Pairs Trading

Discussion in 'Index Futures' started by, Jan 22, 2013.


    The best trading strategy in index futures is "Pairs Trading".

    I have done Pair trading in Index futures like DJIA(YM), S&P500(ES) and NASDAQ100(NQ) for the last 4 years. The best pair or highly correlated/cointegrated pair is YM & ES, but the best "Pair" for the purpose of day trading/short term trading is NQ & ES.

    If you are looking the ratio graph (NQ/ES), one can see there's an upward movement (or we can say there's an upward bias, if we are claiming perfect correlation).
    So what i mean to say is if you are trading in the said pair keep in mind the fact that "NQ will slightly move upward".

    Now I'm thinking to test this strategy in Stocks. Though it's a difficult task (as corporate news may cause sudden jump in individual stocks), I will go on with the same strategy by mixing some fundamentals & technicals.
  2. None of those are pairs in the sense of a pairs trade but are only in the sense that they are mostly highly correlated.

    The ratio of NQ to ES is meaningless.

    There are only about five steps left to derive a useful numerical calculation but at least when looking at the graph of it you saw the slight move upward create the waste to zero in the leveraged instruments, and that's about all that that ratio will tell you so I hope you aren't trading any technical analysis with it.

    For there to be a pairs trade there must nearly perfect negative correlation or a condition in the market that creates this for long enough to backtest and optimize pairs entries. Most of the time either the condition that creates this doesn't exist for long enough, or the instruments themselves are meant to be perfectly negatively correlated.

    The 1:1 indexes do not provide info with regard to the overbought or oversold levels since there is no normalized volatility you'd have if you actually used instruments with high negative correlation (<-0.9).

    Dear bwolinsky,

    First of all thanks for the reply.

    In pair trading one is looking whether the instruments are co-integrated or not. A pair is said to be co-integrated if there occurs “mean reversion”.

    There is nothing to do with correlation/whether both instruments are from same sector/ technical analysis/whether ‘over bought’ or ‘over sold’.

    It’s up to the Trader for his satisfaction mixing some technicals and fundamentals with it.

    I agree with your point a particular pair doesn’t exist for long.
  4. Your pair trading strategy works because market is going up. If it goes down, you will lose money. In other word, you strategy is useless. Why do you bother with pair trading? Just buy es or nq when it goes up and visa versa.

  5. (LOL! Co-integration will always exist, so this loose definition is not a pairs trade).

    The point is not whether they are co-integrated but maybe how much. No matter what co-integration will always exist wherever there is financial data, so this does not meet my criterion of a pair trade.

    They must be negatively correlated by -0.9 to -0.99 to qualify as a pair trade.

    With regard to overbought or oversold, that is how to use the normalized volatility series in a strategy.

    There is no regard to fundamentals possibly other than to sector.

    Right, and then these people expect every anomalous pair related to the other displaying the standard deviations away from the mean commonly found in financial data series but not normalized into values that take on completely independent analysis when looked at by indicators or strategy code where each value is completely unique to each model data series.

    That's a pairs trade. All this other nonsense about pair trading being anything other than this bothers me, but, there it is, nearly all of the pair trading discussed in media does not meet the quantitative value I use to determine if there exists a unique co-interation where correlation is less than -0.9.


    Dear bwolinsky,

    Can you please give an example for the below quote from your reply?

    “They must be negatively correlated by -0.9 to -0.99 to qualify as a pair trade”

    Suppose A&B are two scrips with correlation -0.95, then how you will execute the trade (i.e.; how you will take decision on Buy/Sell)

    Thanks in advance….
  8. The ratios smoothed period moving average should be re-calculated to eliminate non-normalized volatility.

    This value plotted high to low will be all numbers. Against daily data, the best option is to buy on the open with autostops whenever you find the most consistent normalized threshold levels that produce robust results. Optimization should become unnecessary on pairs like QID and QLD, which is primarily where all of the profits my business has produced have come from.
  9. This is very interesting. Which one should be better to long both QID/QLD or short both?

  10. I just finished back testing about this pair. The result is amazing. I really appreciate.
    #10     Feb 8, 2013