Forum Archive :
Match Play at 2away/2away
Proof for doubling immediately

Here is a proof that doubling on one's first turn is a version of optimal
play when at 2 away 2 away match score. It supposes optimal play from
opponent (otherwise one can possibly do better with technically incorect
play).
First of all suppose that there is no market loser possibility before
one's first chance to turn cube. In other words, it is impossible to have
winning chances less than X or greater than 1X when you first have option
to turn cube, where X:= equity when at 2 away 1 away crawford. (It is
generally accepted that X is approximately 30%.)
This is virtually impossible to prove rigorously, but i believe no one
would argue that it is false. If we accept this as a fact then the
following proof works:
By symmetry, one's chance of winning the match assuming perfect play is
50%. Consider the strategy "double at first opportunity and play checkers
optimally for a 1 point match". Since by our above "fact" the double will
always be accepted, the winner of match is winner of game. Probability of
winning the game is 50%. Thus our strategy attains the equity 50% and is a
version of optimal play.
QED
Note that there are other verions of optimal play. The theorem "optimal
play never requires a double if no market losing sequence" together with
the above theorem implies that "double if and only if there is a market
losing sequence" is also a version of optimal play.
Robert Koca (bobk on FIBS)
koca@orie.cornell.edu


Bob Johnson writes:
Hmmmm. Suppose no one doubles for awhile, since the game
seems about even.
Later, my opponent has a miracle roll, or maybe I roll something bad.
Suppose my opponent suddenly has more than 70% advantage; I can drop a
cube, and have a chance to win the match. If, on the otherhand, the
cube was already at 2, then I might feel quite stupid for having doubled
early. I don't like feeling stupid. (I wonder if that measures in?)
Or  the opposite happens and I can unexpectedly and suddenly
double my opponent out of the game. I now have 70% chance for the match.
I am no expert, but I tend to feel something can be gained by waiting so
a more informed decision can be made, rather than immediately basing the
outcome of the match on one game.


Igor Sheyn writes:
I like to double at 2away2away when I am (slight) favorite with maybe
12 market losing SEQUENCES ( since it's hard to lose your market just
on your roll in the very beginning ) and with developing solid threats
down the road. I will NOT double if I am an underdog, or if I have
moderate gammon chances. I don't take the strength of my opponent into
account, as I play only good players on FIBS, and in real life
majority players in tournaments are good by default. I'll probably
wait though if I happen to play a fish. I'd like to hear some more
opinions on this, since the only expert one I've heard so far is Kit's
in his book, and I am still not quite clear on it.
Igor


Kit Woolsey writes:
OK, Igor, let me present it in the following proposition: You and I are
sitting down to play a 2point match. Before we start, I will promise
that I will turn the cube if there is any 2roll sequence by which I can
lose my market next turn.
My claim is that if you adopt any different strategy which involves
risking losing your market, then you have the worst of it. Proof: It is
clear that I will never lose my market  i.e. no position can ever arise
where I will double and you will (correctly) pass, since I have
guaranteed that if this is at all possible I will have doubled on the
previous turn. If it ever happens that you fail to double and lose your
market then you have done less than optimally, since any time you lose
your market you would have done better doubling the turn before.
Therefore, if you adopt any strategy which risks losing your market you
will be at a disadvantage playing me when I adopt the strategy of never
risking losing my market. QED
Kit Woolsey


Albert Steg writes:
I cannot agree with this argument because it begs the question. That is,
in seeking to prove the proposition, it uses the proposition itself as an
axiom. I have asked you to show that it is correct to double at the
prospect of any marketlosing parley, and you have effectively answered
"because my opoponent will surely double with any marketlosing parlay,
unless he plays imperfectly."
What you HAVE proved is that against anyone who is certain to employ your
advice, it is best to follow your advice.
You have also demonstrated, though, that this certainty cannot be assumed.
After all, you yourself do not employ the strategy against players whom you
suspect will not follow your strategy!
As you yourself point out, against an opponent who does not double
at the first marketlosing opportunity, you may do better to back off from
your own advice also. I see no reason for saying then that your opponent
is playing "imperfectly."
Albert


Darse Billings writes:
This discussion seems to be going back and forth... perhaps I can clarify.
It is a simple exercise to show that an automatic double is the game
theoretic optimal play for 2away 2away. The first time I saw this
posted, it seemed so obvious that it didn't even warrant discussion.
*Optimal* play, in game theory, assumes the opponent will play correctly.
The doubling result follows directly from symmetry.
*Maximal* play, in practice, often exploits suboptimal play by the
opponent.
Hence, in *theory* you can do no better than doubling. In *practice*, you
might wait before doubling a weak opponent. Kit has explained this
distinction rather well.
Cheers,  Darse.




Match Play at 2away/2away
 Basic strategy (Darse Billings, Feb 1995)
 Counterexample? (Jim Williams+, Mar 1998)
 Do you need an advantage to cube? (Keene Marin+, Feb 2006)
 Double immediately? (Chuck Bower, Oct 1998)
 Ever too good to double? (Kit Woolsey, July 1995)
 Minimum game winning chances to double (Walter Trice, Mar 1999)
 Practical strategy (Walter Trice, July 1995)
 Practical strategy (Albert Steg+, Feb 1995)
 Proof for doubling immediately (Robert Koca+, May 1994)
 Proof of doubling with market losers (Walter Trice+, July 2001)
 Sample game (Ron Karr, Dec 1996)
From GammOnLine
Long message
Recommended reading
Recent addition

 
