What Consitutes Acceptable System Performance?

Discussion in 'Trading' started by Roscoe, Jan 19, 2004.

  1. Roscoe


    Would anyone care to share their views on what constitutes acceptable system performance: what baseline values they consider would make a system a candidate to move from testing into real trading?

    (If there is interest in this thread we may need to break it into Position trading and Day trading.)

    I offer the following for consideration.

    Performance Statistics for Position trading:
      Statistic                  Value   Description of Stat
      Mathematical Expectancy    > 0.6  $ return for $ risked
      Opportunity * Expectancy   > 2.0  Opportunity = Trades per Year
      Win/Loss Ratio             > 2.0  [ AvgWin / AvgLoss ]
      Profit Factor              > 3.0
      Annual Return on Account   > 50%  where Account = [(Margin*4)+(MDD*2)]
      Net Profit to MDD Ratio    > 3.0
      Percentage DrawDown        < 30%
      Average Trade $$           > $500
    I understand that there are many performance statistics that can be generated.

    What I am trying to determine is:
    * what statistics are truly useful; and
    * what base values are considered acceptable.
  2. Hello Roscoe:
    This is an interesting and very important thread to open. In my opinion, there are a couple of indications that are important, but almost never appear in system test reports as follows;
    1. Max number of consecutive gains and losses
    2. Max days to recover from largest drawdown
    3. Flat time (average time between signals).
    Most traders do not fully understand the "emotional" capital it takes to stay committed to a system in the face of a drawdown period. These indications can provide the trader with some expectation of what it will be like to really use a system over the long run.
    I look forward to reading the comments. Best Regards, Steve46
  3. Other important things would be...
    How many trades has the system done?
    From what date to what date?
    How was the system developed?
    How much optimization was involved (don't come back with "none")?
    Is this your system or someone else's?
    Does it trade only 1 security and if so what happens if you were to trade another security in the same class?

    Well, this will be enough for now.
  4. Roscoe


  5. Roscoe:
    Your suggestions are just fine, if your audience has the background necessary to interprete them. What I am trying to get across is that new traders or those just beginning to trade a system, are unlikely to know what it "feels like" to trade your system, and are likely to gloss over or be unaware of what the "ratio of MCW to MCL" really means. Looking at the equity curve is more of the same. Sure you can get the data from the curve, but its impact is unmistakable if it is stated right in the results. By the way, one of the marketing tricks that is often used to minimize a significant drawdown is to display the graph of the equity curve at a specific resolution and scale. Steve46
  6. Roscoe


    Hi Steve,

    Understood, and an appropriate post too, thanks.

    To expand my initial post: I am trying to determine which of my own MarketSystems are ready for trading, so I do not have to contend with marketing hype (unless I am trying to fool myself :D ).

    One statistic that I have not mentioned is the Sharpe Ratio, primarily because I have little understanding of it and it's uses. I gather that it relates in some way to the equity curve?
  7. rickty



    The Sharpe ratio is defined to be the quotient of the excess gain (above the risk free rate) divided by the standard deviation of the gains. Higher Sharpe Ratios are seen as better, so that implies a lower standard deviation of the gains. which in the limit means that each trade produces the same gain. If one graphs an equity curve with % gain (from some starting capital value) on the vertical axis versus trade number on the horizontal axis, trades that produce the same % gain will plot out a straight line. Consequently, the higher the Sharpe Ratio the greater the linearity of the equity curve.

    I hope this helps.