Gyroscopic Investing

In the interests of transparency, the short, official, authoritative answer is no.  I forgot I use proprietary T-Bond historical data that I spent several hundred dollars to acquire (indirectly), so there is no way for anyone to verify anything I claim publically.  I am thus stuck in trying to prove the equivalent of a negative which is unfair to myself and everyone else that wants independent verification.

Does that mean I want to spend at least several hours to several days day investigating my code, tracking down a paper trail for possible data errors, etc. all for what is a 5% discrepancy at best?  No, it's not worth it.

Does that mean a certain young whippersnapper dickhead known as LC475 that publically attacked and disrespected me personally and still hasn't apologized for it, was right in his speculative conjecture?  No, there's nothing wrong with the T-Bond data which is not a synthetic derivation from 20-year T-Bonds or anything like that.

As I've mentioned before, I like realistic backtests as compared to simple simulations such as Peak2Trough's or others that just use public data straight-up without accounting for real world effects.  I like to include transaction costs which can add a few percentages.  I like to use slippage which can add a few percentages due to not capturing the overnight sessions.  I like to use non-calendar aligned starting and ending dates which can add a few percentages.  I like to rebalance at 365.25 + 1 days to account for long term capital gains tax which can add a few percentages.  I often have bugs in my code that can add a few percentages.  I use publically available historical data that often have bugs and can add a few percentages.  Any or all of these factors can explain the subtly different versions of the numerous Yearly Profit Charts that I've been publishing over the years -- all of which state different nominal CAR and MaxDD values.  As I recall, the worst of them will be around -26% nominal MaxDD, in fact!

So going forward I am no longer publishing or making claims that cannot be independently verified.  Since publically available T-Bond daily data doesn't go back earlier than 1977, it's also the end of the line for me.  I propose that Peak2Trough become the official reference backtester for all things PP.  It is extremely good and if he in fact uses dividend-adjusted S&P 500 total returns back to 1968 (which is not publically available), then it even surpasses what I have.

Good post MG.

For what it's worth I have frequented the Motely Fool Mechanical Investing board since the late 90s and while it doesn't always get applied, the basic rule (which would be good here) is beat an idea to within an inch of its life but never attack the person. No intent to do that on my part for sure (the person part)...so back to the idea. I really just don't see how it could have happened. There is expired series data on FRED that you can use that goes before 1977 for bonds which is where I took my yield data from. http://research.stlouisfed.org/fred2/series/LTGOVTBD

What I'm not sure on is the maturity, 20 or 30 yr? I assumed 30 year which would be worse case, 20 yr of course would have been less a draw down. The 10 year as mentioned which is widely available was only -5% for the year. Regardless, given this data stocks and gold would have had to of cratered harder to get one down to -25% and I couldn't even get close to that. Finding good public data for both stocks and gold is not difficult so unless there was a huge overnight major spike in something I again just don't see it anywhere close to 25%.

As for me...I would love to be proven wrong in the pursuit of better understanding and knowledge.

Kbg wrote: What I'm not sure on is the maturity, 20 or 30 yr? I assumed 30 year which would be worse case, 20 yr of course would have been less a draw down. The 10 year as mentioned which is widely available was only -5% for the year. Regardless, given this data stocks and gold would have had to of cratered harder to get one down to -25% and I couldn't even get close to that. Finding good public data for both stocks and gold is not difficult so unless there was a huge overnight major spike in something I again just don't see it anywhere close to 25%.

As for me...I would love to be proven wrong in the pursuit of better understanding and knowledge.

The problem with monthly data is it can hide drawdown troughs.  Things can happen really fast when people panic in just days to two weeks.  I do wonder how Peak2Trough is backfilling his T-Bonds to 1968 since he is providing daily values.  He doesn't seem to frequent the forum anymore to answer questions.

I may be wrong about the -25% happening in 1969 as I was just going by memory from the last time I did that backtest last year, but it definitely showed up before 1980 since I remember being amazed anything could top that one.  FRED did have errors in its gold prices at one time; I contacted them about it and it looks like it is fixed now since it meshes with their source.  At one time, I also had an error in the transition from the proprietary T-Bonds I use and the official 30-year CMT T-Bonds but that was an obvious error (if you're looking for it) and it didn't occur until 1977.

Anyway, I think the bigger concern now is definitely the -25% real maximum drawdown in 1980 that Gosso put up.  Yeeowtch!

Here are my results for the real and nominal drawdowns for the PP and 60/40:

[img width=600]http://i62.tinypic.com/2ez1vet.png[/img]
(Click to enlarge)

So drawdowns measured at the end of month in nominal terms hasn't dropped below -13% since April 1, 1968 (assuming my data is correct, it might not be).  Daily data would show a greater drawdown, but I find daily data annoying to work with.

To calculate the bond data pre-1977 (and to fill in the gap from 2002-2005) I used 20 year yields and then assumed they had a 30 year term.  I figured the yield on 20 and 30 year bonds is pretty close (see FRED and click "Max").  My biggest problem with 20 year yields is they are an average of the daily yield closes for the month rather than the month end close, so this will introduce some error.  Edit: I found the daily 20 year yield from HERE, however it didn't change the data in any meaningful way, but I still updated the graphs to reflect the new data.

To calculate the total return of the S&P500 (ie dividends reinvested), I manually added the dividends to the price index. Could also use Robert Shiller's data.  In 1993 I switched over to using the adjusted closing price of SPY.  I also switched to SHY, GLD, and TLT once they were active.  I haven't accounted for commissions, spreads, fees, or taxes.

To calculate real returns I usedHeadline Inflation rather than Core.

Edit #2: Fixed title of graphs, both were previously labelled "Real", doh!

MG, can't you just run your simulation and have it report the date of the max DD?  Or is your code gone now?

I honestly don't care that much about REAL DD.  I care about real CAGR (long term real performance).  And I care about nominal DD (short term emotional/margin issues).  But real DD doesn't bother me as long as real gains recover relatively quickly.

MachineGhost wrote:Does that mean a certain young whippersnapper dickhead known as LC475 that publically attacked and disrespected me personally and still hasn't apologized for it, was right in his speculative conjecture?  No, there's nothing wrong with the T-Bond data which is not a synthetic derivation from 20-year T-Bonds or anything like that.

Speaking of not having apologized, I believe you owe various people for:

* Chastising anybody who questioned your claims.
* Calling people names for not taking the simple steps of verifying your claims themselves (which you are now acknowledging was impossible).
* Claiming that your proof was readily available on this forum, claiming that it's only laziness or stupidity which keep people from finding it, and then sending people on wild goose chases for something that didn't exist.
* Calling people "dickheads" for the crime of being persistent enough for you to finally admit the truth.

Speaking of not having apologized, I believe you owe various people for:

* Chastising anybody who questioned your claims.
* Calling people names for not taking the simple steps of verifying your claims themselves (which you are now acknowledging was impossible).
* Claiming that your proof was readily available on this forum, claiming that it's only laziness or stupidity which keep people from finding it, and then sending people on wild goose chases for something that didn't exist.
* Calling people "dickheads" for the crime of being persistent enough for you to finally admit the truth.

+1

* Instead of apologizing, you portray yourself as the victim. I would completely overlook your error, but your arrogance and defiance that everyone else was incorrect makes your confession meaningless to me.

Good luck in rebuilding your credibility.

MachineGhost wrote: I forgot I use proprietary T-Bond historical data that I spent several hundred dollars to acquire (indirectly), so there is no way for anyone to verify anything I claim publically.  I am thus stuck in trying to prove the equivalent of a negative which is unfair to myself and everyone else that wants independent verification.

Ah, thanks!  So now all of a sudden you have a Super-Secret, Unidentified Source which explains all but which Cannot Be Revealed.  How convenient!  Now we know. 

Thank you for explaining.

Furthermore, this source is in fact Incomprehensible based on everything we know, as Kbg points out.  But having no possible rational way of being true does not mean that it's not true.  Does it?

Now I also have to reconcile your sudden discovery of Secret Deep Mysteries with your statement to me via PM: "I got it from here: http://www.stern.nyu.edu/~adamodar/pc/d ... tretSP.xls", but I'm sure this contradiction is only apparent.  I just need to have more faith.  When we die and are at the pearly gates, you will explain it all to us.  I'm sure.

Good thing I am on your ignore so you do not have to be bothered reading all this tripe!

MachineGhost wrote: Does that mean a certain young whippersnapper dickhead known as LC475 that publically attacked and disrespected me personally and still hasn't apologized for it, was right in his speculative conjecture?

By the way, of course, I have actually not called you names nor insulted you nor even made any disrespectful statement to you, Oh Great Ghostly One.  I think that what has happened is that I was insufficiently reverent towards you.  Now that's a lower bar, and I admit that I did demonstrate irreverence by daring to question your pronouncements and thinking on my own.  While that is certainly a great sin against you, and I'm not excusing myself for it, it is not the same sin as disrespecting nor as public attack.

I have some old daily data (it is complete up to May 2013), but it should shed some more light on the PP drawdowns:

[img width=600]http://i57.tinypic.com/6pycfc.png[/img]

So if annual rebalancing was done on May 31 rather than Dec 31 then the PP did indeed experience a -25% nominal drawdown.  However it was extremely brief and was mainly the result of gold spiking 38% in only 6 days at the parabolic top in January 1980.

You could make the drawdown even bigger if you used intraday trading figures -- the top was $875, I believe.

At least, you could make it look even bigger.

Also, would it be possible to get your data, Gosso?  That would be great!  I'd just like to pore over it.

Yikes, this sure did get toxic fast.

Anyway, I appreciate the transparency. In hindsight, it's no surprise that libertarian-leaning Harry Browne readers won't go for argument by assertion, or anything resembling it.

I tried Googling for "Peak2Trough" and all I found was this unmaintained and apparently uninstallable library for the "R" programming language:
http://cran.r-project.org/web/packages/ ... index.html
Is there a more accessible "Peak2Trough" or is that it?

KevinW wrote: Yikes, this sure did get toxic fast.

Anyway, I appreciate the transparency. In hindsight, it's no surprise that libertarian-leaning Harry Browne readers won't go for argument by assertion, or anything resembling it.

I tried Googling for "Peak2Trough" and all I found was this unmaintained and apparently uninstallable library for the "R" programming language:
http://cran.r-project.org/web/packages/ ... index.html
Is there a more accessible "Peak2Trough" or is that it?

http://www.peaktotrough.com/hbpp.cgi

Made by a former member!

Thanks.

Gosso wrote: So if annual rebalancing was done on May 31 rather than Dec 31 then the PP did indeed experience a -25% nominal drawdown.  However it was extremely brief and was mainly the result of gold spiking 38% in only 6 days at the parabolic top in January 1980.

I just looked it up. Wow, a good chunk of the draw down was due to a blow off spike up in gold. The stuff nearly doubled in slightly over a month from Dec 79 - Jan 80. Afterward it stabilized around(ish) $100-200 up from where it started. So for this specific event, unless you were just cursed by the gods and bought the spike up one could probably subtract most of this "DD" as a non-event. Even more bueno, you rebalanced because one would have definitely hit a band and should have rebalanced.

Lesson here for all, holy cow gold can be volatile.

Kbg wrote:

Gosso wrote: So if annual rebalancing was done on May 31 rather than Dec 31 then the PP did indeed experience a -25% nominal drawdown.  However it was extremely brief and was mainly the result of gold spiking 38% in only 6 days at the parabolic top in January 1980.

So the draw down was due to a blow off spike up in gold???

Well a drawdown needs a peak and a trough, so if either one is caused by an extreme temporary move then the drawdown will reflect that extreme volatility.

Bottom line folks...while no way in 1969 it was most likely a yes way in late 1979-1980.

Your mellow PP is not as mellow as you think if you are just looking at annual data. However, both 79 and 80 were excellent years for the PP by year's end with 79 being the best ever.

Kbg wrote:

Gosso wrote: So if annual rebalancing was done on May 31 rather than Dec 31 then the PP did indeed experience a -25% nominal drawdown.  However it was extremely brief and was mainly the result of gold spiking 38% in only 6 days at the parabolic top in January 1980.

I just looked it up. Wow, a good chunk of the draw down was due to a blow off spike up in gold. The stuff nearly doubled in slightly over a month from Dec 79 - Jan 80. Afterward it stabilized around(ish) $100-200 up from where it started. So for this specific event, unless you were just cursed by the gods and bought the spike up one could probably subtract most of this "DD" as a non-event. Even more bueno, you rebalanced because one would have definitely hit a band and should have rebalanced.

Lesson here for all, holy cow gold can be volatile.

Silver is even more volatile. I bought silver bullion on the day of the top in 1980, at $48... on margin!
The good news is that it taught me not to use margin with precious metals, a very valuable lesson.

MachineGhost wrote:As I've mentioned before, I like realistic backtests as compared to simple simulations such as Peak2Trough's

Excuse me? 

Just because you're wrong about the drawdowns on the HBPP and use bad, monthly data doesn't make my simulation "simple".  And the irony of you getting huffy in this post because someone "disrespected" you without cause...

My backtester was never intended to measure things for which we have either no data, or nearly infinite variability.  Eg, bid/ask spreads, tax consequences, etc. 

Further, for clarification, I use publicly available, daily FED interest rate data on all bond series back to 1962 and synthetically calculate the value of the bond from the rate.  Is it perfect?  No, as it doesn't account for inefficiencies in the day to day pricing of bonds.  If anyone knows of a free source giving actual daily bond prices, I'll use that instead.

And yes, I use publicly available, daily, dividend-adjusted S&P 500 data.  Albeit, apparently, simply.

Peak2Trough wrote:

MachineGhost wrote:As I've mentioned before, I like realistic backtests as compared to simple simulations such as Peak2Trough's

Excuse me?

Just because you're wrong about the drawdowns on the HBPP and use bad, monthly data doesn't make my simulation "simple".  And the irony of you getting huffy in this post because someone "disrespected" you without cause...

Quiet, simpleton!

Seriously, Nice to see you here