Buy In October: A Strategy Worth More Than 20% A Year For 10 Years

[badbally]

http://seekingalpha.com/article/921611- ... source=msn

Saw this article today and wanted to get some feedback whether anyone here thinks it can work and if someone can do additional back testing?

I joined the PP forum about 2 months ago and I appreciate the mind meld that I get from some of the regular PP contributors.

I'm always looking for idea's on how to juice up the returns for my VP and I thought this could be a possibility. 

Thoughts?

[Gosso]

http://seekingalpha.com/article/921611- ... source=msn

Saw this article today and wanted to get some feedback whether anyone here thinks it can work and if someone can do additional back testing?

I joined the PP forum about 2 months ago and I appreciate the mind meld that I get from some of the regular PP contributors.

I'm always looking for idea's on how to juice up the returns for my VP and I thought this could be a possibility. 

Thoughts?

I thought I'd take a stab at this.  Here are the results from placing $100 into the S&P 500 in 1958:



Trading strategy for the "Seasonal" was to invest from March 1 to April 30 and Oct 1 to Dec 31, when not invested funds were placed in T-Bills.

Trading strategy for the "200 Day +/-" was to invest when above 200 day average, or place funds in T-Bills if below.  I also applied a buffer of +3 and -5% on the 200 day average to avoid minor whip saw and reduce the number of trades.  Strategy was monitored daily after the close.

Here are the stats:
                      CAGR    STDEV
Seasonal          10.7%      10.5%
B&H                  9.5%      16.5%
SMA (200 day)  10.6%      12.2%

My general impression is that the Seasonal Strategy is a decent trading system, which lowers volatility and market risk.  However, it will significantly under-perform during bull markets.  I'm not sure why these months are best to invest in, maybe because people are happier around the holidays and Spring?  It may continue to work for a while but it is only a matter of time before it begins to under-perform the B&H.

[MachineGhost]

The Halloween to May period returns about 6.5% more on average than the post-May to pre-Halloween period.  It has been getting stronger in recent years.  I think the sun solar cycle plays a big role as well as institutional effects (taxes, Wall Street off during summer, etc.)

Gosso, how about running a test that combines both the 200MA and the Halloween-May effect?

[Gosso]

The Halloween to May period returns about 6.5% more on average than the post-May to pre-Halloween period.  It has been getting stronger in recent years.  I think the sun solar cycle plays a big role as well as institutional effects (taxes, Wall Street off during summer, etc.)

Gosso, how about running a test that combines both the 200MA and the Halloween-May effect?

Combining the two trading strategies from 1958 to 2012 saw a CAGR of 8.9% with a STDEV of 9.3%. 

Using levered ETF's when invested in the S&P500 gives the following for the Combo trading system:

2x = CAGR 11.6%, STDEV 18.1%
3x = CAGR 13.7%, STDEV 27.5%

I also tested an "Off Season" trading system, which had a CAGR of 4.4% and a STDEV of 12.5% (this is worse than T-Bills over the tested timeframe).  Adding the 200SMA to the "Off Season" increased the CAGR to 7.1% with a STDEV of 8.6%.

The same seasonal effect can be seen in the Nikkei (total return in Yen) from 1985 to 2012:
              CAGR    STDEV
B&H          0.6%    23.6%
200SMA    5.7%    14.0%
Seasonal    3.8%    18.0%
Combo      5.9%      12.6%
Combo x2  8.9%    26.7%
Combo x3  11.0%    43.1%

The downside is that seasonality under-performs during bull markets.  B&H with the Nikkei from 1985 to 1990 saw a CAGR of 26.9%, while the Combo saw 13.5%.

Overall it seems like summer is typically a period of lower returns, but I have no clue if this will continue into the future.  I think I'll just stick with the SMA.

[D1984]

The same seasonal effect can be seen in the Nikkei (total return in Yen) from 1985 to 2012:
              CAGR      STDEV
B&H          0.6%    23.6%
200SMA    5.7%    14.0%
Seasonal    3.8%    18.0%
Combo      5.9%      12.6%
Combo x2  8.9%    26.7%
Combo x3  11.0%    43.1%

The downside is that seasonality under-performs during bull markets.  B&H with the Nikkei from 1985 to 1990 saw a CAGR of 26.9%, while the Combo saw 13.5%.

Overall it seems like summer is typically a period of lower returns, but I have no clue if this will continue into the future.  I think I'll just stick with the SMA.

Gosso, could you please run two more simulations? One, see how the "200-day SMA" and the "200-day SMA/buy in October combo" did with the Nikkei starting January 1, 1990 and continung to the present? And two, see how (use Yahoo Finance data for the DJI or Federal Reseve data for the S&P 500...which I guess was the S&P 100 at the time...the 500 didn't start until the late 50s) the "buy in October/sell on Dec 31 and then buy back again March 1st and sell on April 30th" did from January 1929 to December 1932? Assume the funds were in t-bills/cash when they weren't invested in the market.

Finally do you have software that calculates the 200-day SMA or do you use a web service (Yahoo Finance, YCharts, etc) to do it for you? If you have software please let me know what it is because I have some numbers going all the way back to the 30s and 40s I'd like to run manually. Thanks.

[Gosso]

Gosso, could you please run two more simulations? One, see how the "200-day SMA" and the "200-day SMA/buy in October combo" did with the Nikkei starting January 1, 1990 and continung to the present?

Sure.  Returns from Nikkei (total return in Yen) between Jan 1, 1990 and Oct 1, 2012:

                CAGR      STDEV
B&H        -4.6%      21.9%
200SMA    2.1%      10.5%
Seasonal    1.7%      18.3%
Combo      4.3%      12.3%

Nikkei Total Return in Yen

And two, see how (use Yahoo Finance data for the DJI or Federal Reseve data for the S&P 500...which I guess was the S&P 100 at the time...the 500 didn't start until the late 50s) the "buy in October/sell on Dec 31 and then buy back again March 1st and sell on April 30th" did from January 1929 to December 1932? Assume the funds were in t-bills/cash when they weren't invested in the market.

From 1928 to 1933 the Dow returned the following (dividends excluded):

                            CAGR      STDEV
B&H                    -21.1%      39.6%
200SMA                -0.1%        24.2%
Seasonal              -20.7%        31.7%
Combo                  -0.6%        14.0%
Off Season            -0.7%        24.9%
Off Season + SMA  1.2%        11.1%


* Note: The start date is 1900.

The 200SMA got whipsawed in late 1932 which is why it spikes down.  A volatility filter might have helped reduce this.

Finally do you have software that calculates the 200-day SMA or do you use a web service (Yahoo Finance, YCharts, etc) to do it for you? If you have software please let me know what it is because I have some numbers going all the way back to the 30s and 40s I'd like to run manually. Thanks.

I developed my own Excel spreadsheets (so there could be errors ), which are setup to handle daily closing prices and calculate the 200 day SMA.  I can then mess around with different timing systems.

[D1984]

Sure.  Returns from Nikkei (total return in Yen) between Jan 1, 1990 and Oct 1, 2012:

                CAGR      STDEV
B&H        -4.6%      21.9%
200SMA    2.1%      10.5%
Seasonal    1.7%      18.3%
Combo      4.3%      12.3%

Thanks for these. Why does the first chart itself say "(10) +/-" instead of "(200 day +/-)"...did the first one use 10 month MA instead of 200-day? Also, would it be possible to restart the graph for the Nikkei and Dow at 1-1-1990 and July 1st 1929 respectively?

Where are you getting Nikkei 225 total return in Yen data from? CRSP? Global Financial Data? Because if it is Yahoo Finance that is only price and not total return IIRC.

What is a "volatility filter" and how do those work?

Are your spreadsheets proprietary (and are they for sale and if so how much)? I have Nikkei (price only) data from inception at 1949 and it appears that the late 40s/early 50s were a period when seasonal investing (at least without combining the 200-day MA) might have actually worked against an investor. If the spreadsheets aren't proprietary (and are free for use), can you post a copy on an online file host somewhere? I don't want to have to keep bothering you with requests to run backtests and if I had the spreadsheets myself I wouldn't need to 

[Gosso]

Sure.  Returns from Nikkei (total return in Yen) between Jan 1, 1990 and Oct 1, 2012:

                CAGR      STDEV
B&H        -4.6%      21.9%
200SMA    2.1%      10.5%
Seasonal    1.7%      18.3%
Combo      4.3%      12.3%

Thanks for these. Why does the first chart itself say "(10) +/-" instead of "(200 day +/-)"...did the first one use 10 month MA instead of 200-day? Also, would it be possible to restart the graph for the Nikkei and Dow at 1-1-1990 and July 1st 1929 respectively?

The (10) +/- was a relic from copying over from an old spreadsheet that used monthly data.  But you are right it is the 10 month SMA, which is equivalent to the 200 day SMA.  However in that chart I used the 200 day SMA.

Where are you getting Nikkei 225 total return in Yen data from? CRSP? Global Financial Data? Because if it is Yahoo Finance that is only price and not total return IIRC.

I used Yahoo! and then added a 2% annual dividend.  Not a perfect total return, and perhaps a bit sloppy, but I need some sort of dividend to properly measure the levered ETF's.

What is a "volatility filter" and how do those work?

http://www.advisorperspectives.com/dsho ... gement.php

http://www.advisorperspectives.com/dsho ... Part-2.php

Although I haven't been able to fully replicate their success.  Once the 200 day SMA is applied then adjusting for volatility almost made no difference, at least with my backtesting.  I was thinking I could use volatility to adjust my buffers around the 200SMA to help reduce whipsaw.

I have Nikkei (price only) data from inception at 1949 and it appears that the late 40s/early 50s were a period when seasonal investing (at least without combining the 200-day MA) might have actually worked against an investor.

The Dow between 1928 and 1956 showed that June, July and August were the strongest months.  September was still the worst month...it seems September is usually a good time to buy...not sure why though?  So it appears seasonality will not work over all time periods.

Are your spreadsheets proprietary (and are they for sale and if so how much)?
[...]
If the spreadsheets aren't proprietary (and are free for use), can you post a copy on an online file host somewhere? I don't want to have to keep bothering you with requests to run backtests and if I had the spreadsheets myself I wouldn't need to 

Hmmm, to open source or not to open source?   I could give one "as is" but it would take a lot of work to figure out how it works.  If I find the time I might polish one up and distribute it.

[Gosso]

The Dow between 1928 and 1956 showed that June, July and August were the strongest months.  September was still the worst month...it seems September is usually a good time to buy...not sure why though?  So it appears seasonality will not work over all time periods.

Maybe this...

http://en.wikipedia.org/wiki/List_of_st ... ar_markets

There's an awful lot of crashes in Aug-Oct.  Buying in Sept might be akin to catching the bottom, on average.

Siegel has a whole chapter on calendar anomalies in Stocks for the Long Run.  There's another that spans the first few weeks of January - possibly linked to people dumping losers for cap loss harvesting.  

Possibly.  The more I look at seasonality, the less I like it.  For all we know September will be the best month moving forward.  Plus it completely failed pre 1950.  Although maybe it moves in long cycles...somehow tied to the collective unconsciousness and the position of the planets... 

[MachineGhost]

Possibly.  The more I look at seasonality, the less I like it.  For all we know September will be the best month moving forward.  Plus it completely failed pre 1950.  Although maybe it moves in long cycles...somehow tied to the collective unconsciousness and the position of the planets... 

The equinoxes do shift each year.  A true seasonal calendar would align to the fall and spring equinoxes for the starting and ending dates.  That may be why the calendar isn't that hot.