home *** CD-ROM | disk | FTP | other *** search
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!saimiri.primate.wisc.edu!ames!ncar!hsdndev!husc-news.harvard.edu!zariski!grabiner
- From: grabiner@math.harvard.edu (David Grabiner)
- Newsgroups: rec.games.bridge
- Subject: Re: Which inference is better, WAS - "finesse or play for the drop"
- Message-ID: <GRABINER.92Nov19221731@boucher.harvard.edu>
- Date: 20 Nov 92 03:23:39 GMT
- References: <1992Nov16.131237.19210@ms.uky.edu><lsimonse.722011812@vipunen.hut.fi><BxwHCq.
- 55K@irvine.com><1ee8ukINN8ij@agate.berkeley.edu><1992Nov18.224411.4092@u.washi
- ngton.edu><GRABINER.92Nov19121848@boucher.harvard.edu> <davidm.722213002@questor>
- Organization: /etc/organization
- Lines: 67
- Nntp-Posting-Host: boucher.harvard.edu
- In-reply-to: davidm@questor.Rational.COM's message of 19 Nov 92 22:43:22 GMT
-
- In article <davidm.722213002@questor>, David Moore writes:
-
- > grabiner@math.harvard.edu (David Grabiner) writes:
-
- Please check your attributions: I didn't make either of the statements
- below. Eddie Grove made the first one, and Bryant Fujimoto made the
- second.
-
- >>>>This is wrong, of course. Unless you use a perfect random source for
- >>>>your shuffles, the conditional distribution of the opponents' cards
- >>>>will tend to depend upon your cards.
-
- >>> Would someone please explain this last point more fully? I am afraid
- >>> it isn't obvious to me.
-
- > Try this experiment. Sort a deck of cards into order (2C 3C .... KS AS) and
- > then shuffle them. Now count the number of times a card is followed by the
- > next higher card; that is, they have "stuck together" during the shuffle.
-
- > If the shuffle was totally random, the average number of such occurences in
- > a deck would be 1. In practice, the number will be quite high, even if you
- > make a good attempt at shuffling.
-
- It won't be 1 unless you shuffle the deck several times. Even a good
- riffle shuffle (randomly merging two piles) will break only half the
- pairs. Seven shuffles are needed to come close to randomizing a deck,
- if the previous order is fixed.
-
- > Let's suppose you get 10 such pairs.
-
- That's probably a bit too many for the average shuffler; three normal
- shuffles might be expected to leave about 10 pairs, but many people
- shuffle more than three times.
-
- > Let us also suppose that the
- > probability of the K being over the Q in a pack is 10% (before the
- > shuffle).
-
- I don't think this is likely. If the deck came fresh from the factory,
- was sorted by hand, or was just used for a winning game of solitaire,
- the probability that the king is over the queen is either 0% or 100%.
- But in a deck that was last used for a game of rubber bridge, the main
- reason that would make the king would lie over the queen would be if it
- covered a queen on the last deal. I don't think that a king covers a
- queen on 1/3 of the 25% of hands that have the king lying over the
- queen.
-
- > Then the probability of a finess working is raised to
- > roughly 20%*10% + 50%=52%, and the probabiliuty of it failing is
- > lowered to 48%.
-
- This calculation is incorrect. If the king lies over the queen, and
- they aren't split, the king definitely lies over the queen. Otherwise,
- it is 50-50, but you only get 50% of the *remaining* cases, which happen
- 98% of the time by your figuring. You get .02 + .5*.98=.51, and even
- closer to .50 if you make more reasonable assumptions.
-
- And don't forget that 2-2 breaks are also more likely in a poorly
- shuffled deck; if the spade trick containing the ace stayed together as
- a clump of four cards, it's better than 50-50 that the other two spades
- are split. You should still play for the drop.
-
- --
- David Grabiner, grabiner@zariski.harvard.edu
- "We are sorry, but the number you have dialed is imaginary."
- "Please rotate your phone 90 degrees and try again."
- Disclaimer: I speak for no one and no one speaks for me.
-