From: Gary Miller (garymiller@starband.net)
Date: Fri Jun 13 2003 - 11:52:24 MDT
It seems like in investing as horse racing, recent data points should be
higher in value than events which took place longer ago. For example a
stock or horses recent performance should be a better indicator of
future performance than how it ran two years ago.
Is there any theory on the correct incorporation of weighting for the
temporal data in the model creation. I think that ignoring the temporal
aspect of the data series is why many models fail.
-----Original Message-----
From: owner-extropians@extropy.org [mailto:owner-extropians@extropy.org]
On Behalf Of gts
Sent: Thursday, June 12, 2003 10:10 AM
To: extropians@extropy.org
Subject: RE: Investing
Dennis Fantoni wrote:
> You mentioned a statistical test of a trading system. My plan is to
> divide my historic data into several sets, and then test and build my
> systems using only one set of data.
Good. You are thinking correctly here. You can use one set to
"curve-fit" the data, and another set to test your curve-fit to
determine whether it persists through time. No real need for a third set
of data. It is important that your test be for statistical significance
rather than only for profit/loss. If you don't know stat then you should
learn it. The basic tests are not too complicated.
> Fun, that You mention gambling (on the wrong side of the
> counter) as a possible venue of income. I have found several
> non-effective markets in gambling and have profited from a few,
> although not much money. If it is possible to get a positive income
> from gambling , it should be possible to live on investing too, given
> a reasonable amount of capital for investment. Or perhaps the stock
> markets are effective but gambling markets are not? If so - how can
> that be?
In the early 80's I made my living counting cards at the blackjack
tables in Reno and Las Vegas. I played cards for an entire year, seven
days per week. It was my business. In fact I decided to become an
investment advisor while sitting at a blackjack table in Reno. I was
ahead for the night by about $5,000. The dealer asked me what I would do
with my winnings if I were to win $50,000. I said, "Well, I guess I
would invest it in the stockmarket." The dealer asked if I knew anything
about the market. I answered, "No, not really, but now that you mention
it I really think I should learn about the stockmarket. I think I would
be good at it." I went home and started studying to get my series 7
general securities broker license. The rest is history. I was a broker
for 12 years.
Blackjack is the only beatable game in the casino, so don't waste your
time trying to find gambling systems for other games. Even blackjack is
difficult now as the casinos have become wise to card-counting. Even
back then, 20 years ago, I was barred from playing cards in six casinos.
The casino bosses would tap me on the shoulder and say "Excuse me, sir,
but we have decided that we no longer wish to have you as a blackjack
customer. You're welcome to play slots or craps or any other game but
blackjack." They're even tougher now than they were then. In fact it is
almost impossible now even to find a beatable game; the pit bosses and
some of the dealers are trained in card-counting so they can detect it
more easily and evict the counters before they make any money. They also
shuffle early and use other tactics to thwart the counters.
As I mentioned my statistical project these days is horse racing. I read
an interesting research report in a statistics journal about an apparent
inefficiency in paramutual betting at the track: it seems that the most
boring and conservative bets, (show bets on heavy favorites), have
bigger pay-offs than they should have in an efficient market for bets.
The statistics suggest that novices tend to bet on long shots for the
sheer thrill and entertainment value, creating an opportunity for
professionals to make a profit on the favorites.
-gts
This archive was generated by hypermail 2.1.5 : Fri Jun 13 2003 - 12:03:49 MDT