My backtested system consistently beats the S&P500

Discussion in 'Professional Trading' started by StockTrader1985, May 8, 2007.

  1. "My question is, how do I verify that this backtest is legitimate? Are there statistical tests that I can run to ensure that the system's success would also apply to today's markets?"


    "And lastly, if I am confident in my system..."

    #31     May 9, 2007
  2. stk trader how much are you asking for your system ? i take it still beating the pants off this lame market ? jake
    #32     May 9, 2007
  3. There's nothing wrong with using S&P as a benchmark, based on what he mentions.

    Institutionally... he needs to be outperforming the S&P every year. 60+% DD is way too high. Your datasets are way too small. The std. dev/volatility of the performance is way too high. Finally, it'll get killed with trading costs...

    Anyways... it's not a marketable system.

    Good luck to the OP.

    PS. The problem isn't about whether the system is robust or tradable. It's just not a good system to start off with... Based on these replies on "viability", I wonder what kind of systems the people replying are developing...

    :confused: :confused: :confused:

    #33     May 9, 2007
  4. asap


    "My question is, how do I verify that this backtest is legitimate? Are there statistical tests that I can run to ensure that the system's success would also apply to today's markets?"

    you're looking for stochastic modeling such as the monte carlo method, that encompasses a technique of statistical sampling employed to approximate solutions to quantitative problems.
    #34     May 9, 2007
  5. That had nothing to do with the reply to the post...
    #35     May 9, 2007
  6. "However, I've always been under the impression that slippage doesn't make a huge deal when backtesting equities."

    Wow, just wow. :eek:
    #36     May 9, 2007
  7. Well I'm going off reported financial information through Bloomberg. I guess the data could be flawed

    well the way I do it is I assume all my stocks are equally valued. I don't know what the technical term for that is, but i choose 100 stocks that fit my criteria based all stocks' previous year's Q4 financial reports. Then, I see how all those 100 stocks performed over the year, and then find the mean of the returns

    That's why I don't think slippage would be a huge deal. Tomorrow, I'll run through the backtest but I'll get the correct bid/ask prices for all the trades, that should take care of slippage. then I'll take commission into account. However, I was doing some rough calculations today and I don't think commissions would do much harm to my overall returns

    Haven't thought about it yet. perhaps if there were some bids? :p (don't take this seriously please)

    well, I think you are a little confused, because I don't see how you could deduct that this system isn't tradable and that it's "not good". The stdev of the annual returns is only slightly higher than the stdev of the S&P over the same years. I beat the s&p virtually every year. The DD isn't ever 60%, don't know how you came up with that :confused:

    Like I said, I'll run through the numbers tomorrow taking slippage and commission into account. I still don't think it would be a huge deal, since I'm placing end of day trades, holding my trades throughout the course of 1 year (so the number of trades isn't very high)
    #37     May 10, 2007
  8. BJL


    Sure hope you do not use Q4 information to decide which stocks to buy at the beginning of the year. That earnings information is generally not released until some weeks/months later.
    #38     May 10, 2007
  9. Large cap is good.

    This should get you started on alpha.

    If you can show statistically risk adjusted outperformance (edit: in a diversified portfolio) over a reasonable period you should get someones attention.

    Here's some standard portfolio statistics for a mutual fund -

    Obviously very good performance.

    Good luck.

    #39     May 10, 2007
  10. My bad. The DD of 60% was incorrect. 1998 and 2002 was underperforming and took 2002 and 2003 incorrectly... though...

    Shouldn't the Std. Dev. be lower???? From my understanding...

    Lower the Std. Dev., lower the volatility...

    Am I missing something?

    PS. I still think your model is useless. Please correct me.
    #40     May 10, 2007