The Motley Fool Discussion Boards

Previous Page

CAPS / CAPS Feedback

URL:  http://boards.fool.com/the-problem-is-that-if-one-is-better-than-average-25172822.aspx

Subject:  Re: CAPS is not meant to be a mirror of performa Date:  2/14/2007  12:51 PM
Author:  woof321 Number:  4900 of 8287

The problem is that if one is better than average at picking stocks, then the system encourages making as many picks as possible. When I first started playing, I assumed, like most I suspect, that I should only enter my "best" picks. Adding a bunch off extra picks that in aggregate would beat the market would give me a few extra points but they would dilute my accuracy, so there'd be a trade-off. I soon realized that that isn't how this game works, however.

In baseball, we look at both quantitative (H, HR, RBI) and qualitative (BA, OBP, SLG) metrics when considering hitting performance. A hitter that often swings at bad pitches might add a few hits to his total, but will lower his qualitatives - sometimes it's better to just take the walk. There's a trade-off.

Not so in CAPS, unfortunately. Points seem to work fine as a quantitative metric. Accuracy was intended to measure quality, but fails. Instead, by letting players decide when to close positions, accuracy can be "banked" and as such it is essentially a second quantitative. In September we were all marvelling at the leaders' 70% accuracies. Now they have over 75%. They'll soon have over 80%. Qualitative measures should not be increasing over time unless skill is improving. Here's turbotrager's accuracy over 5 months:

October 54%
November 61%
December 64%
January 66%
February 69%

CAPS5star1star picked stocks the same way, but didn't bank. Has a higher score, but an accuracy that does not increase over time. Mindless banking is worth over 3000 positions in ranking in this case. Since both points and accuracy are working like quantitatives, a good strategy is to always have 200 picks open and accumulate both accuracy and points over time. As more people realize this, it'll become even more common, and player and stocks ranking even less meaningful. I've tried to answer all the counter arguments to my case, even "why don't you prove it?". Seems the only counter argument left is "let's wait a few years and see what happens". Well, the good lord gave us mathematics so we wouldn't have to, and the longer you wait the harder it will be to change without overly penalizing those that "played by the (broken) rules".

One argument that still comes up a lot is this:

If someone can very consistently pick stocks that are going to outperform the S&P by 5% or more it's time to start a hedge fund. Go long the individual stocks and short the S&P and pocket that 5%.

Oh no, that's not the way. And you're not listening to what I say! The whole point of my argument is that it doesn't matter how often you beat the market if you don't also look at the sizes of the gains versus the losses. Would you invest in a hedge fund that had a 100% loss for every nineteen 5% wins? That's 95% accuracy, right? It's also called losing to the market. Better a fund that has a 100% win for every nine 5% losses. Only 10% accuracy but a much better place for your money. I know it's hard to accept, but "accuracy", as being measured here, is meaningless. Sorry.

Remember the roulette truncated martingale from high-school statistics? It works like this - go into a casino and put a $1 bet on black. If you lose, bet $2. Lose again, bet $4. If you eventually hit, you'll have won $1 (for example, if you hit on the fourth spin, you win $16, but have bet 1+2+4+8=$15, for a $1 net gain) at which point you can "re-up" by walking out to the parking lot and coming back through the door. Now repeat. Excellent accuracy, right? Occassionally you'll lose all your money, but most of the time you'll win $1. Instead if you decide not to be greedy and just walk out any time you lose a $32 bet, you'll be 98% accurate and can probably do this for years. Is your near-perfect accuracy meaningful? Of course not - you lose so much on your rare losses t