What am I missing?

Discussion in 'Strategy Development' started by doncarlosx, Apr 8, 2008.

  1. What is a statistically significant number of trades?

    Strategy #4 rolls out of the ol' noodle yesterday evening.

    It backtests for +1600ish net over the past four days trading a single NQ. My software is not designed for back testing so I'm curious - how often do failing systems have short periods of such crazy gain?

    I know it can't be this easy, I just can't figure out what I'm missing.
  2. Boib


    Trade it with cash tommorrow; then you'll know.
  3. only 4 days? I've seen long term losing systems not have a losing day for several MONTHS trading e-minis averaging 3 to 4 trades a day.... Even a bad system can have streaks of greatness
  4. bespoke


    Don't be fooled by randomness.
  5. Depends....

    +1600 on a single lot (100 shares) aint' much, depending on the condition (trade frequency, # of symbols traded). Especially with this volatility.

    4 days of testing??? ... May work but as a system developer, I wouldn't be able to trade it.
  6. It's on one car of NQM8, so just one symbol. It definitely benefits from chop and suffers from big trends.

    What is the relative importance of days vs. number of trades? I know at first 4 days screams "statistically insignificant" but it was over 183 trades, which seems like a lot.

    Is this a long term losing system or a short term winning system and then conditions changed? No smartassery intended here; I'm just genuinely curious if there is any reason not to trade a strategy as long as it is working and it includes stops?

    I can't. My platform can't automate trades and I'm the worst discretionary trader I've ever met. It is up 16 points though...

    ..on 47 trades. x_x

    Maybe I'll name it the butter strategy for its magnificent churn.

    Net profit, but it really should have just said "Short 08:30, Cover 10:50"

    Thanks for the feedback.
  7. Markets change. You can be trading 1 trade a day or executing 100 trades a day. Which ever the case, your sample size is limited within the tendency of "that day".

    40 trades a day on NQ? Hrmmm... Are you trading a counter trend model using 1 min. (tick) data? 70-80 %Profitability? Risk/Reward less than 1?

    I am sensing...

    A system that's tested incorrectly... are you using Limit Orders for the testing on a static TOS data? (Tick data or 1-min. bar data) Can you confirm that ALL orders will be FILLED if this system is automated? (Let's take out latency for now) I am sensing that FIFO will kill this system.