Intraday trading

Discussion in 'Trading' started by gifropan, Nov 4, 2007.

  1. I have been working on developing intraday trading strategies for some time. One thing I have always had to contend with is the problem of whipsaws in trend following strategies which generate too many trades and accumulate losses without the market having made any significant move either up or down. To improve the strategies I have often tried to define a filter which avoided certain trades. Where as using filters does reduce the number of trades it does not necessarily improve trade selection. One recent idea I have had is to use a Time Of Day (TOD) filter rather than a technical filter like an envelop or something similar. By TOD I mean restrict the trading to times of day which produce more reliable signals for the particular model in use. The assumption being that if one restricts the trading to the first two hours or the last two hours of the day perhaps there is less randomness in the market.

    Any opinions would be most welcomed.
     
  2. Is there evidence that the first two and the last two hours of trading are less noisy?
     
  3. I think there is empirical evidence that this might be the case. I think it is a fact that the near opening and near the close the volumes are higher and therefore one could imply that the market could be more likely to trend than in periods with little volume. Perhaps someone with more knowledge and experience can enlighten us.
     
  4. Hi Gifropan,

    Good idea.

    Filters you said they reduce number of trades and not neccessarily give better results.

    YET me and many others still use them. Why?

    Because its tightens the distribution of the possible returns. And that's something very important in my opinion.

    Another way of doing this and possible increasing win% is to look at culling your universe, maybe only to the top 20/50/100 stocks.

    Nizar.
     
  5. ronblack

    ronblack

    It sounds like a good idea but I have the feeling it is based on recent observations only. What I suggest is analyzing historical tick data to study the distribution of market price changes during hourly intervals and see whether they are any spikes, or the distributions are uniform or of some other shape. Then you'll know whether you are not just fitting your system to recent observation only.

    Ron
     
  6. GaryN

    GaryN

    I have found the ADX to be a very good filter for trend vs. nontrend.
     
  7. lindq

    lindq

    IMHO, 'intraday trend' is a non sequitur.

    For every 'trend' that you think you're seeing, you'll get sucked into 5 trades that toss you a loss. Because typically, unless the market is moving very strongly in your favor, the moment you identify a trend is the moment that smart traders are going to be positioning for a reversal.

    Trading intraday, better that you look for strong divergence and trade the bounce long and short. Then you'll profit from the whipsaws, instead of cursing them.