System results depending on market hours

Discussion in 'Trading' started by tranet, Mar 22, 2018.

  1. tranet

    tranet

    I use a simple system with 2 signal patterns - pullbacks in trend, I tried to backtest it for RTY futures, but it has strange behaviour in January-March 2018. It is highly profitable in time 9.00 - 15.00, but it is absolute loser from 8.30 - 9.00. Same setup, same parameters, stops, targets, but absolute opposite results in first 30 minutes comparing to rest of the day. I know that first minutes could be quite volatile, but this seems too much. What do you think about such a difference

    all--_830-9.jpg
    all--_9-15.jpg
     
  2. kevinkdog

    kevinkdog

    I assume you are meaning Central time, which would put 8:30-9:00 as the first half hour of the stock market day. It could be that with your approach that the "opening" period acts a lot different than rest of day - hard to say why.

    Have you tested it for a longer period of time than just last 2.5 months? Does the same performance trend hold true for 2017 or 2016?

    As an aside, your average trade is about $17. That is kind of low, unless you are using a decent slippage estimate. What are you assuming for slippage of market and stop orders?
     
    Last edited: Mar 22, 2018
  3. tomorton

    tomorton

    First off, this absolutely wasn't a "normal" market period.

    Then again, maybe it is indeed simply the extra volatility during the early market. But if you're looking at the half-hour in question, you should really compare with another half-hour period, not the remainder of the day.


    PS: I've never made an equity curve chart but how come it starts at 0? How can that be?
     
  4. kevinkdog

    kevinkdog

    That looks like a standard Tradestation chart. It really should say "Net Profit/Loss" since it does not include your initial starting equity in it. They decided "Equity" was the proper word, I guess.
     
    tomorton likes this.
  5. lindq

    lindq

    My first question would regard your use of stops in your coding, if any. A higher volatility environment may require that you adjust stops, and/or profit targets.
     
  6. tranet

    tranet

    Answers to previous comments -

    yes it is central time - Chicago CME market. And it was just tested for last 3 months, as RTY was moved back to CME last Summer and in first months volume was low. Now it seems good for trading.

    And yes, it is Tradestation - so equity is counted as Net profit.

    About stops and profits - it uses dynamic stoploss what means that after entry stop is moved one tick below lowest low (pivot) (short just opposite). Profit target is 4-6 RTY points, depends on signal pattern.

    The same system for NQ futures (showing last 15 months) does not show so huge problem in first 30 minutes - the result is 0, but not huge loss.

    nq_full.jpg
    nq_830.jpg
     
  7. kevinkdog

    kevinkdog

    I would look at TF from 2016 and 2017, since RTY was not around for some of that time.
     
  8. i960

    i960

    Or better yet, look at TF all the way back to 2007 and ER2 on CME before that. I would not rely on any system that's only been working off of data since RTY was moved back to CME less than a year ago. Crazy.
     
    truetype and kevinkdog like this.
  9. tranet

    tranet

    Problem is that I do not have TF data, I use mostly tick charts and I used ES and NQ. The system works on various timeframes and markets. ES volatility (in my timeframes) changed about year ago and NQ is much better although more volatile especially in beginning of trading day, anyway I am looking for other choice and I tried to backtest RTY. Tradestation provides only last 6 months of tick data - or you have to have them in my cache - as I have ES/NQ tick data from maybe 2010. I was just curious if anyone else had similar experience with such difference of system results between first hour and rest of the day.
     
  10. kevinkdog

    kevinkdog

    IMO, 6 months is just not long enough to evaluate a strategy.
     
    #10     Mar 23, 2018