t***l 发帖数: 3644 | 1 公平硬币一枚,一次次的扔,想停就停,等停的时候获得bonus如下:掷出正面的次数
除以掷硬币的次数。问什么时候该停?
顺便问一下,金人性事的桌子策略员容易申请吗? | s*******s 发帖数: 1568 | 2 金人性事?去sex版问问?lol
【在 t***l 的大作中提到】 : 公平硬币一枚,一次次的扔,想停就停,等停的时候获得bonus如下:掷出正面的次数 : 除以掷硬币的次数。问什么时候该停? : 顺便问一下,金人性事的桌子策略员容易申请吗?
| t***l 发帖数: 3644 | 3 被大侠嘲笑了,我说的是goldman sachs,估计sex版的人不懂的。
【在 s*******s 的大作中提到】 : 金人性事?去sex版问问?lol
| s**e 发帖数: 1834 | 4 This is still an open problem, only some asymptotic solution
is available.
http://arxiv.org/PS_cache/arxiv/pdf/0907/0907.0032v2.pdf
【在 t***l 的大作中提到】 : 公平硬币一枚,一次次的扔,想停就停,等停的时候获得bonus如下:掷出正面的次数 : 除以掷硬币的次数。问什么时候该停? : 顺便问一下,金人性事的桌子策略员容易申请吗?
| b******n 发帖数: 637 | 5 为什么不是在最早H大于T的时候停呢?
【在 s**e 的大作中提到】 : This is still an open problem, only some asymptotic solution : is available. : http://arxiv.org/PS_cache/arxiv/pdf/0907/0907.0032v2.pdf
| s**e 发帖数: 1834 | 6 Mainly because by keep tossing, your down side is covered by at least 1/2.
【在 b******n 的大作中提到】 : 为什么不是在最早H大于T的时候停呢?
| b******n 发帖数: 637 | 7 这就是我为什么尽量早停的原因。
1/2.
【在 s**e 的大作中提到】 : Mainly because by keep tossing, your down side is covered by at least 1/2.
| s**e 发帖数: 1834 | 8 Here is an example that why you should not always stop at H > T.
Suppose (H,T)=(51,50). If stop now, the ratio is 51/101.
But if you keep tossing another 2 times, you will get:
(1) (H,T)= (53,50), ratio = 53/103 (with probability 1/4)
(2) (H,T)= (52,51), ratio = 52/103 (with probability 1/2)
(3) (H,T)= (51,52), ratio = 51/103 (with probability 1/4)
However, in the case (3) above, you can keep tossing,
and can achieve at least 1/2.
Here we have
53/103*1/4 + 52/103*1/2 + 1/2 * 1/4 > 51/101.
So we should not stop when (H,T)=(51,50).
【在 b******n 的大作中提到】 : 这就是我为什么尽量早停的原因。 : : 1/2.
| b******n 发帖数: 637 | 9 嗯,我把题看错了,汗。
我以为Reward是#H/#T.
【在 s**e 的大作中提到】 : Here is an example that why you should not always stop at H > T. : Suppose (H,T)=(51,50). If stop now, the ratio is 51/101. : But if you keep tossing another 2 times, you will get: : (1) (H,T)= (53,50), ratio = 53/103 (with probability 1/4) : (2) (H,T)= (52,51), ratio = 52/103 (with probability 1/2) : (3) (H,T)= (51,52), ratio = 51/103 (with probability 1/4) : However, in the case (3) above, you can keep tossing, : and can achieve at least 1/2. : Here we have : 53/103*1/4 + 52/103*1/2 + 1/2 * 1/4 > 51/101.
| b******n 发帖数: 637 | 10 等下你最后一个1/2*1/4怎么来的啊。不应该是51/103么。然后就是左小于右了。我的
policy仍然成
立吧。
【在 s**e 的大作中提到】 : Here is an example that why you should not always stop at H > T. : Suppose (H,T)=(51,50). If stop now, the ratio is 51/101. : But if you keep tossing another 2 times, you will get: : (1) (H,T)= (53,50), ratio = 53/103 (with probability 1/4) : (2) (H,T)= (52,51), ratio = 52/103 (with probability 1/2) : (3) (H,T)= (51,52), ratio = 51/103 (with probability 1/4) : However, in the case (3) above, you can keep tossing, : and can achieve at least 1/2. : Here we have : 53/103*1/4 + 52/103*1/2 + 1/2 * 1/4 > 51/101.
| | | s**e 发帖数: 1834 | 11 Even the Reward是#H/#T, your still do not always stop when H > T. :-)
The same example works.
(H,T) = (51,50) with ratio=51/50, and toss another 2 times.
【在 b******n 的大作中提到】 : 嗯,我把题看错了,汗。 : 我以为Reward是#H/#T.
| s**e 发帖数: 1834 | 12 by law of large numbers, you can achieve at least 1/2.
So in the formular 51/103 was replaced with 1/2.
【在 b******n 的大作中提到】 : 等下你最后一个1/2*1/4怎么来的啊。不应该是51/103么。然后就是左小于右了。我的 : policy仍然成 : 立吧。
| w*********m 发帖数: 196 | 13 看他的意思应该是expected payoff最小是1/2,因为如果出现第三种情况就一直扔,
直到大于或等于1/2。 | b******n 发帖数: 637 | 14 赞链接,你怎么搜出来滴?
【在 s**e 的大作中提到】 : This is still an open problem, only some asymptotic solution : is available. : http://arxiv.org/PS_cache/arxiv/pdf/0907/0907.0032v2.pdf
| G********d 发帖数: 10250 | 15 直观上看至少可以做到0.75吧
一般的概率第一次就赢了 停下获得1
另一半的概率第一次败了 因为常返性 必须回到0 停下获得0.5
所以是0.75 | s**e 发帖数: 1834 | 16 This problem was discussed on this board before, and somebody
gave a link. I don't remember if this one is the same link though.
【在 b******n 的大作中提到】 : 赞链接,你怎么搜出来滴?
|
|