Pair Trading Indices--Tradable?

Discussion in 'Strategy Development' started by matador04, Apr 25, 2009.

  1. If I want to long ES, short YM (after looking at several factors) and don't want to expose myself to more than X contracts, what is the best way to scale into it at Y units? So for I have this:

    1. ES spread is 860.25/860.50, and YM is 8030/8032 I will put in LMT ES @ 860 and LMT YM @ 8028. Risk is that I will get filled on one and not the other, thus losing on direction.

    2. Set STP orders to buy ES at every 2 points on the upside, and STP orders to sell YM on the downside for every 2 points of ES. Risk is I will get killed in a choppy, sideways market and will have to wait for a move outside of the range.

    3. After every order is filled, I reset the STP order for ES/YM (I got filled on YM at ES 858, ES 856, ES 854, etc.) As ES ticks back up to 856, I buy ES at 856, 858, 860, etc. Then I sell more YM at ES 858, 856, etc. Risk of convergence between ES and YM is ongoing, as I will be on the wrong side of the trade. I will make money if the market trends because I will have not hedged away the upside/downside.

    4. Once I've scaled into the trade, I will exit the entire position once a profit target is reached. Risk is I will never reach that target :p

    Maybe I shouldn't ratchet down/up on buys/sells and just wait for a direction? Although I am trying to profit on divergence...

  2. Syprik



    1. Given the IIT reference as a "gem," sounds like you're a bit green and looking to dive into an extremely crowded index arb sector.
    2. Prepared to compete against an already large and quickly growing DMA customer base co-located @ CME LNET/Jackson direct = no telco exposure?
    3. Have a workable que/spread algorithm edge?

    It's one thing data mining a regression arb, it's a completely diff ball-game to be consistently positioned for the arb, esp if it's high frequency with low-expectation per flip. Que algo on Globex is often a significant lynchpin. Add in other factors such as most of today's ATS DMA's no longer employ risk management platforms in effort to further reduce execution speed.

    Likely more workable arbs (usually=lower frequency) out in ETF and common vs preferred land at the moment. Strongly encourage you don't keep your pairs research limited to the super popular emini's.

    Even if this route never pans out as intended, it's still great research/exposure to grow the mind :) Good luck.
  3. I certainly am green in this space. Thanks for your thoughts and potential roadblocks.

  4. CBuster


    agree with Syprik.

    Your best way forwards will depend on the timeframe / frequency that you want to target. Splitting broadly into 3 categories:

    High Freq (seconds, sub minute timeframes) - ie Looking to "arb" the spread and target a couple of ticks. ES vs YM is not a pure arb of course but can still use the correlation to make markets in theory. However, all of Syprik's comments apply. This is generally the domain of some very big players with heavy IT infrastructure. Very unlikely that you can play in this space.

    Intraday (minutes to hours timeframe) - here you would be looking to take a view on the direction of the pair rather than playing one of the outrights. still not easy but possible , in the sameway as directionally trading the outright is possible. a number of guys in my office do it. at a minimum you should employ an auto-spreader program - an algo which will automatically work a bid or offer on one of the products and then, if filled, automatically get you an execution on the other side. many of the better trading platforms offer these (see TT X-Trader if you are talking futures as an example). this will at least leave you unexposed to the market direction. there are alternatives (manually legging) but i don't recommend it if you are new to the game.

    Hourly / daily and above timeframe - Can theoretically execute manualy although will need to be careful. My preference would still be to use an autospreader program.
  5. Midas


    I agree here. Stick with stocks and etf's.

    I got started by finding stocks and etf's with the highest coorelation and putting on pair trades anticipating a reversion to the mean. Keep it simple.

    This guy is doing something similar and has been keeping a journal on this site.

    He uses this software to screen and backtest various pairs:

    These guys have good resources as well:

    Good luck.