How do you judge systems in real-trading and compare systems?

Discussion in 'Strategy Development' started by mizhael, Apr 30, 2010.

  1. Hi all,

    I have the following problems that I need your diagnosis.

    I have three daily systems that are running real.

    From paper trading, I know the Sharpe ratios are A>B>C.

    However after real-trading for a few months, the results are mixed.

    The problems are:

    1. because these are long/short systems, I cannot compare them.

    If I compute the Sharpe ratio in the past few months, the number of data-sample is just too small and the real-trading history too short.

    If I want to compare which made the most money during the past few months, I don't have a common basis to compare - because their notional size (gross and net) are very different. The only thing I know is that on the inception day, the three systems had about 0.16 million, 0.28 million and 0.21 million in gross notional amount. And their cumulative PnLs are different too. Is there a way to really compare the systems other than the long-term historical Sharpe ratios?

    I ask this because I want to upgrade to new systems, but there is no way to judge if the new system will make more money or be better. The only number that I can use as a guideline is the long term historical Sharpe ratio. I am paper trading a few systems for a while, but am still unable to compare the systems.

    2. How do I tell if a system is good or bad in real-trading? I only know that the systems are making mild profits, but the results are kindof disappointing...

    Please share with me your experiences... thanks so much!
  2. Anybody?

    This is an important issue but it troubled me for so long:

    How would you compare the real-performance of trading systems?

    Moreover, let's say you run 3 systems in real-trading,

    and run another 10 candidate systems in real-paper-trading (the only differences are slippage when doing the trades) simultaneously,

    is there a way to decide which one from the candidate pool should be put on to replace the systems that are currently running?
  3. Comparison can only be made wrt to an objective function. What is your objective trading? If it is low variability of returns, then you use the Sharpe ratio. If it is absolute returns, then you use the ROR.

    You can also have more advanced objectives, like the highest return with the lowest time invested in the markets. It is up to you. Asking how to compare systems without stating your trading objective is like asking which woman to get married to, without disclosing your goals in life. Is it sex, money, family, all of them?