The Day Of Week effect, does it still work

[MGJ]

Here is the equity curve.

On November 3, 1997, the CME changed the S&P futures contract specs. Before then, the big point value was $500 per point and the minimum price movement ("tick") was 0.05. Afterwards, the big point value was $250 per point and the minimum price movement ("tick") was 0.10.

Simulating the system with slippage (1 tick on entry, 1 tick on exit) but no commissions, it appears the system's performance characteristics changed when the contract spec changed.

Attached is the entire trading system in Excel, you can experiment with it however you like. Inside the spreadsheet, pre-11/3/1997 is highlighted in blue, post-11/3/1997 is highlighted in yellow. Notice that I've programmed Joe Krutsinger's exact system: If today is Friday, then buy the open of the next day and exit on that day's close. That's what Joe wrote and that's what I tested.

 

[Murray Ruggiero]

I used $250.00 per point all the way though, it is just easier and gives you an idea of how it's working. I think using $250.00 for the period before 1997 does change the equity curve a lot.

Another difference in what I am doing is I am asking the question is the next trading day a Monday. Not is today a Friday. This is a difference, but not a lot,about 10% or so. This is due to Monday being a holiday. When Joe coded this he used a system and assumed the next trading day after Friday was a Monday. This was the easiest way to do this. I used my Event Analyzer addin for TradersStudio which is currently in beta. Using that it is easy to ask the question is today Monday. Finally I used continuous back adjusted data for the S&P500 from Pinnacle data.

 

[Murray Ruggiero]

PS, my study did not include any slippage and commission and assumed a tick size of .10 and as I stated earlier $250.00 a point throughout the history. I wanted to just see what bias existed currently on different days of the week.

 

[ForexGuru]

Quote from ForexGuru:

But what about "black swans"? When the market moves up on Friday with a move enough to wipe out any previous gains using this strategy. Same thing when the market makes a big move down on Monday. Any data on those events?

----------------------------

ForexGuru

Today is the sort of day I was talking about, even though not really a "black swan" the move today was enough take a closer look at the question "Day Of Week effect, does it still work?". I would have to say NO, and did it ever work?

-----------------------------------

ForexGuru

 

[BrightPropGuy]

Quote from Murray Ruggiero:

.... I already planned on doing Third Friday analysis since that being a special case is obvious to anyone who trades....

Murray (and/or others),
What is the result for going long/short at Open and exiting at Close for only the Third Fri of the month?
Does ES, EMD, or ER2 behave differently on Third Fridays?

 

[dtrader98]

Quote from MGJ:

Here is the equity curve.

On November 3, 1997, the CME changed the S&P futures contract specs. Before then, the big point value was $500 per point and the minimum price movement ("tick") was 0.05. Afterwards, the big point value was $250 per point and the minimum price movement ("tick") was 0.10.

Simulating the system with slippage (1 tick on entry, 1 tick on exit) but no commissions, it appears the system's performance characteristics changed when the contract spec changed.

Attached is the entire trading system in Excel, you can experiment with it however you like. Inside the spreadsheet, pre-11/3/1997 is highlighted in blue, post-11/3/1997 is highlighted in yellow. Notice that I've programmed Joe Krutsinger's exact system: If today is Friday, then buy the open of the next day and exit on that day's close. That's what Joe wrote and that's what I tested.

Tried to PM you on this, but I have no idea if it went through since ET has problems updating my PM outbox.
Make a long story short, I reran your numbers and I think you somehow erroneously counted some non mondays (did you count every seven days?). For instance, 6/1/1982 was not a monday.
There's quite a few cases of other non-mondays showing up on the list.

This is a rerun with every day evaluated for o-cl system, including your slippage rates. Note the coincidence of days are not exact, since I am running categorical data only here. Also, there are not an equal number of occurences of days for all the years of data analyzed (i.e. monday had less trading days than wed). But, the outcome is what I think you want to see.

In this case, considering the original reference pt, wednesday is clearly in the lead

MGJ. Since you were kind enough to share your original work, PM me with your email and I"ll send out the modified xcel file for your use/scrutiny.

<img src="http://elitetrader.com/vb/attachment.php?s=&postid=1580669" border="0" alt=""><br /></font></p></font></p></font></p></font></p>

 

[NihabaAshi]

Quote from rcanfiel:

Discussions of this concept are not meaningful...

And you have to avoid falling into the seasonal tendency trap...

I truly believe your lost.

I've been using market seasonal tendencies profitably since the 80's.

I've discussed one particular tendency with in-depth details here at ET (very few were interested based upon the few replies) and I'm glad there's guys like Murray willing to discuss stuff he knows about.

Regardless, in my opinion, the real discussion for any type of tendency is the trade management (stuff after entry) as I've shown via an in-depth example here at ET of another tendency shared by an ET member in which I merged the tendency with my trade methodology.

Further, I enjoy learning about a tendency because I want to do my own research/analysis to see how the tendency performs when I merge my own personal trade methodology with the tendency.

http://www.elitetrader.com/vb/search.php?s=

Simply, these types of discussions are meaningful to some traders here at ET and very profitable to guys like me.

Therefore, since your on record that they are not meaningful...

I suggest you don't read this thread nor return with questions.



Mark

 

[rcanfiel]

Quote from NihabaAshi:

I truly believe your lost. I've been using market seasonal tendencies profitably since the 80's. ..

Mark, I considered you as a knowledgeable contributor, but THIS post is truly bad. You cannot strip out 80% of someone's post, change the meaning, misunderstand the content, and expect to say anything useful. I happen to be quite interested in seasonal trading. Let us try this, restoring what I said:

Discussions of this concept are not meaningful. They do or they don't. Pound the most recent 5 years of S&P (or other instrument) through the computer. See if they still hold up or not for an instrument. It is called a "study." What people believe is irrelevant.

I am telling Mr. Ruggiero that conducting a poll about whether seasonal effects work is useless. The only thing that has value is to put in multi years' worth of data, and SEE if it still works. Beliefs do not matter.

And you have to avoid falling into the seasonal tendency trap. Run 100,000 different instrument-day comboes through the computer, and some will SEEM tradeable. You have to perform validation analysis on them to ensure they are not just those that look good from simple probability that SOME must be on top.

I am telling Mr. Ruggiero that using historic trends is not automatically a reliable trading method. This is known to knowledgeable traders, that just because something "works in 12 of 15 years" does not automatically make it a seasonal tendency. Now, reread what I said, and try to understand. This is a VERY commonly known problem of seasonal trading.

 

[MGJ]

Quote from MGJ:

Notice that I've programmed Joe Krutsinger's exact system: If today is Friday, then buy the open of the next day and exit on that day's close. That's what Joe wrote and that's what I tested.

Yes, 6/1/1982 is a Tuesday. It is also the next market day after Friday, 5/28/1982. That's what Joe wrote and that's what I tested. Joe entered at the open on Tuesday 6/1/1982 and exited at the close, and so did my spreadsheet. You can see this for yourself by looking at the other page within the Excel file (excerpted below).



Since I used Excel's WEEKDAY function to identify Fridays, then selected the market-days-after-Friday to put in the worksheet named (drum roll.....) "The_Day_After_Friday", a missing Monday doesn't throw off the entire rest of the spreadsheet.

By the way, Murray was wrong when he said that "Monday is a holiday" is the only problem with Joe's algorithm. For ten points, name one other problem. For five thousand points, name two other problems.

 

[rcanfiel]

Quote from dtrader98:

, it appears the system's performance characteristics changed when the contract spec changed.

Not a reasonable explanation. All days were affected by the size change. You need a logical explaantion, not a coincidental grasping for straws