Alternate LMBF methods

[Cactus]

OK,here are the monthly returns as of 1/30/13.

[TD="bgcolor: #FFFF99, align: left"] Date [/TD]
[TD="bgcolor: #FFFF99, align: center"] G FUND [/TD]
[TD="bgcolor: #FFFF99, align: center"] F FUND [/TD]
[TD="bgcolor: #FFFF99, align: center"] C FUND [/TD]
[TD="bgcolor: #FFFF99, align: center"] S FUND [/TD]
[TD="bgcolor: #FFFF99, align: center"] I FUND [/TD]

[TD="bgcolor: #FFFF99, align: left"] 30-Jan-2013 [/TD]
[TD="bgcolor: #CCFFCC, align: right"]0.12%[/TD]
[TD="bgcolor: #CCFFCC, align: right"] (0.63%) [/TD]
[TD="bgcolor: #CCFFCC, align: right"]5.45%[/TD]
[TD="bgcolor: #99CC00, align: right"]6.49%[/TD]
[TD="bgcolor: #CCFFCC, align: right"]4.77% [/TD]

The best fund is S so LMBF-1 goes into the S Fund by noon eastern tomorrow. We'll see tomorrow how that compares with LMBF.

As for me, I think S has topped and is set for a pullback, but hey, I'm just recording the system here not putting my money on the line.

 

[Cactus]

Thanks for testing this system out, Cactus. There is the possibility that better results were achieved for a very simple reason: because you were in the stronger fund on the first day of the month rather than waiting for day #2 to be invested in the stronger fund. Frixx'x chart has a lot to say about that, doesn't it?

Yes, it does. I like his chart. The fact that he is using the S&P500 gives him a wider time range of data to look at. That's a big plus. I'm looking at the TSP funds and we only have daily data for that going back to mid 2003. After that you're dealing with monthlies back to inception.

 

[MrJohnRoss]

Thanks for testing this system out, Cactus. There is the possibility that better results were achieved for a very simple reason: because you were in the stronger fund on the first day of the month rather than waiting for day #2 to be invested in the stronger fund. Frixx'x chart has a lot to say about that, doesn't it?

You may very well be on to something important here. A nine year total return of 163.5% compared to 102.5% is nothing to sneeze at.

Keep up the good work,

JR

 

[Cactus]

I agree with JTH. When designing a medium-term system, if one day makes a large difference in the result, it is unlikely to reproduce that result. Over time I can only assume that it will regress to a mean. Why not try the LMBF+/-(1-28) and see which date gives you the best return? If you tabulated that for 50 years I think you'd find that the results are fairly similar. It's the same system and acts on the same performance no matter which day you start unless you can align it with some other trend or logic. If you can statistically demonstrate the first day of any month is likely to be positive, then it may be an effect worth pursuing.

Yes, I think you are right. This could easily be a fluke of the small time span chosen and better results could be achieved over different days of the month in different time spans. I think it does show that one day can make can make large difference in your annual return, which surprised me, but maybe even that is merely a factor of a volatile market which we've had during this time.

Running the numbers with +/1 1-28 days is a good idea but better suited to a program than a spreadsheet. I'm too error prone with a spreadsheet which is why I thought of trying this to begin with.

 

[craigerv]

I agree with JTH. When designing a medium-term system, if one day makes a large difference in the result, it is unlikely to reproduce that result. Over time I can only assume that it will regress to a mean. Why not try the LMBF+/-(1-28) and see which date gives you the best return? If you tabulated that for 50 years I think you'd find that the results are fairly similar. It's the same system and acts on the same performance no matter which day you start unless you can align it with some other trend or logic. If you can statistically demonstrate the first day of any month is likely to be positive, then it may be an effect worth pursuing.

 

[JTH]

Interesting to see how one day makes such a difference. I'm really surprised how 12 times LMBF-1 chose a different fund 12 times. I wouldn't have thought 1 day difference would ever cause that except once or twice in 10 years. What external force happens at the end of the month/1st of the month to cause this difference... we'll probably never know. Being a long term investor, in both TSP and personal stock accounts, I never took a one day change as significant.

Thanks for taking the time to run the numbers and produce this info.

That's one of the problems with with many self-created systems traders create. Sometimes they manipulate the data just enough to get the desired results (often by just a day.) This is why I stress the importance of long/multiple time frame based back-testing. More than a few times we've seen members post how great their system is, then when it goes to crap, they suddenly disappear from the forum...

 

[Frixxxx]

Alternate Investing: Imagine if you were on the "S&P500 First Day of the Month Club" translate: In on the First, out the rest of the month.

 

[Hallatauer]

Interesting to see how one day makes such a difference. I'm really surprised how 12 times LMBF-1 chose a different fund 12 times. I wouldn't have thought 1 day difference would ever cause that except once or twice in 10 years. What external force happens at the end of the month/1st of the month to cause this difference... we'll probably never know. Being a long term investor, in both TSP and personal stock accounts, I never took a one day change as significant.

Thanks for taking the time to run the numbers and produce this info.

 

[Cactus]

I'll start by describing a simplification I call the Last Month Best Fund Minus 1 Day (LMBF-1) method. What I didn't like about LMBF is that you are in a different fund on the 1st day of the month every time you make an IFT. That happens quite frequently (9 - 10 a year) with this method. Being prone to errors, I wanted to be in a fund the whole month so my return is simply the monthly return for that fund. Monthlies are easily verifiable.

My solution was basically to throw out the last day of the month when figuring out the best fund. I did that so I can make any needed IFT by noon on the last day of the month instead of the first day of the following month. This way you are in the same fund the whole month instead of day 1 being different. You also have the emotional benefit of using a precious IFT at the end of the month instead of the beginning in case you later decide to bail out of this method (not recommended).

I back-tested this method and compared it to the LMBF. As you can see from the 2 tables below, this didn't have a large effect on the fund chosen, though there are some, but it did have a surprisingly large effect on the rate of return. What a difference a day makes.

At first I thought throwing out the last day of the month removed some kind of End-Of-Month effect in choosing the best fund but this doesn't appear to be strictly the case. The biggest difference in fund chosen/year is 3 with an average of 1.33/year. There are 3 years (2006, 2008, 2009) with 3 differences. Two of those years (2006 & 2009) have return differences on the order of 10%, which is the largest we see, but 2008 has a return difference of less that 1% which is among the smallest.

It looks like another factor for the larger returns come from already being in the chosen fund as we enter the new month. Does the FRTIB/Blackrock give us some kind of dividend on the 1st of the month? I do remember a couple of years ago folks trying to take advantage of this day 1 effect. I think it was positive 2 times out of 3. Hmmm.

Another interesting difference is that there are fewer total negative months for LMBF-1 (24 vs. 33) and every year has an equal or fewer number of negative months. This should help in the psychology of sticking with a method long term but isn't itself responsible for greater returns. The big loser year, 2008, only had 2 negative months for LMBF-1 and 3 for LMBF.

Other things to note are the LMBF did produce better results than LMBF-1 for 2 years out of 9 (2007 & 2011). The total return for 2004 - 2012 are 102.50% for LMBF and 163.51% for LMBF-1. Those are the only years I have complete daily data for. I wish I could go back to year 2000 at least, but this does give us 9 years of data to look at and consider.


Table 1: LMBF Returns

Note: IFTs/pos represents the number of IFTs in a year and the annual position relative to the 5 TSP Funds respectively. Both methods beat all 5 TSP funds for the last 2 years, but that is only 2 years out of 9.


Table 2: LMBF-1 Returns

Note: The olive colored fund blocks represents months where the LMBF-1 chose a different fund from LMBF.


LMBF-1 does look like an interesting modification to pursue. The way it works, again is: 1) Determine the best fund on the penultimate (next to last) day of the month, 2) Make any necessary IFT before Noon Eastern, 3) come back next month and do it again.

I will track the LMBF-1 here and compare it to the LMBF as well as any others that interest me throughout the year.

 

[Cactus]

This thread is for testing modifications of the LMBF method. If you have a modification you would like to track, describe it here and post your data. Then post the IFT's and monthly returns as they come in so we can all track your results.