Testing for intra-day pair trades

Discussion in 'Strategy Development' started by OHTodd, Jul 26, 2010.

  1. OHTodd



    I was hoping that someone may have a good suggestion for testing the viability of a pair based on intra-day (say 1-minute closing prices) data.

    I've found pairs that seem to co-integrate well based on ADF results on daily closing prices, but this doesn't seem to translate well to a shorter time frame.

  2. You've answered the question you meant to ask in the first place. If they were cointegrated at your trading time frame what would be the point? The cointegration itself (better, it's lack) is what you want to trade.
  3. OHTodd


    So do you think ADF is an acceptable test for short-term time frames (time series based on one-minute bars)?

    thanks for the input.
  4. One minute close prices are too noisy to use without adjustment. Mostly due to bid/ask bounce and other very short-term artifacts. You'll also have to watch out for Epps effect. The most effective bar length for intraday pair-trading with unadjusted trade prices, balancing noise and epps effect, is between ten and twenty minutes.
  5. The EOD correlation will not translate to an ultra-short time frame of 1min.
  6. Use ADF to generate the framework for a trade; in other words, the structure of a "neutral" portfolio. That doesn't tell you how/when/where to price or place the trade. That pricing aspect is a completely separate matter.