Pair Trading Strategy Journal

Discussion in 'Journals' started by jonnysharp, Aug 18, 2008.

  1. coreed

    coreed

    Manually - always. For the following reasons:

    Pairs signals have to be filtered. Is the divergence statistical or have the fundamentals changed the relationship between the two stocks?

    Can the short stock be borrowed?

    As far as liquidity take/make - 90% make(BATS)
    If liquidity is low I will either use BYX who rebate takers, SMART Route it or pay up to get in the trade.
    Paying up for the take is a lot cheaper than seeing the stock run away from you. It pays not to be stubborn in this game.
     
    #2491     May 8, 2011
  2. BD1

    BD1

    For those of you interested in Pairtradefinder. The newest version (3) was released today and has added co-integration. This should produce better results vs the prior version which was based on correlation.
     
    #2492     May 17, 2011
  3. luxor

    luxor

    Would you ignore correlation and just look at high cointegration?
     
    #2493     May 17, 2011
  4. BD1

    BD1

    Yes. According to PTF, cointegration doubles the average profit per pair trade.

    I have started backtesting with the new version and I do see that many of my best actual pairs trades have higher cointegration value vs my losing pairs. This is just a quick reaction and I have more analysis to do.

    Here is what PTF says about cointegration and mean reversion (and there is lots about it if you google it):
    "Co-integration(CI) is similar to
    correlation(CR) in that it measures a
    relationship between 2 stocks, or etf's,
    options, futures, currencies.

    Why CI is better the CR is because
    CI measures the tendency for a pair
    to revert back to it's mean, where CR
    only measures the similarity in a relationship
    between the 2 instruments.

    To use a simple analogy to illustrate this
    point, consider this;

    A man and a women are walking down
    the road together, this is correlation,
    however the women stops to look in
    a shop window and the man keeps walking,
    then the women runs to catch up to the
    man, co-integration measures would
    recognize this relationship has a
    tendency to return to the mean(average),
    whereas in this case correlation would
    have broke down because the women stopped."
     
    #2494     May 17, 2011
  5. trom

    trom

    Thanks for the update. How do we download v3.0? I only see a link to 2.99a...
     
    #2495     May 17, 2011
  6. You will have to buy the upgrade, must have gotten an email. I am rerunning my pairs. I used to manually sort through ratio charts to pick "heartbeat" patterns. Using high cointegration does that for you it seems. i am happy with what i see.

    I still think this is 'a' tool. you still have to do your research, watch and search for news and make a judgement on short term movement.
     
    #2496     May 17, 2011
  7. luxor

    luxor

    Does anyone know how PTF calculates cointegration?
    Is it accurate? I've compared a few of my pairs against the figures at catalystcorner.com and the cointegration numbers can be very different. A high cointegration in PTF shows up as low in catalystcorner and vice versa.
     
    #2497     May 18, 2011
  8. No idea, although you would think it will depend on time period you are looking at. And who says catalyst corner got it right. Still need to see the chart and understand the relationship. Backtesting does show it gets better with high co-integration. don't know if same magnitude with increasing correlation pairs...maybe i will run a comparison..profitability at 80, 90 and 99 % correlation and cointegration..
     
    #2498     May 18, 2011
  9. DBS67

    DBS67

    Cointegration cannot usefully be reduced to a single number. The PTF promos said it would be "dumbed down" but oversimplifying runs the risk of misleading.
     
    #2499     May 19, 2011
  10. luxor

    luxor

    #2500     May 19, 2011