As I have access to a bloomberg terminal at work, I like to play around with historical price & time data, of an evening. My latest curiosity is what I think of as seasonality. Not in the Earthly or agricultural sense of the word; I mean focussing on whether there are specific times in a day or week or month or year that price (for whatever product) is likely (based on historical data) to rise/fall over the next X periods of time. This thinking is not without precedent e.g. I recall a period a couple of years back when the S&P 500 experienced rallies every Tuesday for many consecutive weeks. I'm mindful of what ET member Garachen has noted in the past; that one must be able to rationalize why a strategy should work. My attempt to explain why focussing on TIME may yield fruit is that large institutional players, who may be price insensitive, participate in the markets per a schedule. e.g. "Every day at 11:30am, for the next two weeks, we are going to buy 5000 contracts". I have no idea if any institutions behave in such a fashion*. Does anyone have useful insight on this particular point? Or any input as to whether the thinking behind my exploration of this subject has any credible basis? *This might seem overly stupid behaviour, but there are plenty of examples... e.g. UK's sale of gold reserves, which was telegraphed to the rest of the market via official announcements. Thanks
stocktradersalmanach has graphs cutting the week in separate days, than the day in half an hour sections, indicating the percentage of times the market is up or down during those particular times. Not so sure how useful or reliable it is although I used to add to longs during or just before the last 90minutes of the trading day, did a few daytrades like this as well, with mixed results. It's probably better if you can study shorter term trends as their calculation is an average over several years/decades.
There are "regularities".... opening, late-day, and overnight. Not all that precise, but worth the effort to get a handle on them.
I think this would be an interesting study to perform, but you have to consider macro and financial conditions along with it; because.... this Tuesday phenomenon happened when thefed was doing QE, and a good chunk of the money from fed treasury purchases from prime brokers on the prior week's Wednesday and Thursday tended to flow into equities on the following Tuesday. Also, I've heard that at the end and beginning of quarters, there's a lot of batched transactions scheduled from mutual fund managers and pension funds; these tend to be MOC orders that hit the market around 2pm, and the activity from market makers committing to fill the orders will push the market higher or lower as they try to get a ahead of the fund buyers/sellers.
It's good to keep these kind of things in mind -- but don't get too obsessed with it. In trading it's generally good to keep and open and flexible and limber mind.
Just when you done ton of back testing and shows over long periods there are spurts of something looking like a reliable pattern and start trading it, you wonder "what was I thinking". It is like anything dealing with times or patterns, it is 50/50. Buy the Rumor and fade the fact, but be out ten seconds before the release of time or reports, they will sink you. Forgot who did the study of being long the last 4 business days of the last month and 2 business days of new month works well. Something about Mutual Funds dumping losers and buying something else. There use to be "unique" times when there were actual day sessions trading S&P futures, but not so much any more as Bond and Currency markets don't now close. Close of New York is still good if it closes very strong or weak in upper/lower 95% of big range bar as it will often reverse in last fifteen minutes of ES, guess that could be true for most markets. Have to agree with Lawrence-Lugar.
Because short term trends contain mostly randomness because of the tug of war of supply and demand between algos, institutional traders, soveriegn fund flows, etc, it is hard to exploit any statistically significant patterns / seasonality from the data. Information assimilated into long trends ( expected earnings, monetary flows, economic factors, etc. ) tends to be more robust and is what gives the trend "persistence" ( bull markets last for "years"); the evolution of this information/ trend tends to act like an oil tanker and it takes time to change direction. Analysis conducted on monthly / yearly time frames ( U.S. equities market ) has tended to yield clues towards providing statistically significant outcomes. Yet, as the financial industry is based on sales of products and services / management, and the "trading" depts of hedge funds and banks are focused on "short term", research of this type is low on the totem. The most basic evidence based process involves the use of moving average / price crossovers using 10 period MA (on a monthly basis) applied to SP500 index and long bond fund proxy. The "monthly" time frame tends to reflect the "long trend" information assimilation well, and "contains" the whipsaw activity that would trigger transactions unnecessarily if one used more frequent MA time frames ( 200 day MA is most popular ). Using the QQQ ETF as the equity selection and VUSTX ( EDV ) for the bond allocation has produced risk adjusted CAGR of 14.2% since 1985, with 40 transactions. There are other more complex methods and variables, based on "long term" trend, that have produced statistically significant, repeatable outcomes but that is beyond the scope of this post.