About the Sharpe Ratio and Expectancy in the analyze of intraday trading systems

Discussion in 'Risk Management' started by Eddy, Jan 6, 2003.

  1. Eddy


    I am a bit puzzled about the Sharpe ratio calculation and its use.

    Here is the background :
    I am presently backtesting over 4 years some ES trading systems based on intraday bars (typically, 3 minutes one). Til recently, I was focusing mainly on Profit Factor, max Drawdown, consecutive losers, etc, ie the usual performance indicators (as nicely presented by C. Wright in his latest article on "The Science of Strategy Evaluation"). Looking for a single ranking criteria, I was wondering what would be the Sharpe Ratio of my equity curve…. On first sight, the Sharpe Ratio Formula sounds pretty simple :
    Average monthly return (%) minus the risk-free rate divided by the standard deviation of monthly returns
    But, when I started calculating (in Excel) and analyzing the resulting figures, following issues appeared :

    1) First issue :
    The Sharpe ratio seems to be originally intended for daily trading strategies. For example, Tradestation will not report a Sharpe ratio for intraday strategies (and I am not sure, by the way, if this figure is available in Wealth Lab) Anybody knowing WHY this do not appear in TS ?

    2) Second issue :
    The monthly returns (in $) can be directly found from the equity curve. But what about the reference for getting the corresponding percentage ? Is that systematically the equity at the start of the backtest period ?
    If I am trading a fixed number of ES contracts, it looks logical to take the initial available equity as constant reference. But what if I increase the number of lots when I get more equity available ? Does the Sharpe ratio (using a constant reference) still make sense for a fair trading system performance comparison ? In this case, in order to evaluate the % monthly return to be used for the Sharpe ratio calculation, wouldn't it be better (just a first intuition) to "rescale" the reference used for the % calculation each time the number of lots traded in the backtest changed ??

    3) Third issue :
    It concerned the impact of the leverage used in my ES trading equity in term of Sharpe Ratio analysis. For this point, I was pointed to a great article of Bob Fulks who provides quite clear answers.
    Have a look to : http://www.miapavia.com/homes/ik2hlb/sr.htm

    4) Final point
    While searching for alternative single figure to be used to “rank” quickly trading systems, I found this article of Alex Matulich : http://www.unicorn.us.com/trading/expectancy.html
    He proposes in fact a simplified way of calculating the expectancy as originally defined by Van Tharp. Anyone having feedback on using this approach ?

    Thank you for your inputs,

  2. i think the Sharpe Ratio has been discredited because the janus funds had a low sharp ratio and we know what happened to janus investors.
  3. omcate


    I thought only hedge funds or people doing arbitrage care about Sharpe Ratio. Daytraders only care about returns. Any comment ??



  4. LOL... amateur...

    Sharpe Ratio and Expectancy is very important in Money Management.

    How the hell do you think you figure the optimal risk? You have to know the consistancy and expectancy of the trades you make.

    This doesn't only concern system traders. Discretionary traders should also be equally conscious about their risk.

    Unless, you work for a prop. firm who tells you how much contracts(shares) you can trade...
  5. The Sharpe Ratio is only applicable to a totally mechanical trading system... otherwise its a pretty meaningless statistic...
  6. acrary


    TS requires at least 2 years of data before it will compute the Sharpe Ratio. It works fine for me in my testing with intraday data.

    As a % it's: (ending equity - beginning equity) / beginning equity
    for each period.

    The Sharpe ratio is a time independent measure of dispersion of returns. If you allow varying the number of contracts per-trade you'll get a wider dispersion and a lower Sharpe Ratio. For system evaluation purposes it would be better to test using a fixed number of contracts throughout the test.

    There's no need to be locked into the monthly returns either as your intra month equity could vary widely. A simple way to calculate the Sharpe Ratio is to enter all trades in column A of a spreadsheet then apply in column B =average(a1:a99)/stdev(a1:a99) where 99 is the number of trades in the first column.
    It is generally assumed the higher the Sharpe Ratio the better. However, as traders we're supposed to maximumize gains and minimize losses, so the standard deviation could be large if we're trading correctly. A better mesaure is to use only the losing trades in the standard deviation calculation so that you're not penalized for large gains.

    Jack Schwager has a very nice discussion on this topic in his book "Managed Trading - Myths & Truths".

    I think Schwager's RRR is a better indicator of overall quality of a method as it uses drawdown and annualized returns to gauge risk:reward. This is how our trading desk was evaluated.
  7. CalTrader

    CalTrader Guest

    Not really an answer but a philosophy ....

    Any particular indicator that you consider as a candidate for inclusion into your trading methodology needs to have high utility: that is, it must provide demonstratable signal for your trading system. Your definition of "demonstratable" is the value add: If you are not sure why you are using an indicator then you need to revisit and re-test the indicator against your historical trades and within your current model.

    There are a lot of popular indicators which wax and wane in their adoption: some maintain their utulity over time and some are naturally useful. Many times the reason I watch a particular indicator is simply because everyone else does ....
  8. Sharpe ratio also works for me with intraday data. As far as dumping the trades from Tradestation into excel, how do you get rid of the dates in the column in the performance report? The Sharpe ratio, in my view, has to be taken into account along with other performance figures.
  9. acrary


    There's no need to mess with the dates. Just Export the trades using the tab delimited text output and the standard trade format (not advanced). When you import into Excel, just leave the cells as a General column format. You can manipulate the P/L column without having to worry (it'll skip over the blank fields) and calculate everything properly. If you really wanted to delete the dates, time, price, etc. all you'd do is click on the top of the column in Excel and choose edit/delete to remove the column.
  10. Eddy


    Looking at the Sharpe ratio is my first attempt to find an indicator evaluating the CONSISTENCY of the returns when comparing two intradays trading systems having similar global performance (ie similar total PL, % winner, profit factor, etc..)

    In this area, I found the max Drawdown (or max consec. Losers ) to be sometimes misguiding when used as ranking parameter…
    Indeed, the maximum DD is only one aspect of the underwater equity curve. From my very little experience following live a trading system, my biggest concern is to know HOW long it will take til a new equity high is reached (ie I would like to know the typical distribution curve of this information) : I am ready to trade a system having a bigger DD value IF it is associated with shorter recovery time…
    Acrary, thank you for proposing this simplified version of the Sharpe Ratio : looks indeed that taking only losing trade for the std deviation will not hamper system with large gains.
    The RRR looks to be a good compromise in order to merge information coming both from the “shape” of the Drawdown periods and the evolution of the returns… I will investigate that deeper, starting with a reread (from Chapter 6 onwards) of Chande “Beyond TA” book ! By the way, it is nice to see that lot’s of the notions explained there (about equity curve analysis) become now easier to catch and use (compared to the first reading the book, one year ago)
    Anyhow, it is interesting to know RRR was a tool for trader evaluation purposes. I wonder what was considered at your trading desk to be a good RRR figure ?

    Together with RRR, I will focus as well on some volatility measurements of the Drawdown duration and depth. Moreover, I intend to run some Montecarlo simulations in order to investigate the possible range of drawdowns.
    I just finished reading the great free book from Larry Sanders on the possible use of such simulations for evaluating different types risks strategies.. It can be downloaded at : http://www.tradelabstrategies.com/
    This is a great presentation on issues around trading, risk and randomness…

    Finally, i agree one should first focus on developing a robust system using a fixed number of lot, and then try to improve the global performance thru some money management techniques in order to increase “safely” the number of lots traded.

    Happy backtest,

    #10     Jan 7, 2003