由买买提看人间百态

boards

本页内容为未名空间相应帖子的节选和存档,一周内的贴子最多显示50字,超过一周显示500字 访问原贴
Quant版 - 大家来扔硬币-an interview question
相关主题
[合集] 请教两个题目?A probability Q
An interview question of unfair coin's stop timetoss a coin 100 times, what's the probability at least 60 of them are heads?
old probability Q[合集] Probability Question
一道概率题,大家看看[合集] probability 问题
[Markov Chain] 世界矿工的一道笔试题another interesting probability question
问一个面试题,关于概率的A probability question
[合集] a question问一个mathproblems上的coin toss问题
[合集] 一道概率题,被问倒了。一道概率题
相关话题的讨论汇总
话题: pr话题: win话题: 硬币话题: 50%话题: case
进入Quant版参与讨论
1 (共1页)
p********e
发帖数: 3
1
再次在扔硬币问题面前倒地,发来听听各位大牛的idea
A,B两人比赛,A有51个硬币,B有50个硬币,独立的扔硬币,当得到的Head多时则获胜
,那么请问A的概率?
当时给出了条件概率的公式,但是无法计算啊,是不是应该有什么巧妙的办法呢?求助
r*******s
发帖数: 303
2
1vs1
d*j
发帖数: 13780
3
假设 A也有50个硬币, 这里用对称性
然后 condition on 最后一个
答案是 50%

【在 p********e 的大作中提到】
: 再次在扔硬币问题面前倒地,发来听听各位大牛的idea
: A,B两人比赛,A有51个硬币,B有50个硬币,独立的扔硬币,当得到的Head多时则获胜
: ,那么请问A的概率?
: 当时给出了条件概率的公式,但是无法计算啊,是不是应该有什么巧妙的办法呢?求助

p********e
发帖数: 3
4
都是50的时候也不是50%,因为还有平的时候
p********e
发帖数: 3
5
我觉得就是50vs50的时候,一方赢的概率也不好算呢
d*j
发帖数: 13780
6
恩, 但是加了一个就可以了, 呵呵
假设都是50个, P(A win) + P(B win) + P(Eq) = 1
这里 P(A)= P(B)
然后 A 多了一个, 所以 condition on 最后一个, 对于 P(A win) or P(B win), 最
后一个无所谓, 对于P(Eq), 如果最后一个是 H, A win
所以最后A赢的概率是 P(A) + 1/2 * P(Eq)
看看是不是第一个式子除以 2

【在 p********e 的大作中提到】
: 我觉得就是50vs50的时候,一方赢的概率也不好算呢
r*******s
发帖数: 303
7
head赢得概率和tail赢得概率是一样的。所以是50%
d*j
发帖数: 13780
8

因为数量不同, 所以或者这个多 或者那个多, mutually exclusive
所以他们概率和为一

【在 r*******s 的大作中提到】
: head赢得概率和tail赢得概率是一样的。所以是50%
a**m
发帖数: 102
9
Pr[H(A)>H(B)] = Pr[T(A)>T(B)] = Pr[H(A)<=H(B)]
and
Pr[H(A)>H(B)] + Pr[H(A)<= H(B)]=1.
Therefore, Pr[H(A)>H(B)] = 1/2

【在 p********e 的大作中提到】
: 再次在扔硬币问题面前倒地,发来听听各位大牛的idea
: A,B两人比赛,A有51个硬币,B有50个硬币,独立的扔硬币,当得到的Head多时则获胜
: ,那么请问A的概率?
: 当时给出了条件概率的公式,但是无法计算啊,是不是应该有什么巧妙的办法呢?求助

J****g
发帖数: 103
10
看不懂。。。 请问这个跟random walk,或者gambler: 51/(51+50)有什
么区别吗?

【在 a**m 的大作中提到】
: Pr[H(A)>H(B)] = Pr[T(A)>T(B)] = Pr[H(A)<=H(B)]
: and
: Pr[H(A)>H(B)] + Pr[H(A)<= H(B)]=1.
: Therefore, Pr[H(A)>H(B)] = 1/2

相关主题
问一个面试题,关于概率的A probability Q
[合集] a questiontoss a coin 100 times, what's the probability at least 60 of them are heads?
[合集] 一道概率题,被问倒了。[合集] Probability Question
进入Quant版参与讨论
a**m
发帖数: 102
11
H(A) is the number of head person A tosses, T(A) is the number of tail
person A tosses.
The first equality is by symmetry of head and tail. The second one is just because that A has exactly one more coin than B has.

【在 J****g 的大作中提到】
: 看不懂。。。 请问这个跟random walk,或者gambler: 51/(51+50)有什
: 么区别吗?

m********0
发帖数: 2717
12
适用的。

【在 p********e 的大作中提到】
: 我觉得就是50vs50的时候,一方赢的概率也不好算呢
q********u
发帖数: 53
13
为什么 P(A win) or P(B win), 一种可能B比A多一个H。最后一个A如果是H,应该变成
Eq。



【在 d*j 的大作中提到】
: 恩, 但是加了一个就可以了, 呵呵
: 假设都是50个, P(A win) + P(B win) + P(Eq) = 1
: 这里 P(A)= P(B)
: 然后 A 多了一个, 所以 condition on 最后一个, 对于 P(A win) or P(B win), 最
: 后一个无所谓, 对于P(Eq), 如果最后一个是 H, A win
: 所以最后A赢的概率是 P(A) + 1/2 * P(Eq)
: 看看是不是第一个式子除以 2

d*j
发帖数: 13780
14
那就是打平了, A还是没有赢, 不用考虑了

【在 q********u 的大作中提到】
: 为什么 P(A win) or P(B win), 一种可能B比A多一个H。最后一个A如果是H,应该变成
: Eq。
:
: 最

J****g
发帖数: 103
15
这种情况怎么算呢:before A tosses 51st coin. B win & B had one more H than A
. Now A tosses and the 51st coin shows H, which means Eq.

【在 d*j 的大作中提到】
: 那就是打平了, A还是没有赢, 不用考虑了
d*j
发帖数: 13780
16
不用算啊, 不是说了吗, 这样最后是平了, 我们需要的是A赢
所以只有 EQ + A 最后一个 coin 是 H, 对我们有帮助

A

【在 J****g 的大作中提到】
: 这种情况怎么算呢:before A tosses 51st coin. B win & B had one more H than A
: . Now A tosses and the 51st coin shows H, which means Eq.

X***0
发帖数: 45
17
坑主 对题目 理解不清 阿
当一方抛出的head比另一方多,游戏就结束,可能抛到第3枚 B就win了, 也许A win
所以 51对50 真正对 A 有利的情形 只可能发生在 前50次 A B 抛的head 均等,而第
51次 A投出的head的几率仍是50%
而之前 A B win 的机会 均等
所以 总的来说 还是 50%
a**m
发帖数: 102
18
no! what is for comparing is the total tossing of heads.

【在 X***0 的大作中提到】
: 坑主 对题目 理解不清 阿
: 当一方抛出的head比另一方多,游戏就结束,可能抛到第3枚 B就win了, 也许A win
: 所以 51对50 真正对 A 有利的情形 只可能发生在 前50次 A B 抛的head 均等,而第
: 51次 A投出的head的几率仍是50%
: 而之前 A B win 的机会 均等
: 所以 总的来说 还是 50%

J****g
发帖数: 103
19
恩, 谢谢谢谢, 刚回去又重新读了一遍题。

【在 d*j 的大作中提到】
: 不用算啊, 不是说了吗, 这样最后是平了, 我们需要的是A赢
: 所以只有 EQ + A 最后一个 coin 是 H, 对我们有帮助
:
: A

m*****a
发帖数: 636
20
我是不是漏了什么东西。
这个证明似乎在 50 VS 50 的时候也适用。
但50VS50的case, 按daj一开始给的解释是小于50%的

Pr[H(A)>H(B)] = Pr[T(A)>T(B)] = Pr[H(A)<=H(B)]
and
Pr[H(A)>H(B)] + Pr[H(A)<= H(B)]=1.
Therefore, Pr[H(A)>H(B)] = 1/2

【在 a**m 的大作中提到】
: Pr[H(A)>H(B)] = Pr[T(A)>T(B)] = Pr[H(A)<=H(B)]
: and
: Pr[H(A)>H(B)] + Pr[H(A)<= H(B)]=1.
: Therefore, Pr[H(A)>H(B)] = 1/2

相关主题
[合集] probability 问题问一个mathproblems上的coin toss问题
another interesting probability question一道概率题
A probability questioncoin flip终极版
进入Quant版参与讨论
p*****k
发帖数: 318
21

the punchline is Pr[T(A)>T(B)] = Pr[H(A)<=H(B)], which holds only
when A has exactly one more coin than B.
if B has at least as many coins as A, then T(A)>T(B) implies H(A) [they cannot be equal unless T(A)=T(B)], so the prob of A winning
is < 50%;
similarly, if A has at least two more coins than B, the case T(A)>T(B)
includes the possibility of H(A)>H(B), so the prob of A winning is > 50%

【在 m*****a 的大作中提到】
: 我是不是漏了什么东西。
: 这个证明似乎在 50 VS 50 的时候也适用。
: 但50VS50的case, 按daj一开始给的解释是小于50%的
:
: Pr[H(A)>H(B)] = Pr[T(A)>T(B)] = Pr[H(A)<=H(B)]
: and
: Pr[H(A)>H(B)] + Pr[H(A)<= H(B)]=1.
: Therefore, Pr[H(A)>H(B)] = 1/2

m*****a
发帖数: 636
22
嗯,懂了,谢谢
解释的清楚

the punchline is Pr[T(A)>T(B)] = Pr[H(A)<=H(B)], which holds only
when A has exactly one more coin than B.
if B has at least as many coins as A, then T(A)>T(B) implies H(A) [they cannot be equal unless T(A)=T(B)], so the prob of A winning
is < 50%;
similarly, if A has at least two more coins than B, the case T(A)>T(B)
includes the possibility of H(A)>H(B), so the prob of A winning is > 50%

【在 p*****k 的大作中提到】
:
: the punchline is Pr[T(A)>T(B)] = Pr[H(A)<=H(B)], which holds only
: when A has exactly one more coin than B.
: if B has at least as many coins as A, then T(A)>T(B) implies H(A): [they cannot be equal unless T(A)=T(B)], so the prob of A winning
: is < 50%;
: similarly, if A has at least two more coins than B, the case T(A)>T(B)
: includes the possibility of H(A)>H(B), so the prob of A winning is > 50%

h*****u
发帖数: 38
23

这个思路最好

【在 r*******s 的大作中提到】
: head赢得概率和tail赢得概率是一样的。所以是50%
d*j
发帖数: 13780
24


【在 h*****u 的大作中提到】
:
: 这个思路最好

w**********y
发帖数: 1691
25

问你个问题哈..
如果A有100个硬币,B有1个硬币.
head和tail赢的概率一样么?
A赢的概率还是50%么?then.......

【在 r*******s 的大作中提到】
: head赢得概率和tail赢得概率是一样的。所以是50%
t*******e
发帖数: 172
26
I prefer
sum (n)(n+1)= sum (n)(n+1)
(i)(j ) (i)(j )
i=j
and the sum of both is 2^{2n+1}
l*********o
发帖数: 3091
27
对daj解法的疑问:
daj的解法实际是AB同时flip coin,平则继续,一方胜出则停止。
经过50轮还平的话,A有一次单独flip的机会。赢的概率50%.没有平的机会。
试想按daj的叙述:on the other side, 如果最后一个是 T, B win
所以daj最终的结果是P(Awin_final) = P(Bwin_final) = 50%
l*********o
发帖数: 3091
28
H:head
for A n+1 toss,
(n+1)H nH (n-1)H ... 0H
C(n+1,0) C(n+1,1) C(n+1,2) ... C(n+1,n+1)

for B n toss,
nH (n-1)H (n-2)H ... 0H
C(n,0) C(n,1) C(n,2) ... C(n,n)
total_A_win_case=
C(n+1,0)*(C(n,0)+C(n,1)+C(n,2)+...+C(n,n))
+C(n+1,1)*(C(n,1)+C(n,2)+...+C(n,n))
+...
+C(n+1,n)*C(n,n)
A_win_probability = total_A_win_case/2^n/2^(n+1)
p*****k
发帖数: 318
29
not to beat it to death, but for daj's approach, note after 50 tosses
of each player, if A is behind with # of H's, he can never win with
the extra toss (the best case is a tie); on the other hand, if A is
already ahead, he will surely win. so half-half split here due to
the symmetry.
the only case left to examine is tie so far, for which case, A wins
again with half chance (another half results tie).
btw, there is a closed-form solution for the chance of tie:
C(2n+1,n)/2^(2n+1)
the easiest way
1 (共1页)
进入Quant版参与讨论
相关主题
一道概率题[Markov Chain] 世界矿工的一道笔试题
coin flip终极版问一个面试题,关于概率的
coin toss 题[合集] a question
有两道题目求解[合集] 一道概率题,被问倒了。
[合集] 请教两个题目?A probability Q
An interview question of unfair coin's stop timetoss a coin 100 times, what's the probability at least 60 of them are heads?
old probability Q[合集] Probability Question
一道概率题,大家看看[合集] probability 问题
相关话题的讨论汇总
话题: pr话题: win话题: 硬币话题: 50%话题: case