If your trading strategy and back-test rely on time, is that fishy?

Discussion in 'Strategy Development' started by mizhael, May 12, 2010.

  1. Hi all,

    Let's say my trading strategy and backtest rely on hitting the price of "open" and certain times such as "11:27 am", etc...

    Is that a fishy thing and may not be reliable?

    Can VWAP etc. help me in these cases?

    Thanks!
     
  2. depends how big your sample size is. markets act differently at different times of the day on average. im sure theres ways to take advantage of it.
     
  3. Opening up your decision making to a parameter like time with a wide variation leads to random optimizations. Stick to price-specific events like support and resistance. Things happen when they happen. You'll drive yourself nuts in backtesting by effectively trying to find out statistically when they should occur. As to VWAP, in my experience it has no value. Anything simple doesn't work, or we'd all be rich and they'd all be broke. Only extremely subtle and illogical things work.
     
  4. I think as long as you have general ideas of timeframes of the day and not specific numbers, there isn't a problem with overfitting.

    Awhile back I made this chart - you can see how the average volume in stocks changes throughout the day:

    http://www.hrtco.com/volumepatterns.jpg

    I noted some times that had spikes in volume. I think the consensus was that the 3:40 and 3:50 spikes were due to imbalance data being published. 10AM was because there's often numbers released at that time?

    The horizontal line is where volume would be if it were evenly distributed throughout the day. Of course every day is different but you might parameterize 3 or so timeframes of the day,
     
  5. VWAP is exactly the opposite of hitting a given price at a certain time. It is an average, So I basically do not see the point of the question.

    Timing means finding the proper time to act. If your system already defines the time then it is an arbitrary selection and it won't work in the future. If your system finds the time then it is a biased selection you have to analyze and determine it has predictive power. No easy, but I think the idea is good.
     
  6. It makes good sense to trade by time analysis....

    for liquidations, entries, exits..

    Sometimes, time is an indicator before price....

    Especially when you combine strategies or events....

    At a high frequency, sometimes time is your only indicator/tool to trigger the next trading event.

    Example

    Strategy 1 ---- a trade did not take place between event 1 and event 2

    Strategy 2 ---- 100 calls were purchased in a 1 minute period, but the price stayed the same..

    Strategy 3 ----> takes the time analysis into consideration that the trade was accelerated, but price stayed the same.... price based strategies would miss this opportunity.... the strategy could hit the bid here and chase it up somewhat... even though the initial price did not move and price based strategies would have missed this trade.

    By finding a series of these events in the Market you can likely structure returns in a unique, controlled way. And they aren't just volume or price based. ... its Market Microstructure.

    This is one of the principles of complex event processing...using time as a source of information and detangling that from price elements.