Reader Strategy: SPY/DIA Pairs Trade


This strategy is another example of the importance of the right data, and not confusing the intricacies of indices with the intricacies of ETFs (or futures, or options, or anything else).

The strategy comes from friend of the MarketSci Blog Russ: buy the S&P 500 and short the DJIA 30 at the open if the S&P 500 opens lower than the DJIA 30 (in % terms, relative to the previous close). Short the S&P 500 and buy the DJIA 30 if it opens higher. Exit all positions at the close.


The graph above shows the strategy results applied to the respective indices from early 1998 to present.

Geek note: This test is frictionless (i.e. ignores transaction costs and slippage).  I realize that because we’re trading at the open (and therefore not using leveraged mutual funds) this assumption is unreasonable. I’ll explain in just a moment.

In a frictionless world, this strategy would have been very effective, not necessarily in terms of profits (15.8% annualized return), but in terms of risk-adjusted profits (only 4.2% annualized std. dev.) Notice that this is a simple pairs trade where the trader is going long and short two different market indices and capturing just the performance difference between the two.

But as we’ve repeated ad infinitum on this blog, intraday index data (as opposed to ETF data for instance) has to be taken with a grain of salt…especially the opening price. I would say that any strategy that relies on intricate differences in intraday price (like this one) is meaningless when tested on the indices themselves.

For example, here is the same strategy applied to ETFs in red, compared to the original impossible-to-mimic index strategy in blue.


Still profitable, but only barely. Total returns were approx. 0.018% gain per day, meaning friction in excess of about $3.51 per $10,000 traded brought profit to zero.

Lesson reminded?

Strategies (particularly the shorter-term variety) that are based on small changes in price need to be tested on the actual asset being traded. Anything else is just asking for trouble.

P.S. A big thank you to Russ for asking us to put this one to the test and giving us the idea for this post.

Happy Trading,


To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.

9 Responses to “Reader Strategy: SPY/DIA Pairs Trade”

  1. 1 Jens

    Interesting! I wonder how this would look if you did the same back test with DOW and SP futures? Pretty liquid contracts, should not be a problem getting filled at open.

    • 2 heywally

      That is my question – how would this do with the futures? You could even jump start the open a bit using them.

  2. 3 toptick

    Yes, know your data! Old data in index time series might be missing highs and lows. For the DJIA, old reported daily highs/lows were the average of the component highs/lows, not the index’s intra-day high/low).

    A quick look at SPX or DJIA shows that the “open” is essentially the previous day’s close. (My guess is that the first index computation in a day includes lots of prices for stocks that haven’t had a regular-hours trade yet, so they use the previous close rather than last before-hours trade.) The reported open for these indices in no way approximates a tradable price.

    NDX, COMP, and RUT don’t have these problems. Definitely, test on what you can trade (e.g., ETFs or futures), and be aware of the limitations of your test if you use index data.

  3. 4 eber terandst

    Ah , the (in)famous Open price ! ! !
    Yes, I have seen a number of “brilliant” backtests using this type of data. Certainly the “discoverers” never tried to really trade this . . . or they would have found how they got creamed by the brokers at the Open.
    The availability of sophisticated backtesting platforms has created a large number of opportunities for people to be rapidly separated from their money.
    I applaud this healthy dose of realism ! !

  4. Jens: “I wonder how this would look if you did the same back test with DOW and SP futures? Pretty liquid contracts, should not be a problem getting filled at open.”

    Jens, I used to trade the Dow/S&P futures arb from outside the Dow pit. A simple strategy like this simply won’t work after slippage and commissions. (And you haven’t experienced slippage until you’ve put a market order into the S&P pit. Thank God the eminis have put these corrupt/incompetent brokers out of business!)

    The Dow/S&P spread is actually a very sophisticated trade, because there’s a lot of “dirt” in the mix, beginning with the way the indexes are calculated. The basis is also critical to the trade.

    Moreover, there was the issue of which way to weigh the “extra” dollars. When I traded it was $10 Dow futures against $250 S&P futures. It was almost impossible to get a perfect dollar-for-dollar hedge. So was it better to have nine Dow or ten against three S&Ps? One gave you a little more Dow, the other gave you a little more S&P.

    Yes, there was a correct answer to this question. I will leave it as an exercise for the reader to figure out which way we went. (And we were always overweighted in the same index as a hedge of our hedged position.)

    Finally, with regards to opening cash prices, we never initiated a trade until all 30 Dow stocks were open, so we knew where cash REALLY was. And of course the last ones to open were those with the biggest order imbalances and thus the most likely to have a big impact on the cash when they finally opened.

  5. 6 pairtrading

    You can make that return 10 fold if you only initiate the pair trade when the 2 instruments have meaningfully diverged from one another.

  6. 7 justin

    What was the standard deviation of the ETF returns? I’m wondering if using all these leveraged ETFs that are popping up these days may amplify the returns without additional risk.

    • 8 marketsci

      RE to Justin: very low (0.206% combined daily), but remember that’s frictionless. Employing leverage ETFs would definitely increase that vol (and perhaps the return, but that would need to be tested). But I still don’t think it would be tradeable after accounting for trading frictions. michael

  7. 9 Black Label

    interesting. instead of looking at the movement between cycles, have you looked at using the full moon / new moon as pivot points? for example, if the thrend is up going into the full moon is a reversal likely (vice versa) and same for the new moon?

    thanks for the great ideas.


Leave a Reply

Please log in using one of these methods to post your comment: Logo

You are commenting using your account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s