Reader Feedback: The Importance of the Right Data
This post comes from Andrey S. of Russia (who was also responsible for the test of Monthly Seasonality in Latin America). Andrey is critiquing a trading strategy presented at another site (which shall remain nameless because I’m a nice guy).
The trading strategy calls for shorting the Dow 30 Index at the open if the opening price is below the previous close, otherwise buying at the open. The original strategy didn’t specify when the position was closed, so I’ve assumed positions are closed at the closing price.
Strategy results from 02/1998 frictionless:
[logarithmically-scaled]
Cha-ching! 69% annualized return – bring it on!
Oh wait, you can’t actually trade the opening price on the Dow 30 Index. Well I’m sure it will work as advertised on the ETF right?
Original strategy results (blue), applied to the DIA ETF (red), and applied to the index but trading the ETF (green):
[logarithmically-scaled]
Back to the drawing board.
This is a reminder of the importance of making sure what you’re studying and what you’re trading make sense together.
Opening prices on indices are notoriously inaccurate (I made a similar but not as blatant mistake myself in a long-ago post), and trading opportunities are often already traded out in the overnight market (which we saw for instance in my posts on East Asia vs US Stock Market Performance).
A giant thank you to Andrey for the post idea. As I’ve written many times before, I love reader feedback that contributes to this collective discussion.
Happy Trading,
ms
To stay up to date with what’s happening at the MarketSci Blog, we recommend subscribing to our RSS Feed or Email Feed.
Filed under: Random Stuff | 13 Comments
ms,
I haven’t really looked at this too closely, so might be way off, but…what if you used say the YM futures contract, and opened/closed positions in the AH market?
Would this then approximate the initial strategy returns [claimed]?
jog on
duc
Quite a post.
That´s the good thing of searching edges among futures contracts: it is a lot of work building the data joining each contract month. Much more harder if edges are being executed through options, have to have the delta and find the strike zone for entering.
keep this grate blog.
Ugh…I made the same mistake recently. I’ll have to re-run the tests on the SPY rather than the SPX.
Excellent post. I’ve done this as well when designing systems to trade the S&P 500. Good stuff until I ran it against something like the SPY. I did, however, take a few of the ideas and adjust the basics for the SPY data (which is filled with errors from some data vendors).
~Shepard
=^.^=
Sad, but very true. I have also had several bad experiences in trading ETFs at the open.
I use TDAmeritrade and several times the execution price was quite different than the one found in popular sites like Yahoo, Google or TC2000. When I asked TDAmeritrade, the invariable answer was that they routed the order to an exchange different from the one quoted in Yahoo, etc. One way or another, the order was always routed to the “wrong” exchange ( wrong from my point of view, of course).
In my opinion, this is chicanery, plain and simple. The practical lesson would be to trade with limit orders, but this a different ball game from using the open.
eb
nice. Thanks for posting this.
Great post. I built a strategy once doing something similar. I looked at which sector ETFs had the best performance the first five-fifteen minutes of trading and the would look for mean-reversion or momentum depending on what the rest of the market was doing (if the market corrects, load up on MR, but if it keeps going, take advantage of the momentum). Anyway, the biggest difficulty (that I didn’t realize until many hours later) was that the five minute ticks that I was using wasn’t catching the right prices. Different ETFs were opening up at different times. Some weren’t actually open 10 minutes after the bell. Suffice it to say, the returns looked amazing and then when I corrected for all those difficulties, they went away.
RE to duc: I would expect the same (bad) result because I think the root of the problem is that the index opening quotes are completely bunk, but I’ll add it to the to-do list to test.
Jog on sir.
What data should one then use for testing out various strategies ?
ETF’s ? Leveraged MF’s ? Futures contracts like ES & YM. Can you let me know the names
of the particular products.
Further, is data on Yahoo finance reliable enough to test strategies ?
RE to piyush: it depends on what you’re trading – test on what you’re trading (or something sufficiently similar). Yahoo data’s reliability also depends on the specific type of asset. Indices, stocks, etfs, etc? Yes, it’s usually appropriate for EOD data.
I was developing strategies for indexes only. So the values for SPY (Open, High, Low, Close) correspond to the US Spot market hours only ? These securities are traded electronically (23-24hours) ? or only when spot market is open ?
thanks :)
RE to piyush: the OHLC values for SPY correspond only to when the US market is open (09:30 – 16:00 ET).
also is there an actively traded ETF for the DJIA