T.R | Title | User | Personal Name | Date | Lines |
---|
1298.1 | Solution hidden behind form feed: | CHOVAX::YOUNG | Cooper Bullet-Proof Vests, Inc. | Fri Sep 21 1990 14:21 | 13 |
|
The first player should guess either 1 or 10.
Reasoning:
1 or 10 has the same chance of being right as any other number and
has the advantage that, if wrong, it reveals NO information to your
opponent. Since any information ("higher or lower") revealed by an
incorrect guess can be acted on by your opponent before you, you
should try to reveal as little information as possible.
-- Barry
|
1298.2 | Probability of Player 1 Winning | VAXRT::BRIDGEWATER | Blasting out of the past. | Fri Sep 21 1990 15:20 | 19 |
| If both players play optimally, the probability of player 1 winning
is:
1/2
Guess # Probability of Player 1 winning on this guess
--------------------------------------------------------
1 (1/10)
2 (9/10)*(8/9)*(1/8)
3 (9/10)*(8/9)*(7/8)*(6/7)*(1/6)
4 (9/10)*(8/9)*(7/8)*(6/7)*(5/6)*(4/5)*(1/4)
5 (9/10)*(8/9)*(7/8)*(6/7)*(5/6)*(4/5)*(3/4)*(2/3)*(1/2)
Each of the above probabilities is equal to 1/10. Summing them
gives the total probability of player 1 winning, 1/2.
- Don
|
1298.3 | | TRACE::GILBERT | Ownership Obligates | Fri Sep 21 1990 15:39 | 13 |
| Let the number of numbers be n. If n is even, it's a fair game
(expectation=1/2), and choosing any number yields an optimal strategy.
If n is odd, an optimal strategy is to choose any odd ordinal number
(the first, third, fifth, etc), and this yields an expectation of (n+1)/(2n).
Now, ...
Assume that the players *must* choose some reasonable number; for example,
if the number is known to be > 3 and < 7, the player must choose 4, 5, or 6.
Is it possible to randomly choose the initial number (i.e., with a non-uniform
probability distribution) so that the *second* player gets the advantage?
- Gilbert
|
1298.4 | Proof of previous. | CADSYS::COOPER | Topher Cooper | Fri Sep 21 1990 18:37 | 56 |
| RE: .3 (Gilbert)
>Let the number of numbers be n. If n is even, it's a fair game
>(expectation=1/2), and choosing any number yields an optimal strategy.
>If n is odd, an optimal strategy is to choose any odd ordinal number
>(the first, third, fifth, etc), and this yields an expectation of (n+1)/(2n).
Since you got it in before I did, let me present the proof which you
"left as an exercise".
Assume that N is even and the statement is true for all smaller N.
Call the winning number W.
Choose a as the initial choice.
There are three possibilities: 1) W = a (prob=1/N) 2) W < a (prob =
(a-1)/N) and 3) W > a (prob = (N-a)/N).
If a is even, then the segment below a is of odd length, so if W is
there the second player has the advantage with a probability of
a/(2a-2) of winning. This means that the first player has a chance
of (a-2)/(2a-2) of winning if W < a. The segment above a is of even
length so the probability of the first player winning under those
conditions is simply 1/2. Total probability of the first player
winning is then:
a-2 a-1 1 1 N-a
---- * --- + - + - * --- = 1/2
2a-2 N N 2 N
If a is odd, then it works out to:
1 a-1 1 N-a+1 N-a
- * --- + - + ----- * --- = 1/2
2 N N 2N-2a N
If N is odd and a is even, both segments remaining after a has been
selected are odd, so:
a-2 a-1 1 N-a+1 N-a N-1
---- * --- + - + ----- * --- = --- < 1/2
2a-2 N N 2N-2a N 2N
If N is odd and a is odd, both segments remaining afterr a has been
selected are even, so:
1 a-1 1 1 N-a N+1
- * --- + - + - * --- = --- > 1/2
2 N N 2 N 2N
Base cases: If N is 1, the first player wins with probability (N+1)/2N
= (1+1)/(2*1) = 1. If N is 2, the first player has 1 chance in 2 of
picking the correct choice (it was not actually necessary to use this
base case -- it just was simple and more comfortable).
Topher
|
1298.5 | | TRACE::GILBERT | Ownership Obligates | Mon Sep 24 1990 11:31 | 17 |
| Rewording the problem in .3:
Assume that the players *must* choose some reasonable number; for example,
if the number is known to be > 3 and < 7, the player must choose 4, 5, or 6.
Is it possible to choose the initial number with a non-uniform probability
distribution (which is known to both players) so that the *second* player
gets the advantage?
re .1
> 1 or 10 has the same chance of being right as any other number and
> has the advantage that, if wrong, it reveals NO information to your
> opponent.
Actually, it may reveal that "the answer is not 1" or "... not 10". And
reply .4 shows that this is important.
|