A fresh look at data and back testing

[kevink]

The wonderful Portfolio Charts site continues to grow in sophistication and usefulness.

This new post is a real tour-de-force exploration of the utility and dangers of back testing, data mining and relying on such info to construct portfolios.

https://portfoliocharts.com/2019/07/23/how-to-study-portfolios-when-the-data-is-full-of-bubbles/

 

[MBAustin]

Interesting. I didn't try to understand it in depth, but my takeaway is (as always) to take any analysis that pushes (or bashes) a specific asset class with a few grains of salt.

 

[NW-Bound]

Yes, take away the bullish years, and the stock market return looks meager.

On the other hand, a stock lover will say that if we take away the sorrowful years, the market looks great.

So, what does it really mean, other than showing that all assets go through strings of ebullient years, followed by years of despair?

I try to keep reminding myself that we have been through 10 years of wonderful return, and it's time to lower my expectations for the immediate future.

I survived the bad years of 1999-2009. I hope to survive the next onslaught.

 

[LRDave]

This was painful to watch. Not the data - the presentation.

That said, this currently mimics our plan - more spent in the 60-70 years and declining spending from there.

 

[kevink]

I think that both that specific article and the site as a whole (especially the “Portfolios” section) are invaluable for early retirees. Sequence of returns risk and start date sensitivity matter a lot to those living off of assets with limited or nonexistent ability to replenish them.

A comparison of the metric he calls the Ulcer Index for, say, the classic 60:40 or Three Fund portfolio with the Larry Swedroe, Golden Butterfly or All Weather shows the value of broader diversification and a more defensive mindset for those who’ve already (in William Bernstein’s words) “won the game.”

 

[NW-Bound]

Anything that weans investors off the expectation that the market will keep on returning 10+%/year is good.