Bruce Momjian wrote:
> Greg, tell me you didn't get involved with Postgres because of Poker
> Tracker. :-O :-)
>
Nah, both came out of my working on stock trading systems. I just wrote
a bit more about this whole subject at
http://blog.2ndquadrant.com/en/2011/07/pushing-allin-with-postgresql.html
if anyone would like to see what the actual queries look like against
the Poker Tracker database.
One of the things you have to come to grips with when writing a trading
system is that you can never assume something won't happen just because
it's really unlikely. Traders have adopted the term "black swan" to
talk about these sort of completely unexpected things that destroy their
bankroll. On-line poker is quite good at teaching you this lesson
quickly, whereas traders might go a decade or more before encountering a
similarly unlikely event.
Lose an ace-high flush to a straight flush; lose 4 of a kind to a higher
4 of a kind; rare, but I watched them both happen to me on multiple
times when playing. If I'd bet so much that I'd have been wiped out by
either event, even though I was *sure* I would win that hand, I'd have
been done for. And a couple of times, I've seen the most rare thing of
all: the initial 3 cards come out, and I have a hand where I can only
lose if the other player gets the exact two cards they need. For example:
Greg: KK
LuckyBum: QQ
Flop: K82
The only way the other player can win here is if the remaining cards are
QQ, giving them 4 of a kind over my full house at the end of the hand.
Assuming no other players are involved, that's 45 cards left in the
deck, and the odds of them getting one of those two followed by the
other are 2/45 * 1/44. And yet I've lost this sort of 990:1 long shot
multiple times. It definitely gives you a better gut feel for
"gambler's ruin", one that translates back into stock trading--and into
thinking about how to really achieve high-availability for a computing
system, too.
--
Greg Smith 2ndQuadrant US greg@2ndQuadrant.com Baltimore, MD
