Another 'Why History Is Not a Guide' for Today's Trading

I usually don’t like to talk about strategies that don’t work, but this one was interesting (and I think proves an important point), so here goes…

I was inspired by the Quantifiable Edge’s post Bank Action and the Market to look at how the market has performed historically the day following relative strength/ weakness in banking stocks (represented by the index BKX).

Below are results from early 1993 (when BKX launched) through the end of 1999:

[logarithmically-scaled]

The red line represents going long the S&P 500 tomorrow (from today’s close) if the bank index outperformed the S&P 500 today and the blue line if it underperformed. This is a proof of concept, so these results are frictionless (i.e. do not account for transaction costs or slippage).

That is an extremely consistent result. Remember, this isn’t some long-term strategy that might “appear” to be working solely as a result of market conditions at the moment. This is a very short-term strategy changing positions on average once every 2.0 days.

While I don’t think this observation was ready for prime time all by itself, I would have certainly used the general concept as part of some broader strategy.

But look at what has happened to the strategy in our current decade:

[logarithmically-scaled]

Ouch. Zero predictiveness.

Lessons learned?

Most of these we’ve talked about before. The markets are evolving. Trading strategies should either be adaptive, or we should at least be very sensitive to moving on to new strategies as current ones wane.

And something that I don’t know that I’ve specifically said (but I hope I’ve shown through this blog), we should constantly be in search of new trading advantages as the market offers them. They’ve always been there and (I believe) they’ll always be there…we just have to find 'em.

Happy Trading.

P.S. this is NOT a critique of Quantifiable Edge’s post. QE is looking at this BKX relationship at extremes, while I’m looking at it as much more of an everyday observation.

P.P.S. was there some change in the methodology used to calculate the BKX that led to the breakdown of this strategy? I dunno…any thoughts?

Comments

[John Lounsbury:]

Michael - - -

Thanks for another good article. This is another example of why blindly following a trading system can lead to disappointment. I have found it necessary to be constantly monitoring any strategy to make sure it is "still working".

Looking at the graphs, my eyeball indicates that the strategy had periods of under performance during the "good years" (first graph): 1H/1993, late 1994, 1H/1996, 2H/1997 and most of 1999.

There were also periods of out-performance during the "bad years" (second graph): 2H/2007 and 1H/2008.

So to use a metaphor, you have to watch for periods when your goose stops laying golden eggs, but make sure you stop waiting for more golden eggs when the goose turns into a turkey.

Mar 04 11:17 AM

[Fred Voetsch:]

Actually, history is a very good guide but you have to know where in history to look. We are now seeing a correction to the excesses of the past 2-3 decades and so looking to them for answer, except in reverse, perhaps, is foolish.

We are also not in a high (working) population growth, inflationary period so the 1970's are also not a good time for comparison.

We are in a deflationary deleveraging and so looking way back to the 1930's would be the most likely to offer useful data.

Mar 04 12:48 PM

[Michael S:]

RE to Freddyv: I'll have to apologize for SA's title - it was NOT the original from my blog (but the editors here take a lot of liberty in modifying authors' republished works). I agree that history IS a guide. The point of the post was that sometimes the market makes abrupt shifts and we have to be prepared for that as mechanical traders. michael


On Mar 04 12:48 PM freddyv wrote:

> Actually, history is a very good guide but you have to know where
> in history to look. We are now seeing a correction to the excesses
> of the past 2-3 decades and so looking to them for answer, except
> in reverse, perhaps, is foolish.
>
> We are also not in a high (working) population growth, inflationary
> period so the 1970's are also not a good time for comparison.
>
> We are in a deflationary deleveraging and so looking way back to
> the 1930's would be the most likely to offer useful data.

Mar 04 03:25 PM