how exactly do the index arb'ers do their thing???

Discussion in 'Trading' started by Sky123987, Jan 22, 2008.

  1. You could use any number of products in your basket for the SPX. Obviously the SPY is a great hedge but how much edge do you really think will be in very simple arbs like that?
     
    #21     Jan 23, 2008
  2. taowave

    taowave

    Nope..except in rare occurences
     
    #22     Jan 23, 2008
  3. #23     Jan 23, 2008
  4. I was wondering if you could look at this again nitro. Thanks again for all your help... not only on this post, but on many others
     
    #24     Jan 28, 2008
  5. #25     Jan 29, 2008
  6. Sig

    Sig

    This is an oversimplification, but by definition any portfolio you put together with a Beta of 1 will track the S&P 500. You don't need to hold all 500 stocks or for that matter any of the 500 stocks.
     
    #26     Jun 5, 2016
  7. nitro

    nitro

    A bit late but I totally forgot about it.

    First let me state that, there is a world of difference between trying this today and in 2008. Today, you probably need to be on the microwave network between NY and Chicago to take advantage of this scenario. Or, you need to be really clever and are able to handle small latencies so you don't have to compete with the ELL [extreme low latency] competitors.

    That said, assuming you are either really clever or you have fantastically low latency system,
    what index arbs do is, and probably the most important thing, is to compute the subset of basket of stocks that mathematically track the index the best at any given time. This is not so easy, but there are mathematical tools that help one find a subset that worked well in the recent past. There are several approaches to this, and the more clever you are the better.

    One approach is doing quadratic programming to find the optimal portfolio subset of the basket that composes the index. Another is to do dimensional analysis to see what eigenvector corresponds to the highest eigenvalue. There are still other ways that mix genetic algorithms and or neural nets with the above.

    Needless to say, if you identify those components, and are able to track when the index and components subset has gone out of whack, after commissions you make money more often than not [or you might structure it to take small loses and let it ride when it works.]

    All easy to say. Not so easy to do.
     
    #27     Jun 5, 2016
  8. newwurldmn

    newwurldmn

    Balance sheet and low commissions. I doubt anyone other than the largest dealers can do this now.
     
    #28     Jun 5, 2016