Sell in May?: not today

It's May Day. Usually, I'd mark the occasion by singing the Internationale and transferring my pension fund out of shares. This time, though, I'll not be doing the latter.
This isn't because I've made a judgment that the market will go up; anyone who uses unaided judgment is an idiot.
It's because there's another rule that more than outweighs the "sell in May" rule. It's the "buy when the dollar is weak" rule.
In my day job, I've estimated that the strong positive message for All-share returns in the next six months sent by the weak dollar more than offsets the bad seasonal pattern in stocks.
Surprisingly, these two facts alone can account for half the variation in six-monthly returns since 1990.  Which raises the question: why do people use so much "judgment" and story-telling in thinking about the market when there seems to be so much predictability from simple rules?

Comments

[dearieme]

Singing and selling? You're not young anymore, then?

Hurray, hurray, the First of May,
Outdoor **** starts today.

[chris]

Not for me it doesn't.

[Charlie Whitaker]

"why do people use so much "judgment" and story-telling in thinking about the market when there seems to be so much predictability from simple rules?"

And you're not contributing to the lore of stock trading? If the rules are simple (and succinct) please to write them all down in one place, onetime?

[dsquared]

because these simple rules are much easier to spot in hindsight, and have a habit of stopping working without warning? I mean, have you tried a Hansen's Reality Check type test on some of these rules?

[chris]

D2 - what have you got against hindsight? It's all we have.
Of course, such rules can (and do) stop without warning; they also, of course, fail occasionally through ordinary random noise. But I know from out-of-sample testing with my own money that this hasn't obviously happened to the sell in May rule.
I haven't tried a Hansen SPA test - I'm not sure how to rig up Excel to do it - but no test can prove that a trading rule will work in future, however well it's done before.
And surely, I'm just about the last fella you should be directing your complaint to. Both rules I refer to are entirely consistent with the EMH, and today they boil down to just one rule - hold some equities all the time.

[echo]

If you use matlab you can download an implementation of the SPA test and White's original Reality Check from Kevin Sheppard's website: http://www.kevinsheppard.com/research/ucsd_garch/ucsd_garch.aspx
And by the way neither of the two rules is consistent with the EMH.-if the market was efficient you wouldn't be able to use publicly available information to achieve superior risk adjusted returns.

[dsquared]

what have you got against judgement? :-)

I just think that this sort of analysis is dangerous unless you're carrying out some sort of check that you haven't just datamined. The (IMO erroneous) characterisation of Soros and Buffett as lucky winners of a coin-tossing competition is definitely applicable to trading rules of this sort. It's like that time you compared Anthony Bolton to a quite obscure valuation filter - I would have needed some convincing that this filter rule was known when the Special Sits fund was set up, or that its performance will be robust going forward.