Using stock correlations for day trades

B

boddintrader

Does anyone use correlations between stocks and indexes to day trade? This could be a market index or ETF. This could also be sector index or ETF.

My general trading is mostly to focus on the technical pattern of a stock and use the various inflection points for entries and exit.
If I'm day trading an energy stock, I'll usually look at the market (SPY) and sector (XLE) for help with the general direction. That is, if I'm looking to short the stock, I'll check for market and sector weakness ahead of taking the position. But other then a check for a few down bars, general trend, or an upcoming inflection point, I don't do much with this.

So I've been trying to figure out how I can use the market and sector more to my advantage. The general idea is that markets move sector, and sectors move stocks, but is this true in day trading? Would I find more high-probability opportunities if I were trading stocks with high market and sector correlation?

What are some of the general rules that you use for correlation? What time frame do you use for measuring correlation? How many periods? What's your threshold for "high correlated"?

My research on correlation points to a 20-day correlation of above .80 being "highly correlated." Not sure how others came up with 20 days. TradeStation's correlation indicator has a 14 period default, which doesn't make any sense to me. Not sure if using daily charts for correlation is useful for day trading. I've used MetaStock to measure correlations on 5-minute charts for 78 (1 day) and 234 (3 days) periods. Not sure if that's any better.

I appreciate any thoughts, feedback, research or links to more information.
 
Quick clarification, the hypothesis I'm investigating is whether or not I can use the market or an ETF as a leading indicator for a .15-.30 move in a stock.

I appreciate the feedback and link, but I'm not heading down the pairs trading road on this one.
 
A good thread!

However to day trade, you are more interested in intraday Corelations not daily.

There is high daily corelations between Oil i.e. XOI and OIH with SPY.......at least I think.

There is roughly 60% corelation between Asian markets and S&P500 also.

:D
 
Here are correlations between the SPY and the nine S&P sector ETF's using 5-minute bars for 1, 3 and 5 days.

Ticke 1day 3day 5day
XLK 0.9829 0.9823 0.9736
XLI 0.9821 0.9811 0.9358
XLY 0.9748 0.862 0.9172
XLP 0.9688 0.5208 0.6053
XLV 0.9448 0.955 0.4322
XLE 0.9431 0.4868 0.4359
XLB 0.9281 0.9338 0.951
XLU 0.903 -0.4816 -0.0494
XLF 0.8408 0.946 0.9504

Doesn't look so useful on only one day's data, but we start to see separation on 3- and 5-day's worth of data. So tech (XLK) has the highest correlation in this period.

Let's look at 30 largest holdings in the XLK ETF correlated to the SPY. Fourteen of the 30 stocks have a correlation above .80 over 5 days on the 5-minute bars, though some of those have negative correlation over one day.

Ticker 1day 3day 5day
ACN 0.9302 0.9688 0.9333
GOOG 0.7398 0.8877 0.9256
ORCL 0.8957 0.917 0.9242
V 0.814 0.9057 0.9096
EMC 0.6982 0.9131 0.9071
INTU -0.4818 0.8688 0.902
MA 0.771 0.893 0.8764
TXN 0.8928 0.9205 0.8675
INTC 0.894 0.9281 0.861
CSCO 0.8096 0.913 0.8473
MSFT -0.3076 0.8504 0.8421
ADP 0.9653 0.8356 0.8419
MSI 0.601 0.865 0.8362
T 0.8457 0.8317 0.8235
YHOO -0.629 0.7454 0.814
ADBE 0.7368 0.9113 0.8062

Also, here the correlations between the 30 largest holdings of XLK and the XLK. Only 11 of the 30 stocks had a 5-day correlation above .80.

Ticker 1day 3day 5day
INTU -0.453 0.8935 0.8972
GOOG 0.7758 0.8861 0.887
CSCO 0.8339 0.9449 0.8832
V 0.8147 0.901 0.8744
ACN 0.9305 0.949 0.8728
ORCL 0.9174 0.8849 0.8622
EMC 0.6069 0.9401 0.8574
ADP 0.9746 0.8343 0.8541
INTC 0.8896 0.93 0.8536
TXN 0.8952 0.9299 0.8485
MA 0.7608 0.8939 0.8125
 
Quote from boddintrader:

Here are correlations between the SPY and the nine S&P sector ETF's using 5-minute bars for 1, 3 and 5 days.
More...
When estimating pairwise ETF correlations using 5 minute bars you should correct for Epp's Effect and microstructure noise, otherwise your estimates will be severely biased. The corrections developed by Zhang et alia in papers between 2005 and 2011 are as good as any. The RTAQ package in R implements this.

Also, I hope you are adequately compensating for the overnight discontinuities in your 3 and 5 day estimates.
 
You must log in or register to reply here.
Back
Top