SEASONALITY STOCK REPORTS

[Qarel]

QP (to Ken): Just stay focused with some minor tweaks that have to be worked out.

Let me propose a tweak. I'll need a bit of a runway, however. Here goes.

First a question. Let's say Ken's method works. That's the assumption everyone here has. Now why does it work? Does he discover seasonality, or just good stocks? You might say: "What's the point, both are fine," but there is a point. If Ken is just finding good stocks, picking a "season" (month, quarter) is not very logical. But doesn't Ken find stocks with good months/quarters? Doesn't that prove seasonality? No, it doesn't. Let's look for a string of 5 years with good same-month results. Let's look at a stock that is flat over the 5 year period. Let's assume that the monthly results have a normal distribution. (You know: a lot about the middle, and the more extreme plus and minus scores get progressively scarcer.) Now you have a 1 in 2 chance (flip a coin) that a month is positive, and a 1 in 32 chance (1 in 2 in 2 in 2 in 2 in 2) that the same month is positive in all five years. We're looking at 12 months, so the chance to find a "green column" is 12 in 32, or better than 1 in 3. Work your way through 200 of such stocks and you'll get on average 75 green columns in 75 stocks (or less when one stock has more green columns). And this is all by pure chance! Flip a coin! No predictive value whatever! (BTW, this is why looking at 2 week periods is such a bad idea. There are 52 (staggered) 2 week periods in a year, so you'll find 1.6 green column on average, just by chance.)

Now things get worse. A lot of stocks load the coin, for they have grown over that period. Which means that their (geometrical) average monthly result is greater than 0. Which means that they have a better chance than 1 in 2 for a positive month, and a much better chance than 1 in 32 for a 5 year green column. The fact that a stock is good (=makes you a profit) makes for a greater chance for a green column, by pure chance alone. Hey wait, that's not so very bad! You could turn that around and say: When you find a stock with a green column, there is a fair chance that it has that column because it loaded the coin, and thus is a good stock.

Is Ken just finding good stocks? No, but probably a good deal of them are just good stocks. A lot of his stocks are noise: some bad stocks that got lucky, average stocks have a chance of better than 1 in 3 to come aboard, and good stocks an even better chance. But all these stocks are just noise for the seasonality pattern. How do we find the real patterns? Enter the tweak: unload the coin! First of all, create a barrier: don't look at positive returns, but at returns over, say 3% (or 5%) monthly. Those are True Green, as opposed to Dappled Green with all the noise mixed in. That will weed out most of the noise. But you might also look at the (geometrical) average monthly return of a stock, and if it is higher than 3% (or 5%), require a higher score than the average for a True Green month.

And remember, this doesn't weed out all chance! You just can't do that. (Except by raising the bar so high that nothing will be able to jump over it.)

Just an idea.

Regards,

Karel