw his investment at any time, there’s no telling how much he’d have had available on a given day. Maybe he’d have a lot of cash, maybe not much at all.
In fact, the next day after the competition, Bill’s gains were wiped out and he was down a full -50% from where he started, meaning that in reality, Bill just got lucky that the competition ended when it did. Clearly, there’s more to the story here.
Susan takes a different approach. She creates a well-balanced portfolio of different assets in different market sectors, and at the end of the competition she posts a very healthy 12% return – much less than Wild Bill.
However, the value of Susan’s portfolio grew steadily and predictably through the competition. On her worst day, she had still earned a return of 8%, but upon the whole it never varied too much from 12%.
The Gothic Hedge Society believes it should be Susan because otherwise the Competition turns into a simple gamble on who’s NAV will be the highest on the day the Competition ends.
That poses the following problem when it comes to designing a scoring system: how do we rank trader performance in a way that is:
We’ll need a metric somewhat more sophisticated than simple return to satisfy those requirements, one that takes into account both average return of a trader’s porfolio over time as well as the volatility of those returns.
Fortunately, the Sharpe Ratio is just such a metric. For you, the trader, your Sharpe Ratio is a measure of your risk-adjusted return; in other words, how much money you made vs. the risk that you took on.
There are two main postulates behind the Sharpe ratio that you should think about deeply if you’re new to this. They are:
Posting a high Sharpe ratio during a time period means that your portfolio earned a big return comparable to the volatility of your account’s value during that time – and that’s good.
Before we move on to calculating the Sharpe ratio with concrete numbers, there are a few more concepts you should become familiar with. Read on!
Think about all the investments out there – is there one that will earn you a GAURANTEED return – a sure thing, with zero volatility?
The answer, of course, is “no”, but US Treasury bills (T-Bills) come pretty close for practical purposes.
At any time, just about anybody can simply buy a T-Bill from the US Government or another large sovereign nation. The thinking is that if a big country like the US defaults (doesn’t pay) a T-Bill debt at maturity, then something is so terribly wrong with the economy, the world political situation, etc., that nobody is worried about finance anymore… they’re worried about surviving the space alien invasion, global war, or other major disaster that must have taken place to cause the country to default. Obviously this turn of events would be very bad, but fortunately, the prevailing financial thinking is that the likelihood of such an occurrence is so low that its probability can be assumed to be 0.
In other words, these investments are thought of as risk free.
Holders of these assets receive a return, of course. The exact percentage varies but whatever its value, we refer to it using the symbol \(r_{f}\).
Since they’re ‘risk-free’, traders should just buy government bonds, earn a zero-risk return, and forget about complex strategies, right? Well… you could do that, but because the investment is risk-free and easily available, it doesn’t (historically) pay very much.
The US Department of the Treasury compiles daily risk-free rates and publishes them on their website, where you can go to view & download the rates and learn about the methodology the USDT uses to calculate them.
In the Competition, your return is calculated each day by observing the Net Asset Value (NAV) of your account – the total end-of-day value of everything you own.
“Today’s” NAV is compared to “yesterday’s” to calculate the percent return you realized “today”.
That means that over a period of \(N\) days, we’ll have \(N - 1\) total observations of a portfolio’s return because you can’t measure a return for the very first day – you have nothing to compare it to.
The formula used to calculate the return you earned on your portfolio is the log ratio of \(V_{N}\) – the value of your portfolio on Day \(N\) – divided by \(V_{N-1}\) – the portfolio’s value on the previous day; i.e., \(R_{N} = \log(\frac{V_{N}}{V_{N-1}})\).
In the competition, daily log returns are calculated for every student’s portfolio as soon as the data becomes available for that day (usually after about 6pm EST on a trading day).
TLDR: We use the geometric mean of period-over-period log returns.
Details: Let’s say you invest $100k into an investment portfolio. Each quarter, you measure the return you earned at the quarter’s end with respect to the how much the portfolio was worth at the beginning of the portfolio. You observe the returns +25%, -25%, +25%, -25% for each quarter. In other words, an initial investment of $100k would be worth $125k, $100k, $125k, and $100k at the ends of quarters 1, 2, 3, and 4, respectively.
In this context, when you discuss means of a sample of returns, what you’re really asking is something like , what’s the overall return that characterizes this investment’s performance at any given time over the period?
There are infinite ways of answering this question (the doubters may start by Googling “Harmonic Mean”, or maybe “Lehmer Mean” to start convincing themselves of the truth of this statement).
This vignette focuses on only two means, discussed below.
You might think: “just take the average” of the returns, which works out to 2.5%. But that’s not quite descriptive of the whole picture is it… after all, if you tell a client that you earned an average of 2.5% on their investment of $100k, then they’d naturally expect to see something around $102.5k when they look up the value of the account right?
…but the value of the account is $100k, meaning that you’re probably going to have a client banging your door down wondering where their $2,500 went. The “normal average”, more properly called the arithmetic mean of the returns is therefore misleading in this way!
A better way to assign an overall characteristic (“expected”) return for a time series of returns is to use the geometric mean, calculated by converting all of the returns into factors by adding 1, multiplying them all together, and taking the \(n\)th root, where \(n\) is the number of returns observed in the period. Finally, convert resulting factor back into a percent by subtracting 1.
In symbols, the geometric mean of a series of returns over \(n\) periods can be expressed as \(\left(\prod_{i=1}^{n}R_{i}\right)^\frac{1}{n}\), or, equivelantly: \(\sqrt[n]{R_{1}R_{1}R_{1}...R_{n}}\).
You can check the numbers from the example above to see that the GMRR for the investment is 0, just as it should be.
You’re going to be managing a portfolio – some set of bonds, stocks, currency trades, etc, on which you earn a portfolio return \(R_{p}\). As an investor, you only care about earning a return that is better than the risk-free rate \(r_{f}\) because otherwise, why are you even bothering to invest? Better to just buy government debt.
In other words, what you really care about is your portfolio’s Excess Return over the risk free rate.
Excess Return is defined as the difference between the return your portfolio earned and the return that was available by buying sovereign debt during a certain time period; in other words, \(R_{p} - r_{f}\).
Earlier in this vignette we presented the two postulates that amount to the idea that, for a trader’s account, if we measure the return earned each day with respect to the day before, then volatility = standard_deviation(returns) = risk. Standard deviation of a portfolio’s returns is a concrete value that we can understand and calculate, and we assign to it the symbol \(\sigma_{p}\).
Now we can put it all together and, as William Sharpe did in 1966, write down a metric that goes down when expected return goes down & volatility goes up, and up when expected return goes up and volatility goes down:
\[\begin{align*} {\tt Sharpe\ Ratio} = \frac{R_{p} - r_{f}}{\sigma_{p}} \end{align*}\]
Let’s say you go to a market that is selling “return”, which it quantifies in units of percent “%”. You can buy returns of 5%, 500%, -0.2%, and so on, but you have to pay. But what payment will the market accept?
The idea behind Sharpe’s ratio is: you can buy (or sell) ‘return’, but you have to pay by taking on (or divesting) ‘risk’, of which volatility is a measure.
Look back at the definition of the Sharpe ratio and see that it has units of “% return” in the numerator, and “% risk” in the denominator. That’s not an accident- the Sharpe ratio is, at it’s core, the “price” of an asset or portfolio in terms of its “risk”.
In general, better traders have higher Sharpe ratios because they’re able get the best “price” (in terms of risk) that the market will give in exchange for the returns they earn.
Now you’re ready to move on to calculation to see worked examples showing how these parameters are calculated in the Competition.