青蛙跟风炒股秘籍--理论依据的错误

xiaobailong

俩对一错: 0.6*0.6X0.4*(3!/2!(3-2)!) =.432
一对俩错:  0.6*0.4*0.4 *(3!/(3-1)!) = 0.288
Control_Risk 发表于 2009-3-1 05:37

算了算，这样三分之一仓位的赢率居然还是60%， 没有任何提高呢。

xiaobailong

算了算，这样三分之一仓位的赢率居然还是60%， 没有任何提高呢。
xiaobailong 发表于 2009-3-1 06:25

但是可以在72笔交易中把前面减少的4.8个交易补回来。

啊？又增加了72笔交易费，盈利了4.8个交易。这不是和跟一位老大一样吗？

把三分之一仓位提高到全仓，可以盈利14.4个交易，这样就有提高了。

结论：三位老大中两位以上意见一致的时候，就跟。这样100次交易中，可以净盈利29.6次。比只跟一位老大的净盈利次数提高了9.6次。

前提是：三位老大一直都有意见，不会出现某位老大弃权，另两位老大意见相反的情况。

egghead1

15# Control_Risk

Sorry, don't agree with your calculation. I think lite is correct. If 3 lao da give out the same signal, let's say buy, and the market subsequently moves in the opposite direction, let's say down, it means all 3 of them are wrong. Therefore, they have to all be wrong for followers to lose money. The probability of that happening is 0.4X0.4X0.4. So the win/lose ratio should be (1-0.4^3)/(0.4^3).

Also your logic about how many trades to make is arbitrary. Someone can easily scale the numbers and get different results.

lite1067

能看见小白龙make mistake，let me feel very happy

:-)

lite1067

15# Control_Risk

Sorry, don't agree with your calculation. I think lite is correct. If 3 lao da give out the same signal, let's say buy, and the market subsequently moves in the opposite directio ...
egghead1 发表于 2009-3-1 09:25

egghead1 is right...

it seems either my point was not correctly understood or many people here need to review the probability textbook :-)

torontonian

egghead1

25# lite1067

Sorry again, actually I am changing my mind. :-) I think control_risk is correct. In this case, it's a conditional probability based on the assumption that you only make a trade if all 3 lao dao have the same prediction. So the probability space is not the whole universe of events. It's only when either they are all right or all wrong. You have to throw out the combinations that one or two of them is right. So the probability space becomes (0.6^3+0.4^3). Therefore, the win/loss ratio becomes (0.6/0.4)^3.

For example, there are 1000 events. Each of lao da will get 600 of them right. But there are only 6^3=216 events that all 3 of them will get it right. And there are 64 events that all 3 of them get it wrong. So if you trade only on those (216+64)=280 events, then you will win 216/280=77% of the time.

lite1067

25# lite1067

Sorry again, actually I am changing my mind. :-) I think control_risk is correct. In this case, it's a conditional probability based on the assumption that you only make a trade if a ...
egghead1 发表于 2009-3-1 11:43

we are talking about two types of probabilities here. i will post a clean argument when i get time...

xiaobailong

能看见小白龙make mistake，let me feel very happy

:-)
lite1067 发表于 2009-3-1 11:13

呵呵，可见无脑跟风是多么大的危险啊。

xiaobailong

we are talking about two types of probabilities here. i will post a clean argument when i get time...
lite1067 发表于 2009-3-1 11:45

我知道你说的是什么，你的理论是：三位老大都错的概率是93.6%。  这一点我们也都同意。

可是这并不等于正确率就是93.6%。 你再好好看看ControlRisk的算法，我觉得他是有道理的。

egghead1

28# lite1067

Remember the condition is that you make a trade only if all 3 of them give the same prediction.

jasmin256

30# xiaobailong
agree: 青蛙真不好当啊

humbletrader

This thread is really great! Thanks, especially Egghead, Lite, Risk_control!

benshou

control_risk是对的。另外，也不可能找到3个独立的老大。作为个体他们是独立的，但三个老大各自的买入卖出决定是不独立的，因为他们观察的是同一个市场。

lite1067

control_risk是对的。另外，也不可能找到3个独立的老大。作为个体他们是独立的，但三个老大各自的买入卖出决定是不独立的，因为他们观察的是同一个市场。
benshou 发表于 2009-3-1 12:19

don't rush your conclusion before reading my complete series...

NewMember

No matter who's right, it's obviously that when 3 Lao Da have come to the same conclusion independently, the chance of them being right is very high.  

Isn't is more important pick out 3 Lao Da than debating on the exact odds?  Let's come out with a list.

balto

benshou

No matter who's right, it's obviously that when 3 Lao Da have come to the same conclusion independently, the chance of them being right is very high.  

Isn't is more important pick out 3 Lao Da tha ...
NewMember 发表于 2009-3-1 15:00

我强烈推荐5分老大，1毛老大和2毛5老大。三个绝对独立。虽然每个的正确率只有50%，但按照楼主的理论，三个凑一起可以有1-（1-0.5）×（1-0.5）×（1-0.5）= 87.5%的赢率。

Control_Risk

我强烈推荐5分老大，1毛老大和2毛5老大。三个绝对独立。虽然每个的正确率只有50%，但按照楼主的理论，三个凑一起可以有1-（1-0.5）×（1-0.5）×（1-0.5）= 87.5%的赢率。

benshou 发表于 2009-3-1 16:28

三个臭皮匠顶个诸葛亮?  哈哈.

ULTRAETF

俩位都不对。  OK, nothing personal, just discuss probability.
三老大都对： 0.6X0.6X0.6 = 0.216
俩对一错: 0.6*0.6X0.4*(3!/2!(3-2)!) =.432
一对俩错:  0.6*0.4*0.4 *(3!/(3-1)!) = 0.288
都 ...
Control_Risk 发表于 2009-3-1 05:37

这个帖子越来越有意思了。