p*******0
发帖数: 420 | 1 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
次，包括pre和
after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
，mysql受不
了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
计新系统要比以前
快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
100只以内值得做
的做5分钟的跟踪。其他都是按天分析。
系统做好以后是可以实时通知的(email, sms等)，甚至可以直接交易（做成功了准备上
ib）。如果版上有
哪些牛牛们对这个东西有兴趣可以和我联系，给我提供点意见和算法。做好以后可以为
大家服务，提供
email list。
我的gmail是p*******[email protected]，或者msn:p******[email protected]
PS:旧的系统大家可以看下 phoenix.linkpc.net，现在这个系统里面没什么数据，因为
要重写才放上来
对照的。系统界面做的比较简陋，因为大部分都是后台运行，前台只是一个调控。大家
可以去stock
portfolios 下real panel看下，这里面都是可以追踪的原始数据。 |
u********e
发帖数: 4950 | |
d****a
发帖数: 2901 | 3 早就用云计算了，哈哈
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
s*******e
发帖数: 432 | 4 very nice. Do you have fundamental data available. |
w****u
发帖数: 172 | 5 首先 为什么要自己做？ finance.google.com已经有很好的资料了
第二，大数据量你承受不了。给你个建议，考虑couchDB 之类的map reduce存储系统存
储近期
的，数据大了用schema的join都是做梦，除非有DBA帮你弄。 老数据做好index然后文
件系统存放
吧，没有必要放数据库了。
第三，怎么AUTO trade？？？？？？。。。。。
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
s*******e
发帖数: 432 | 6 I suggest use volume / float ratio instead of volume only |
s*******e
发帖数: 432 | 7 very nice. Do you have fundamental data available. |
x*********x
发帖数: 1402 | |
x*********x
发帖数: 1402 | |
w******s
发帖数: 16209 | 10 en.我觉得这个也不是autotrade。
描一
大了
守估
找到
【在 w****u 的大作中提到】
: 首先 为什么要自己做？ finance.google.com已经有很好的资料了
: 第二，大数据量你承受不了。给你个建议，考虑couchDB 之类的map reduce存储系统存
: 储近期
: 的，数据大了用schema的join都是做梦，除非有DBA帮你弄。 老数据做好index然后文
: 件系统存放
: 吧，没有必要放数据库了。
: 第三，怎么AUTO trade？？？？？？。。。。。
|
|
|
f****e
发帖数: 227 | 11 自己做的好处是能自动监测pattern and trend吧，至少可以先过滤一下，找一些有潜
质的股票再仔细分析吧。finance.google 不能有这些功能吧？
【在 w****u 的大作中提到】
: 首先 为什么要自己做？ finance.google.com已经有很好的资料了
: 第二，大数据量你承受不了。给你个建议，考虑couchDB 之类的map reduce存储系统存
: 储近期
: 的，数据大了用schema的join都是做梦，除非有DBA帮你弄。 老数据做好index然后文
: 件系统存放
: 吧，没有必要放数据库了。
: 第三，怎么AUTO trade？？？？？？。。。。。
|
p*******0
发帖数: 420 | 12 1. fundamental数据有，但是只有profit,book value一些基本的，没有预测。
2. finance yahoo不会告诉你哪知股票要涨，哪知要跌
3. 现在同时跟踪2000只股票只要大概1个月数据库就很慢了，新系统做好后我只打算留
6个月的5min数据，剩下的都放到daily data里面。
4. 云计算什么的都是个形势，主要是算法和数据，以后升级很容易的。现在我的服务
器是双xeon 3.2gh 4gb ram, 10000rpm。对于目前开发测试我觉得差不多了。
5. 现在我的broker不支持auto trade，所以这个只能算个push alert,不过改auto
trade很容易。
6. 我看看couchDB，现在考虑为每个stock每月建个表
7. 由于是改写，进度很快，估计到下周新系统就可以接收数据了，之后就要写
indicator和signal了。我个人对ta还不是很了解，还指望nns不吝赐教。 |
w******s
发帖数: 16209 | 13 呵呵。 赞热情。
【在 p*******0 的大作中提到】
: 1. fundamental数据有，但是只有profit,book value一些基本的，没有预测。
: 2. finance yahoo不会告诉你哪知股票要涨，哪知要跌
: 3. 现在同时跟踪2000只股票只要大概1个月数据库就很慢了，新系统做好后我只打算留
: 6个月的5min数据，剩下的都放到daily data里面。
: 4. 云计算什么的都是个形势，主要是算法和数据，以后升级很容易的。现在我的服务
: 器是双xeon 3.2gh 4gb ram, 10000rpm。对于目前开发测试我觉得差不多了。
: 5. 现在我的broker不支持auto trade，所以这个只能算个push alert,不过改auto
: trade很容易。
: 6. 我看看couchDB，现在考虑为每个stock每月建个表
: 7. 由于是改写，进度很快，估计到下周新系统就可以接收数据了，之后就要写
|
f****e
发帖数: 227 | 14 看了一下你的网站。你做的不能算auto trading吧。就一 real-time screening, 没什
么意义吧？off-line screening from yahoo will do.
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
y******u
发帖数: 246 | 15 不管是电脑选股还是autotrade，编程不是主要问题，实在不行雇人编。主要是你的
trading strategy行不行。我猜你的程序是找短期突破，多说一点大家才能讨论。 |
w****k
发帖数: 6244 | 16 我也感兴趣做一个类似的东西
几千个股票的fundamental数据哪里来呢？
p/e, eps, eps growth, earning等等
不知道可以从哪里收集
【在 p*******0 的大作中提到】
: 1. fundamental数据有，但是只有profit,book value一些基本的，没有预测。
: 2. finance yahoo不会告诉你哪知股票要涨，哪知要跌
: 3. 现在同时跟踪2000只股票只要大概1个月数据库就很慢了，新系统做好后我只打算留
: 6个月的5min数据，剩下的都放到daily data里面。
: 4. 云计算什么的都是个形势，主要是算法和数据，以后升级很容易的。现在我的服务
: 器是双xeon 3.2gh 4gb ram, 10000rpm。对于目前开发测试我觉得差不多了。
: 5. 现在我的broker不支持auto trade，所以这个只能算个push alert,不过改auto
: trade很容易。
: 6. 我看看couchDB，现在考虑为每个stock每月建个表
: 7. 由于是改写，进度很快，估计到下周新系统就可以接收数据了，之后就要写
|
g****u
发帖数: 695 | 17 有这本事干点什么有意义的事情不好。
听大帅的话，找点别的项目做吧。股票根本就不是你一个人搞个 auto trading
system 做得了的。这跟研究永动机一样，不管你下多大的功夫，用多高深的学
问，根本就不可能有结果。
如果编程是你的强项，你最好的出路是找个 hedge fund 在里面做个 programmer。
看看人家是怎么想的怎么做的，学上一年半载，你就知道你现在的想法有多么的
幼稚。
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
k****n
发帖数: 1334 | 18 我从yahoo搜集price历史数据
从morningstar和ycharts挖掘基本数据
然后根据一些条件扫描出来pick，然后人工分析。
worked pretty well
【在 w****k 的大作中提到】
: 我也感兴趣做一个类似的东西
: 几千个股票的fundamental数据哪里来呢？
: p/e, eps, eps growth, earning等等
: 不知道可以从哪里收集
|
k****n
发帖数: 1334 | 19 exactly，做短线的trading system根本就是徒劳
不是泼冷水，是事实
【在 g****u 的大作中提到】
: 有这本事干点什么有意义的事情不好。
: 听大帅的话，找点别的项目做吧。股票根本就不是你一个人搞个 auto trading
: system 做得了的。这跟研究永动机一样，不管你下多大的功夫，用多高深的学
: 问，根本就不可能有结果。
: 如果编程是你的强项，你最好的出路是找个 hedge fund 在里面做个 programmer。
: 看看人家是怎么想的怎么做的，学上一年半载，你就知道你现在的想法有多么的
: 幼稚。
|
f****e
发帖数: 227 | 20 TA的话，是不是daily既可以了？5min/15min for DT? 村长应该很有经验的。村长给说
说看？
【在 p*******0 的大作中提到】
: 1. fundamental数据有，但是只有profit,book value一些基本的，没有预测。
: 2. finance yahoo不会告诉你哪知股票要涨，哪知要跌
: 3. 现在同时跟踪2000只股票只要大概1个月数据库就很慢了，新系统做好后我只打算留
: 6个月的5min数据，剩下的都放到daily data里面。
: 4. 云计算什么的都是个形势，主要是算法和数据，以后升级很容易的。现在我的服务
: 器是双xeon 3.2gh 4gb ram, 10000rpm。对于目前开发测试我觉得差不多了。
: 5. 现在我的broker不支持auto trade，所以这个只能算个push alert,不过改auto
: trade很容易。
: 6. 我看看couchDB，现在考虑为每个stock每月建个表
: 7. 由于是改写，进度很快，估计到下周新系统就可以接收数据了，之后就要写
|
|
|
g*******w
发帖数: 430 | 21 我所知道的做auto trading的除非是大资金的， 小资金的auto trading还不如纯手动 |
b*******e
发帖数: 6389 | 22 支持下楼主。
类似的系统俺测试过C++/MySQL和C++/纯文本文件，后者的速度是前者的2到3倍左右。
所以俺自己做了个简单的Database Management System，速度很快。个人经验谨供参考。
一
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
f****e
发帖数: 227 | 23 大帅啊，真的是象永动机一样的 hopeless 吗？我几年前有一个 matlab 的 prototype
, 还想着最近有点时间接着把它做下去呢。
有人能跳出来给点希望吗？
【在 g****u 的大作中提到】
: 有这本事干点什么有意义的事情不好。
: 听大帅的话，找点别的项目做吧。股票根本就不是你一个人搞个 auto trading
: system 做得了的。这跟研究永动机一样，不管你下多大的功夫，用多高深的学
: 问，根本就不可能有结果。
: 如果编程是你的强项，你最好的出路是找个 hedge fund 在里面做个 programmer。
: 看看人家是怎么想的怎么做的，学上一年半载，你就知道你现在的想法有多么的
: 幼稚。
|
w****k
发帖数: 6244 | 24 相信自己
不试怎么知道
有很多可以持续盈利的model，其实并不复杂
prototype
【在 f****e 的大作中提到】
: 大帅啊，真的是象永动机一样的 hopeless 吗？我几年前有一个 matlab 的 prototype
: , 还想着最近有点时间接着把它做下去呢。
: 有人能跳出来给点希望吗？
|
w******s
发帖数: 16209 | 25 有厉害的主就是用matlab 也能做的。不过最后都是欠到c++上去下单。
中间是runtime 调用时间长短对一般散户的波段操作应该没有太大影响（非高频）
prototype
【在 f****e 的大作中提到】
: 大帅啊，真的是象永动机一样的 hopeless 吗？我几年前有一个 matlab 的 prototype
: , 还想着最近有点时间接着把它做下去呢。
: 有人能跳出来给点希望吗？
|
s*******e
发帖数: 432 | 26 Thanks very much for your advice. That is what i am looking for.
Best
【在 k****n 的大作中提到】
: 我从yahoo搜集price历史数据
: 从morningstar和ycharts挖掘基本数据
: 然后根据一些条件扫描出来pick，然后人工分析。
: worked pretty well
|
f****e
发帖数: 227 | 27 你说的对。我个人感觉即便对散户来说也是有一些希望的，其实好多常用的TA
indicators are not difficult to calculate and some TA patterns can be detect
automatically.
【在 w****k 的大作中提到】
: 相信自己
: 不试怎么知道
: 有很多可以持续盈利的model，其实并不复杂
:
: prototype
|
f****e
发帖数: 227 | 28 谢谢村长的鼓励。等将来做成了一定回来汇报一下。
【在 w******s 的大作中提到】
: 有厉害的主就是用matlab 也能做的。不过最后都是欠到c++上去下单。
: 中间是runtime 调用时间长短对一般散户的波段操作应该没有太大影响（非高频）
:
: prototype
|
p*******0
发帖数: 420 | 29 谢谢大家关心
1. 我知道这个东西一个人做很困难，所以才来版上请教各位大牛。我觉得现在就已经
有很多人给了我很好的新思路。如果有志同道合的也可以一起探讨合作。
2. computer can beat human brain in the market是我的信仰，
3. 像本版大夫这样的牛人更加强了我对有一些人可以beat the market的信心。
4. 这个系统倒不是一定是要短期交易。我认为交易时间段是由利润和交易成本决定的
（commission，spread等，有margin还有interests）。
5. 现在可以说是一个real time screen,但自己的程序可以写自己的screen criteria.
上一个程序流产的时候我正在写一个identify trend & pattern, resistance &
support的程序，虽然也是screen,但我想很少有软件有这个功能吧
6. 而且这个软件还包含portfolio管理，就是说它不光告诉你买卖什么，还会告诉你多
少，根据不同的strategy控制风险和利润
7. 这个软件的最终产物将会是trading signal,即买卖和数量。对于软件来说会screen
,会analyze,但对于user来说只有买卖和数量，其他神马都是浮云，要不也就不叫auto
trade了。
8. 我要求5min信息主要是因为我认为每笔交易对于股市都是由影响的，得到的信息越
多预测的越准。比如说一个股票一天一直升到最高点和到盘尾一下窜到最高点从daily
chart看是一样的，但对于以后的预示可能完全不一样。
9. 如果我能证明我的模型能挣钱，你认为我会发愁没钱运作么。
10. yahoo里面也有fundimental信息，我那个表上的信息都是从yahoo搜集的，从其它
网站的信息不在那个地方。
11. 我过两天试试看纯文本。我专业是会计，对计算机都是初中开始自己慢慢学的，很
多地方还不是很懂。不过我现在的系统一些简单的常用数据都是用文本(track的股票，
信息什么的)。如果信息比较复杂如何很快查询呢(比如有bid, ask, volume, trade，
我想查所有trade between $x and $y)。还望牛刀兄多多指教 |
m********0
发帖数: 2717 | 30 看你这么热情，就说两句。
这板上90%以上是挖苦和讽刺，还有就是来寻求虚荣心的。从你总结的那么多点看。
你要走的路还好远好远，我从一无所知到build你这样的system大概两三个礼拜的样子。刚开始也
很热情，如果只是把make money作为动力，恐怕热情不会持续这么久，如果当作一种消遣，可能对
身心更有益。
2. beat the market做信仰可不行，beat market概念上，只要买write covered SPY
call就by definition beat the market了。。。。
4. 频率是由你想找什么样的inefficiency决定的，高频对于licensed member的机构比
较实
际，他们不需要交commission，还可以拿到大量liquidity rebate，如果你想花M（不
记得具体
多少了）买一个交易所的membership也可以。如果叫commission的话，越高频越难做，没
commission的情况下，SR达到上千上万都有可能，有commission，。。。。如果想入门
，可以参
考一下Irene写的那本Introduction to HFT
5. 看你的描述似乎是想做swing style的，这样做个beat market的strategy挺多，也
不难
做，比较beginner的做法大部分是static的思路，最多每天开盘前必要情况下重新
evaluate一下
结果，adjust parameters一下。我在刚学习的第二个月做过static的backtest。
6. PM的model很多，而且一直在发展，我前面推荐的paper里面有reference很多流行的
model。那些用online PAC learning+boost可以显著地beat传统的PM model。
7. 比你想的要复杂，execution值得优化的方面很多很多，可以参考一下这本书，
Kissell, Robert (2003), Optimal Trading Strategies.
SAC capital的人作序的，比较经典。
8. 做swing，5min对你帮助有限，做day trading，5min只能分析走势，做signal不够。
9. 如果我能证明我的模型能挣钱，你认为我会发愁没钱运作么。
你会，因为卖模型的人太多了。看看C2上面的strategy，或者你看看wealth-lab上的，
反正现成
的strategy没有十万种也有一万种了。你是无法prove的，即便你能拿得出非常好的
track
record，非常漂亮的各种risk/reward指标（这个也一直在发展），如果自己没有一定
规模的资金
实际地操作半年一年，没有人会买你的帐。例外我也遇到过，不细说了。
11. 很多人都说了，strategy远重要于技术细节，技术细节属于慢慢tuning的工作，有
时间了可
以RAID存储，快很多，SSD也快可以RAID了。或者可以尝试一下GPU computing。挺多大
公司已
经用这个做simulation相关的计算了。
criteria.
【在 p*******0 的大作中提到】
: 谢谢大家关心
: 1. 我知道这个东西一个人做很困难，所以才来版上请教各位大牛。我觉得现在就已经
: 有很多人给了我很好的新思路。如果有志同道合的也可以一起探讨合作。
: 2. computer can beat human brain in the market是我的信仰，
: 3. 像本版大夫这样的牛人更加强了我对有一些人可以beat the market的信心。
: 4. 这个系统倒不是一定是要短期交易。我认为交易时间段是由利润和交易成本决定的
: （commission，spread等，有margin还有interests）。
: 5. 现在可以说是一个real time screen,但自己的程序可以写自己的screen criteria.
: 上一个程序流产的时候我正在写一个identify trend & pattern, resistance &
: support的程序，虽然也是screen,但我想很少有软件有这个功能吧
|
|
|
z***o
发帖数: 1254 | 31 展开说说？ 我现在也为没有合适的初步screening股票的工具发愁。 其实步骤不算复
杂，可惜编程功力低下，只会用傻瓜型的软件， matlab顶头了。有合适的更愿意花钱买。
【在 k****n 的大作中提到】
: 我从yahoo搜集price历史数据
: 从morningstar和ycharts挖掘基本数据
: 然后根据一些条件扫描出来pick，然后人工分析。
: worked pretty well
|
F******a
发帖数: 92 | 32 赞一下楼主的热情,我们就需要对auto trading系统的信仰和激情.
以前我也曾尝试做自己的auto trading系统,但由于工作忙不过来,最后放弃了.
如果有什么进展,不妨时时给我们update.
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
f****e
发帖数: 227 | 33 WOW,只能说大牛就是大牛。替LZ谢谢你肯花时间码这么多字，我们旁观者也受益匪浅。
我非常赞同你的一些观点。我是搞工程的，没有多少背景，但我的gut feeling跟你的
很一致。你的解释让我清楚了很多。我读了你推荐的那本论文，很有意思。里面的
references也很有用。
子。刚开始也
消遣，可能对
【在 m********0 的大作中提到】
: 看你这么热情，就说两句。
: 这板上90%以上是挖苦和讽刺，还有就是来寻求虚荣心的。从你总结的那么多点看。
: 你要走的路还好远好远，我从一无所知到build你这样的system大概两三个礼拜的样子。刚开始也
: 很热情，如果只是把make money作为动力，恐怕热情不会持续这么久，如果当作一种消遣，可能对
: 身心更有益。
: 2. beat the market做信仰可不行，beat market概念上，只要买write covered SPY
: call就by definition beat the market了。。。。
: 4. 频率是由你想找什么样的inefficiency决定的，高频对于licensed member的机构比
: 较实
: 际，他们不需要交commission，还可以拿到大量liquidity rebate，如果你想花M（不
|
g*********n
发帖数: 220 | 34 我也很佩服LZ...
很想自己搞一个系统慢慢调教, 可惜自己计算机啥的都不行
LZ勇气可嘉, 我觉得这还是可行的...有想过用NN吗? |
p*******0
发帖数: 420 | 35 非常感谢mitbbs2020大牛
1. 你推荐的东西我一定会去看的
2. 我从php改c++还有一个原因就是要gpu加速。我一个朋友就是做这个的，他答应我以
后给我改写计算代码。
3. 做这个算是个课余爱好吧。白天上班，晚上和周末就做这个。除了想赚钱以外，也
希望通过做这个程序在trade这方面有点研究，以后可以进个ib,hf，或者申请个phd什
么的。 |
p*******0
发帖数: 420 | 36 nn什么东西，头两天看mitbbs2020 里面提到了svm和knn，准备过两天找点书看呢 |
f****e
发帖数: 227 | 37 Neural networks, another AI algorithm. Where did mitbbs2020 mention SVM
and kNN? About classification/Clustering?
【在 p*******0 的大作中提到】
: nn什么东西，头两天看mitbbs2020 里面提到了svm和knn，准备过两天找点书看呢
|
w*******o
发帖数: 6125 | 38 说起NN,据闻当年psp大牛花了十几万买了个基于Neural Network的核武器，
但是这个核武器没有与时俱进，在去年4，5月的在Fed大举
干预市场的时候出了"Short the market to death"的信号，
那是我最后一次看到psp的帖子了，不知道后来核武器改进了没有。
【在 p*******0 的大作中提到】
: nn什么东西，头两天看mitbbs2020 里面提到了svm和knn，准备过两天找点书看呢
|
g*********n
发帖数: 220 | 39 十几万??太牛b了吧
我只希望有个可以慢慢调较的NN
只是这还是需要有TA的基本知识...
顺道问一下, 有啥好用的NN软件吗? |
I********n
发帖数: 1468 | 40 做一些分析工具, 并系统化是很好的, 但要 autotrading, 坦诚的说， 你在浪费时间
。
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
|
|
g****u
发帖数: 695 | 41 I never trust a single word from psp.
But even assume the story is true, it's only another example that automatic
trading systems, no matter how sophisticated they look like, simply don't
work.
【在 w*******o 的大作中提到】
: 说起NN,据闻当年psp大牛花了十几万买了个基于Neural Network的核武器，
: 但是这个核武器没有与时俱进，在去年4，5月的在Fed大举
: 干预市场的时候出了"Short the market to death"的信号，
: 那是我最后一次看到psp的帖子了，不知道后来核武器改进了没有。
|
u********e
发帖数: 4950 | 42 I believe the nature of the market is "simple".
The more sophisticated the system is, the less chance it can win.
Try it out and good luck to you. |
f****e
发帖数: 227 | 43 My experience of data mining shows that indeed sophisticated systems
sometimes tend to overfitting the data. And for black box models you don't
know what is going on in there. Just like blindly following GPS may lead you
to a wrong place. So I think simpler but more intuitive ones tend to be
more robust and might be helpful in decision making - not necessarily auto
trading, but as a facilitating tool.
automatic
【在 g****u 的大作中提到】
: I never trust a single word from psp.
: But even assume the story is true, it's only another example that automatic
: trading systems, no matter how sophisticated they look like, simply don't
: work.
|
f****e
发帖数: 227 | 44 I used c++ for GUI and Matlab's c/c++ math library for computation for
another project. Suggest LZ consider.
【在 w******s 的大作中提到】
: 有厉害的主就是用matlab 也能做的。不过最后都是欠到c++上去下单。
: 中间是runtime 调用时间长短对一般散户的波段操作应该没有太大影响（非高频）
:
: prototype
|
g****u
发帖数: 695 | 45 The market is "simple" in the sense that it's always about demand and
supply and history repeats itself again and again in a very broad base.
Beyond that, we can only say market is extremely complicated. The
information presented and digested by the market is way way beyond
human technology to ever process.
Actually, human being can never master financial market because the
purpose of observers is always to interfere. And the resulted interference
by its nature is to invalid the original observation. It's like a cat
chasing its own tail. In that sense, astronomy has a much brighter
future because human interference is not possible in any foreseeable
future.
【在 u********e 的大作中提到】
: I believe the nature of the market is "simple".
: The more sophisticated the system is, the less chance it can win.
: Try it out and good luck to you.
|
C*G
发帖数: 7495 | 46 大帅，我们文科凤凰WSN看不懂英文
哭死哭死
【在 g****u 的大作中提到】
: The market is "simple" in the sense that it's always about demand and
: supply and history repeats itself again and again in a very broad base.
: Beyond that, we can only say market is extremely complicated. The
: information presented and digested by the market is way way beyond
: human technology to ever process.
: Actually, human being can never master financial market because the
: purpose of observers is always to interfere. And the resulted interference
: by its nature is to invalid the original observation. It's like a cat
: chasing its own tail. In that sense, astronomy has a much brighter
: future because human interference is not possible in any foreseeable
|
A*********n
发帖数: 637 | 47 这是个很好的建议,单纯用C++做数学的算法太麻烦.
【在 f****e 的大作中提到】
: I used c++ for GUI and Matlab's c/c++ math library for computation for
: another project. Suggest LZ consider.
|
s*******e
发帖数: 432 | 48 there's lot of evidence that the market in US is at least weak efficient
which means by looking at historical price, no one can gain anything from it. I believe with the wide availability and current speed of tools for mining and analyzing data from public filing. The US market has already been or will be semi-efficient soon, which means on one can gain after looking at public data filed by companies such as fundamental data derived from sec filing. |
f****e
发帖数: 227 | 49 I agree with you that "human being can never master financial market because
the purpose of observers is always to interfere" and we can never
completely predict the system behavior. However, the system has its own
momentum and inertia, besides, the mass majority of the participants are
followers, not leaders or active participants like frogs on this board. So I
believe it is possible to predict the direct of a trend in a SHORT FUTURE
HORIZON.If the majority of the money is controlled by MM equipped with
sophisticated trading strategies, then it would be much more difficult for
individuals to make money in this market.
【在 g****u 的大作中提到】
: The market is "simple" in the sense that it's always about demand and
: supply and history repeats itself again and again in a very broad base.
: Beyond that, we can only say market is extremely complicated. The
: information presented and digested by the market is way way beyond
: human technology to ever process.
: Actually, human being can never master financial market because the
: purpose of observers is always to interfere. And the resulted interference
: by its nature is to invalid the original observation. It's like a cat
: chasing its own tail. In that sense, astronomy has a much brighter
: future because human interference is not possible in any foreseeable
|
s*********a
发帖数: 129 | 50 还没有赚到钱，先把自己脑细胞杀死多半了。
建议用broker成熟的软件，多研究。不要做无用功了
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
|
|
g****u
发帖数: 695 | 51 I agree with you that market has its momentum and inertia. That actually has
been proved by many academics, include Eugene Fama himself, the developer of
the EMH.
But note that chasing momentum is itself re-enforcing the momentum, or in
other words it's a positive feedback. We know positive feedback only makes
the system increasingly unstable. So when you loss, you loss big. It
eventually becomes a game about getting in and out ahead of the crowd,
while participators are actually the crowd. Like all games, there are always
winners and losers. But can there be a constant winner? I doubt.
The only way to beat the market to bet against the crowd. However, this is
not a secrete. When more and more people come with ideas to bet against
the crowd, they themselves become the crowd. Then it's just another cat
chasing its own tail.
because
I
【在 f****e 的大作中提到】
: I agree with you that "human being can never master financial market because
: the purpose of observers is always to interfere" and we can never
: completely predict the system behavior. However, the system has its own
: momentum and inertia, besides, the mass majority of the participants are
: followers, not leaders or active participants like frogs on this board. So I
: believe it is possible to predict the direct of a trend in a SHORT FUTURE
: HORIZON.If the majority of the money is controlled by MM equipped with
: sophisticated trading strategies, then it would be much more difficult for
: individuals to make money in this market.
|
C********N
发帖数: 1263 | |
g*****u
发帖数: 14294 | 53 butt admiral挣得不错，当然也就没啥进取心。需要反复给自己心理暗示，自己是对的。
【在 w******s 的大作中提到】
: 有厉害的主就是用matlab 也能做的。不过最后都是欠到c++上去下单。
: 中间是runtime 调用时间长短对一般散户的波段操作应该没有太大影响（非高频）
:
: prototype
|
l****n
发帖数: 6773 | 54 RE
has
of
always
【在 g****u 的大作中提到】
: I agree with you that market has its momentum and inertia. That actually has
: been proved by many academics, include Eugene Fama himself, the developer of
: the EMH.
: But note that chasing momentum is itself re-enforcing the momentum, or in
: other words it's a positive feedback. We know positive feedback only makes
: the system increasingly unstable. So when you loss, you loss big. It
: eventually becomes a game about getting in and out ahead of the crowd,
: while participators are actually the crowd. Like all games, there are always
: winners and losers. But can there be a constant winner? I doubt.
: The only way to beat the market to bet against the crowd. However, this is
|
s*******e
发帖数: 432 | 55 Studies has shown that hedges are not better than retailers at all. one explanation for why people are still getting to hedges is that people believe that they can choose or already have chosen the best hedges.
【在 w******s 的大作中提到】
: 有厉害的主就是用matlab 也能做的。不过最后都是欠到c++上去下单。
: 中间是runtime 调用时间长短对一般散户的波段操作应该没有太大影响（非高频）
:
: prototype
|
g****u
发帖数: 695 | 56 I'll tell you what: they should.
You think the hedge fund industry live on sophisticated trading strategies
or algorithms? That's the biggest lie in 2000s. The whole hedge fund
industry lives on one and only one principle: Heads they win, tails you
loss.
Private Equity is a much more meaningful animal.
【在 w******s 的大作中提到】
: 有厉害的主就是用matlab 也能做的。不过最后都是欠到c++上去下单。
: 中间是runtime 调用时间长短对一般散户的波段操作应该没有太大影响（非高频）
:
: prototype
|
w*******o
发帖数: 6125 | 57 Bingo, 其实Hedge Fund总体上也没有大家想象那么牛掰,
当然，就像Start-Up一样，人们往往津津乐道最后成功的几个,
但都是"一将功成万骨枯"
explanation for why people are still getting to hedges is that people
believe that they can choose or already have chosen the best hedges.
【在 s*******e 的大作中提到】
: Studies has shown that hedges are not better than retailers at all. one explanation for why people are still getting to hedges is that people believe that they can choose or already have chosen the best hedges.
|
g****u
发帖数: 695 | 58 进取心 is only meaningful when you put it on the right direction.
I am not trying to convince most of you anyway. You will understand at
some point, if your heart is strong enough to stay in the market for
a long enough time.
的。
【在 g*****u 的大作中提到】
: butt admiral挣得不错，当然也就没啥进取心。需要反复给自己心理暗示，自己是对的。
|
g****u
发帖数: 695 | 59 Even these HF indices won't tell you the whole truth, because they have
deliberately excluded those funds that have been totally wiped out.
【在 w*******o 的大作中提到】
: Bingo, 其实Hedge Fund总体上也没有大家想象那么牛掰,
: 当然，就像Start-Up一样，人们往往津津乐道最后成功的几个,
: 但都是"一将功成万骨枯"
:
: explanation for why people are still getting to hedges is that people
: believe that they can choose or already have chosen the best hedges.
|
g*****u
发帖数: 14294 | 60 “进取心 is only meaningful when you put it on the right direction”
大帅这个讲得有生活智慧。
你们花街上干的营生，不外乎12个大字：坑蒙拐骗，弱肉强食，乘人之危。
fund赚不赚钱不是目的，目的是自己荷包装满。agency problem， 莫有办法。
【在 g****u 的大作中提到】
: 进取心 is only meaningful when you put it on the right direction.
: I am not trying to convince most of you anyway. You will understand at
: some point, if your heart is strong enough to stay in the market for
: a long enough time.
:
: 的。
|
|
|
x****7
发帖数: 895 | 61 人们往往津津乐道最后成功的几个,但都是"一将功成万骨枯" -----cannot agree more
.The overall success rate is low.
【在 w*******o 的大作中提到】
: Bingo, 其实Hedge Fund总体上也没有大家想象那么牛掰,
: 当然，就像Start-Up一样，人们往往津津乐道最后成功的几个,
: 但都是"一将功成万骨枯"
:
: explanation for why people are still getting to hedges is that people
: believe that they can choose or already have chosen the best hedges.
|
w*******o
发帖数: 6125 | 62 好了，大帅，破比立容易，就像谁都会批评政府，但自己上阵也未必好到哪去.
我来个建设性意见吧，我见过很多人想做系统，但很多都中途而废了.对于没多少资源和时间的小散， 我觉得学会判断大方向是最重要的，这也是蒙古大夫说的，不做跌势，在这个基础上，如果你需要一定的工具来做Stock Screening，特别是TA-based的Screening, e.g, MA(10) cross above MA(50), candle stick pattern, balabala
我推荐两个网站，www.finviz.com and www.stockfetcher.com,花几天好好看看，我相信这里面已经包括了95%你自己想做的功能，如果你想搞得Fancy一点，搞搞AI/Statistics Model,Event-Driven, balabala，那是另外一回事了
【在 g****u 的大作中提到】
: Even these HF indices won't tell you the whole truth, because they have
: deliberately excluded those funds that have been totally wiped out.
|
f****e
发帖数: 227 | 63 How do YOU make money then? You have your own strategy of getting in and out
, right? I figured you can turn this logic into a system that does some
heavy lifting for your brain in processing/filtering huge amount of data/
information out there.
I don't think I can ever convince you because obviously you have rich
experience while I am an amateur with no financial background and not even
knowing some basic jargons on this board. But I do appreciate that you share
your opinion and I mean it. I read some of your other posts and I think I
learned something.
has
of
always
【在 g****u 的大作中提到】
: I agree with you that market has its momentum and inertia. That actually has
: been proved by many academics, include Eugene Fama himself, the developer of
: the EMH.
: But note that chasing momentum is itself re-enforcing the momentum, or in
: other words it's a positive feedback. We know positive feedback only makes
: the system increasingly unstable. So when you loss, you loss big. It
: eventually becomes a game about getting in and out ahead of the crowd,
: while participators are actually the crowd. Like all games, there are always
: winners and losers. But can there be a constant winner? I doubt.
: The only way to beat the market to bet against the crowd. However, this is
|
f****e
发帖数: 227 | 64 谢谢对青蛙们的指点 - 大方向是最重要的，不能捡了芝麻丢了西瓜。从我个人的经历
，application is more important (makes more money) than (algorithm)
development.
源和时间的小散， 我觉得学会判断大方向是最重要的，这也是蒙古大夫说的，不做跌
势，在这个基础上，如果你需要一定的工具来做Stock Screening，特别是TA-based的
Screening, e.g, MA(10) cross above MA(50), candle stick pattern, balabala
相信这里面已经包括了95%你自己想做的功能，如果你想搞得Fancy一点，搞搞AI/
Statistics Model,Event-Driven, balabala，那是另外一回事了
【在 w*******o 的大作中提到】
: 好了，大帅，破比立容易，就像谁都会批评政府，但自己上阵也未必好到哪去.
: 我来个建设性意见吧，我见过很多人想做系统，但很多都中途而废了.对于没多少资源和时间的小散， 我觉得学会判断大方向是最重要的，这也是蒙古大夫说的，不做跌势，在这个基础上，如果你需要一定的工具来做Stock Screening，特别是TA-based的Screening, e.g, MA(10) cross above MA(50), candle stick pattern, balabala
: 我推荐两个网站，www.finviz.com and www.stockfetcher.com,花几天好好看看，我相信这里面已经包括了95%你自己想做的功能，如果你想搞得Fancy一点，搞搞AI/Statistics Model,Event-Driven, balabala，那是另外一回事了
|
g****u
发帖数: 695 | 65 That will be lots of writing... Going for lunch now. I'll write as much
as I can when I am back.
out
share
【在 f****e 的大作中提到】
: How do YOU make money then? You have your own strategy of getting in and out
: , right? I figured you can turn this logic into a system that does some
: heavy lifting for your brain in processing/filtering huge amount of data/
: information out there.
: I don't think I can ever convince you because obviously you have rich
: experience while I am an amateur with no financial background and not even
: knowing some basic jargons on this board. But I do appreciate that you share
: your opinion and I mean it. I read some of your other posts and I think I
: learned something.
:
|
m********0
发帖数: 2717 | |
N******r
发帖数: 642 | 67 blah blah blah...
the only way to find out is test the shit out of it... |
Q***k
发帖数: 1419 | 68 真的？
【在 w*******o 的大作中提到】
: 说起NN,据闻当年psp大牛花了十几万买了个基于Neural Network的核武器，
: 但是这个核武器没有与时俱进，在去年4，5月的在Fed大举
: 干预市场的时候出了"Short the market to death"的信号，
: 那是我最后一次看到psp的帖子了，不知道后来核武器改进了没有。
|
Q***k
发帖数: 1419 | 69 hehe - I agree that finding a consistently profitable strategy first and
then implement some auto trading algorithms based on that, is a good way.
【在 s*********a 的大作中提到】
: 还没有赚到钱，先把自己脑细胞杀死多半了。
: 建议用broker成熟的软件，多研究。不要做无用功了
|
a****n
发帖数: 1887 | 70 现在HFT市场已经很成熟了，你做出来的软件很有可能用正确的策略也买不到你想要的
东西。。。
【在 w*******o 的大作中提到】
: Bingo, 其实Hedge Fund总体上也没有大家想象那么牛掰,
: 当然，就像Start-Up一样，人们往往津津乐道最后成功的几个,
: 但都是"一将功成万骨枯"
:
: explanation for why people are still getting to hedges is that people
: believe that they can choose or already have chosen the best hedges.
|
|
|
f****e
发帖数: 227 | 71 大帅，how was your lunch? Still waiting to learn from you.
【在 g****u 的大作中提到】
: That will be lots of writing... Going for lunch now. I'll write as much
: as I can when I am back.
:
: out
: share
|
g****u
发帖数: 695 | 72 Okay, here is what I suggest (and also what I am practicing):
1. The first of first is to keep in mind that winning in a trade in
financial market is a matter of probablity but never a sure thing.
Now everybody will say "I don't need you to tell me this, I know
from the beginning." Well, I know you know, but you didn't really
keep it mind, because when it comes to the next question, not many
people really use their brain.
The next question is: if somebody made some money in one trade or a
few trades, is it because they had a right strategy? or because they
are nothing but lucky.
When people think carefully, they all know how important the answer is
to the above question. But the reality is: People are so used to their
daily physical world, where good method produces good result and bad
method produces bad result in sort of certainty or at least a very high
probability. This gives them an instiction to make judgement that
good results must come from good method and bad results come from bad
method, without even think twice. Unfortunately, in financial markets,
good results can very much likely come from bad method, not forever,
but within some period. Of course, bad method will eventually yield
bad result in total, but the effect can take some time to show up.
It took sometime for people to get rid of this instinction when making
judgement in the market, yet for some people they never really understand.
If you agree with what I said above, you can keep reading the rest part.
Otherwise, forget all negative things I have said, no bad feeling and
keep on your happy trading.
So what the hell this bullshit has anything to do with trading? Well,
it is very related because it will yield a series of negative conclusions
about what you probably have trusted.
The first one I am going to challenge is Technical Analysis, which I am
sure everyone know and 95% of people on this board are putting their
full faith on. My conclusion will be -- TA is an illusion which only
works when you look back but never really works when you look forward.
To be continue...
【在 f****e 的大作中提到】
: 大帅，how was your lunch? Still waiting to learn from you.
|
B*S
发帖数: 945 | 73 赞
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
b*******o
发帖数: 203 | 74 “winning in a trade in financial market is a matter of probablity but never
a sure thing. ”
说得很对，其实TA也是这样，模式只代表了一种概率，比如超买超卖。
如果单纯依靠TA，对于这一种概率的东西，只有大规模重复才有意义。 |
c**t
发帖数: 9197 | 75 WOW.
Tossing a coin is 50%.
TA can do at least 51%, then you got your edge.
But you are right that making money requires no strategy.
Keeping the money you made requires a good strategy.
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
c**t
发帖数: 9197 | 76 you do not need to "predict" the market to make money.
【在 g****u 的大作中提到】
: The market is "simple" in the sense that it's always about demand and
: supply and history repeats itself again and again in a very broad base.
: Beyond that, we can only say market is extremely complicated. The
: information presented and digested by the market is way way beyond
: human technology to ever process.
: Actually, human being can never master financial market because the
: purpose of observers is always to interfere. And the resulted interference
: by its nature is to invalid the original observation. It's like a cat
: chasing its own tail. In that sense, astronomy has a much brighter
: future because human interference is not possible in any foreseeable
|
e********d
发帖数: 1202 | 77 How come?
【在 c**t 的大作中提到】
: you do not need to "predict" the market to make money.
|
b******s
发帖数: 589 | 78 Truly a guru. plz continue:
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
t*****3
发帖数: 254 | 79 TA中有一些东西是具有高概率的预测作用的，当一些条件苛刻时，出现的几率下降，而
成功的概率上升，，，完全否定是肯定不对的。当然现在太多人并不真正懂得技术分析
的作用和有效性。
一个技术指标或形态，如果被许多人知道，那有效性就会大大减弱。但是有些东西是和
大众心理直接相关的。因此，即使很多人知道，但大多数人的心理还是和100年前的散
户相同，因此还是具有相当的有效性。
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
C****a
发帖数: 1639 | 80 I suggest you read more related literatures before wasting more time on
reinventing the wheels.
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
|
|
m********0
发帖数: 2717 | 81 大牛发话了。。。
【在 C****a 的大作中提到】
: I suggest you read more related literatures before wasting more time on
: reinventing the wheels.
|
g*****u
发帖数: 14294 | 82 Mastress offer some links 吧!
【在 C****a 的大作中提到】
: I suggest you read more related literatures before wasting more time on
: reinventing the wheels.
|
c********r
发帖数: 649 | |
C****a
发帖数: 1639 | 84 I think this book shows a nice big picture of what a trading system looks
like.
ie=UTF8&qid=1288306356&sr=8-1
【在 c********r 的大作中提到】
: http://www.amazon.com/Inside-Black-Box-Quantitative-Trading/dp/0470432063/ref=sr_1_1?ie=UTF8&qid=1288306356&sr=8-1
: Inside the Black Box: The Simple Truth About Quantitative Trading (Wiley
: Finance) [Hardcover]
|
t****3
发帖数: 2337 | 85 搞个类似A股大智慧level2的 DDX,Y,Z的模型就行了，关键没有类似的数据。 |
B********e
发帖数: 1062 | 86 忍不住上来说两句
1）要舍得花钱， 能买到的不要自己做， 当然买不起除外
你想做的东西，网上基本都有现成的。
2) 你的portofolio 太大， 搞个100只就足够了
3）弄个e-signal, 看看人家的TA 这么做的
4）弄个trading station, 检验一下你自己的想法。
5）多看书， 少编程;
6) 最好找个fund的工作， 工作一段时间就知道其他人是为了你好
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
c***n
发帖数: 133 | 87 做国内股票的话，lz想要的这些数据都是现成的。
用你的strategy在国内的股市上面成功率有多大？ |
c***n
发帖数: 133 | 88 做国内股票的话，lz想要的这些数据都是现成的。
用你的strategy在国内的股市上面成功率有多大？ |
p********3
发帖数: 33 | 89 握手握手。。我也在业余搞这方面的研究， 不过是研究汇市。。。
体会是个人搞自动交易基本不可行（关键是交易多多少少都是要人为干预的，机器交易
毕竟自适应性不足。日内交易的突发波动性噪音太重，），但是电脑搞筛选过滤和信号
是很好的。。
重要的是两部分。 1。 成功的选股和交易策略，这个手工应该有成熟的方案后，才有
意义。2.机器提供结果后人为分析进行交易。
问题1. 楼主有没有简单易行的SMS接口。有信号直接发到手机上 |
x****o
发帖数: 29677 | 90
汇市研究的结果如何
【在 p********3 的大作中提到】
: 握手握手。。我也在业余搞这方面的研究， 不过是研究汇市。。。
: 体会是个人搞自动交易基本不可行（关键是交易多多少少都是要人为干预的，机器交易
: 毕竟自适应性不足。日内交易的突发波动性噪音太重，），但是电脑搞筛选过滤和信号
: 是很好的。。
: 重要的是两部分。 1。 成功的选股和交易策略，这个手工应该有成熟的方案后，才有
: 意义。2.机器提供结果后人为分析进行交易。
: 问题1. 楼主有没有简单易行的SMS接口。有信号直接发到手机上
|
|
|
g*****u
发帖数: 14294 | 91 手机SMS发，你得搞个wireless modem，程序通过串口发就好了，发个SMS，简单AT命令
搞定。
这个要单独跟云英商开个号。
如果能通过互联网上的服务，发SMS就更好了。
【在 p********3 的大作中提到】
: 握手握手。。我也在业余搞这方面的研究， 不过是研究汇市。。。
: 体会是个人搞自动交易基本不可行（关键是交易多多少少都是要人为干预的，机器交易
: 毕竟自适应性不足。日内交易的突发波动性噪音太重，），但是电脑搞筛选过滤和信号
: 是很好的。。
: 重要的是两部分。 1。 成功的选股和交易策略，这个手工应该有成熟的方案后，才有
: 意义。2.机器提供结果后人为分析进行交易。
: 问题1. 楼主有没有简单易行的SMS接口。有信号直接发到手机上
|
l******h
发帖数: 405 | |
p*******0
发帖数: 420 | 93 我现在做这个系统目的主要也不是赚钱,而是通过做这个系统学习下TA。以前专业都是
分析报表什么的，后来发现做FA其实很难，要各个方面的信息，没有个team很难做好，
感觉TA好像一个人更容易做一些，所以现在又开始研究TA。
我知道市面上有很多这种软件，但我觉得至少要知道下这种软件是怎么运作，那些公式
怎么出来的用的才放心。
我也想找个fund干干，不过现在这种job market不知道fund的情况怎么样。而且我在DC
,现在这边大公司基本不要外国人了，换工作的可能性比较小。 |
p*******0
发帖数: 420 | 94 你google下 sms gate,很多公司都提供这种服务的，他们都有api |
t*******d
发帖数: 1186 | 95 多年前，我曾经想过，用clementine 做一个主要基于neural network的alert 系统。
当然不是全部股票，是有选择的目标股。
太忙，未果。如果被炒了，就来做。 |
N**t
发帖数: 1738 | 96 我不懂炒股, 但你说的这些跟玩poker非常像. 很难说服一个人玩的不对/不好, 如果他
玩了一次赢了一次, 玩了三次赢了三
次, 玩了十次赢了七次.
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
F******a
发帖数: 92 | 97 Hey, 股大帅,
Thanks a lot.
We are still waiting for the part 2/3/4/...
To be continue...
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
l******h
发帖数: 405 | 98 Wealth-Lab Pro provides a lot of TA indicators, but it allows you to define
your own TA indicators as plug-ins (DLLs) too. They are using C#.
http://personal.fidelity.com/products/trading/Trading_Platforms_Tools/Wealth_Lab_Pro_Overview-Tools.shtml.cvsr?refpr=brk23
DC
【在 p*******0 的大作中提到】
: 我现在做这个系统目的主要也不是赚钱,而是通过做这个系统学习下TA。以前专业都是
: 分析报表什么的，后来发现做FA其实很难，要各个方面的信息，没有个team很难做好，
: 感觉TA好像一个人更容易做一些，所以现在又开始研究TA。
: 我知道市面上有很多这种软件，但我觉得至少要知道下这种软件是怎么运作，那些公式
: 怎么出来的用的才放心。
: 我也想找个fund干干，不过现在这种job market不知道fund的情况怎么样。而且我在DC
: ,现在这边大公司基本不要外国人了，换工作的可能性比较小。
|
f****e
发帖数: 227 | 99 大帅, please continue. Not necessarily today but when you have time. I am sure
your generous post would save some of our young frogs' asses, if not lives.
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
m******t
发帖数: 635 | 100 来晚了，先支持下。
我也在做一个自动交易系统，第一步的目标和你类似，也就是一个screener。
数据从Yahoo Finance上抓，目前抓的是S&P 1500大概1500个股票的EOD数据。
split和dividend数据。刚作完adjusted close。下一步打算作stragegy和backtesting。
目前的系统用的Ruby写的，目前用CSV存EOD，下一步打算用C#重写一下（Ruby确实比较
慢）和Tokyo Cabinet （一种Key-Value store)存10min or 5min。
和前面几位的建议一样，如果是要作自动交易系统的话，strategy是最重要的，you
must have an edge. 另外从我做的research来看，tick data cleansing和order
management要比想象中的要难。
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
|
|
t****3
发帖数: 2337 | |
s*******r
发帖数: 35 | 102 Been doing this for 3 years (pet project), still not making money... |
g****u
发帖数: 695 | 103 All right, part 2, Technical Analysis.
I threw out a little hint at the end of part 1 by saying TA doesn't
work, and it got some defensive words back already. It will be
interesting to see what happens after I finish this part, because
if you accept what I am going to say, you probably will have to start
over in your trading adventure (which I think you should).
It is worth a few words to talk about what TA is trying to solve
first. Market price, no matter what the underlying product is, all
comes down to supply and demand. FA usually tries to evaluate the
underlying product and believes that supply and demand will eventually
converge to the true value. But evaluation itself is a very difficult
job if can be done at all, not even ask if the market price ever
converges to the true value. (I'll talk more about FA in next part.)
TA, in a broad sense, is trying to get around this evaluation part,
leave the job for other smart guys, and look at supply and demand
directly to guess the next move. I have to say it WAS a brilliant
idea, at least a good try, in the good old days (maybe 120 years ago).
In old days, people don't even have a chart in front of them. Drawing
the basic price chart at that time was a very professional job that
only a handful people or trading house can do. Being able to read a
chart was a huge advantage not many people can enjoy, that's why I
guess maybe it could work at that time. (Even though, the famous Dow
Jones Theory, the grand father of all TAs, born in 1890s, had been
already proved worse than simple buy-and-hold by 1920s.)
TA methods today can be roughly divided into two categories. I call
them 'chart TA' and 'book TA'. Chart TA is a set of methods based on
historical prices and volumes. Book TA is a set of method based on
exchange books (e.g. Nasdaq Level II quotes.) Chart TA can be further
divided into two sub-categories. I call them vertical analysis and
horizontal analysis.
Vertical analysis is the one attracting the enthusiasm from most of
you, i.e. look at the historical price and volume of a stock or index,
try to dig some information from it, be it patterns (e.g. head-shoulder,
flags, candlesticks, etc.), lines (e.g. trend lines, moving averages,
resistance/support, Elliott levels, Fibonacci, etc.), and their derivatives
(e.g. A/D lines, Arms index, MACD, etc.) I cannot list them all because
there are hundreds of them. If you read Technical Analysis of Stocks &
Commodities Magazine, you'll see new TA indicators are being invented by
somebody somewhere every single day, and they all look great!
Vertical analysis has been applied by traders in all time frames,
from years to ticks. John Magee in his legendary book Technical Analysis
of Stock Trends (first published in 1940s, still being read by readers
overall the world today) suggested that round top and SHS are only
meaningful when formed over 1 or 2 years, and round bottom can take
even much longer. And he gave very attention to the volume pattern
accompanying these price patterns, claiming that without the correct
volume setup the price pattern cannot be trusted. Yet after so many
years of 'evolution', I see people commonly refer to these patterns
in minute charts without even look at the volume. (I am not trying to
say John Magee is right, just to present a spot that many popular
TA methods actually come from nowhere, not even the original idea
of the inventor.)
Horizontal analysis is a group of methods trying to profit from the
correlation between stocks or markets. It includes a lot of different
methods. (You may be thinking long short hedge fund, but no, they
are actually more on the FA side.) I have seen people doing things
from as simple as plotting the charts of two stocks together and
visually identify the correlation to as complicated as using some
statistic theory that I never heard of and a few 16-cpu servers to
trade a basket of stocks and commodities. There might be dozens of
ways being tried but they all come down to one idea: correlation
in prices without fundamental.
Book TA is probably more advanced to most of people on this board
although some of you might have tried to learn reading Level II or
ECN book. The idea is to look at the book or market depth at the moment
to identify the location of liquidity, either to execute large chunk of
order or to predict the movement for the very next moment. This type of
TA by its nature can only be applied to high frequency trading. However,
it's pretty much all they care in many sell-side trading house and market
makers (where the word MM comes from and very much associated with market
manipulation, although they are absolutely not most of you have thought),
and thus an area of real professionals. This is an extremely competent
battle field which has absolutely no space for civilian traders. I knew
trading houses full of PhDs paying millions of dollars a year just to get
milli-seconds ahead of others to access certain market. You get the idea.
Book TA is a pure mathematic science. Which stock you are trading or even
what you are trading is actually not relevent. The touch of this area
is both out of retail traders' capability and out of their interest. So
I am not going to talk about it any more. There are always somebody
believe that they are smart enough to try it. I'll say that's like a
civilian with or even without a gun trying to fight an army.
I'll go back to talk about chart TA again.
To be continue...
【在 g****u 的大作中提到】
: Okay, here is what I suggest (and also what I am practicing):
: 1. The first of first is to keep in mind that winning in a trade in
: financial market is a matter of probablity but never a sure thing.
: Now everybody will say "I don't need you to tell me this, I know
: from the beginning." Well, I know you know, but you didn't really
: keep it mind, because when it comes to the next question, not many
: people really use their brain.
: The next question is: if somebody made some money in one trade or a
: few trades, is it because they had a right strategy? or because they
: are nothing but lucky.
|
g*****u
发帖数: 14294 | 104 Very interesting read!
【在 g****u 的大作中提到】
: All right, part 2, Technical Analysis.
: I threw out a little hint at the end of part 1 by saying TA doesn't
: work, and it got some defensive words back already. It will be
: interesting to see what happens after I finish this part, because
: if you accept what I am going to say, you probably will have to start
: over in your trading adventure (which I think you should).
: It is worth a few words to talk about what TA is trying to solve
: first. Market price, no matter what the underlying product is, all
: comes down to supply and demand. FA usually tries to evaluate the
: underlying product and believes that supply and demand will eventually
|
f****e
发帖数: 227 | 105 大帅, thank you so much for providing such great pro bono service to this
board. It was very informative (to me at least). Probably not all of the
readers would agree with you but I am pretty sure they are all looking
forward to your next part.
【在 g****u 的大作中提到】
: All right, part 2, Technical Analysis.
: I threw out a little hint at the end of part 1 by saying TA doesn't
: work, and it got some defensive words back already. It will be
: interesting to see what happens after I finish this part, because
: if you accept what I am going to say, you probably will have to start
: over in your trading adventure (which I think you should).
: It is worth a few words to talk about what TA is trying to solve
: first. Market price, no matter what the underlying product is, all
: comes down to supply and demand. FA usually tries to evaluate the
: underlying product and believes that supply and demand will eventually
|
m******t
发帖数: 635 | 106 Thanks for the post, very informative. It deserves to be a seperate blog
entry.
If I understand you corretly, the book TA you mentioned is actually
quantative trading, where you can apply some more sophisticated strategies
over a relative bigger number of stocks or other instruments. Though myself
is a very newbie in this arena, I personally don't think those a bit more
advanced topics such as pair trading, risk management, Sharpe ratio,
backtesting are beyond the capability of retail traders like us. You do need
to know what a game you're going to play tho.
Just my 2c, look foward to your part3.
【在 g****u 的大作中提到】
: All right, part 2, Technical Analysis.
: I threw out a little hint at the end of part 1 by saying TA doesn't
: work, and it got some defensive words back already. It will be
: interesting to see what happens after I finish this part, because
: if you accept what I am going to say, you probably will have to start
: over in your trading adventure (which I think you should).
: It is worth a few words to talk about what TA is trying to solve
: first. Market price, no matter what the underlying product is, all
: comes down to supply and demand. FA usually tries to evaluate the
: underlying product and believes that supply and demand will eventually
|
g****u
发帖数: 695 | 107 Book TA is not what you have mentioned, not at all. I don't know why
you get that impression... please read again. If you are interested
what it's really about, you have to look at the market depth, Nasdaq
Level II quote, or ECN books. If you don't really know what these
things are, it's okay to forget them altogether. As I said, it's not
a field for retail traders.
Pair trading 应该在 horizontal chart TA 的范畴里。Just remember,
whenever you forget fundamental and look into historical price for
answers, you are in chart TA.
Risk management is another story, not related to what I am trying to
solve here. Backtesting is not only in the capability of retail traders
but also something they should have done before they use any TA indicators.
If they ever do that seriously, they will understand none of TA works.
I'll talk about it in more detail later.
myself
need
【在 m******t 的大作中提到】
: Thanks for the post, very informative. It deserves to be a seperate blog
: entry.
: If I understand you corretly, the book TA you mentioned is actually
: quantative trading, where you can apply some more sophisticated strategies
: over a relative bigger number of stocks or other instruments. Though myself
: is a very newbie in this arena, I personally don't think those a bit more
: advanced topics such as pair trading, risk management, Sharpe ratio,
: backtesting are beyond the capability of retail traders like us. You do need
: to know what a game you're going to play tho.
: Just my 2c, look foward to your part3.
|
g*****u
发帖数: 14294 | 108 如何解释蒙古大夫的成功率？去年到最近。
用的不外乎pattern等等，算chart TA吧。
【在 g****u 的大作中提到】
: Book TA is not what you have mentioned, not at all. I don't know why
: you get that impression... please read again. If you are interested
: what it's really about, you have to look at the market depth, Nasdaq
: Level II quote, or ECN books. If you don't really know what these
: things are, it's okay to forget them altogether. As I said, it's not
: a field for retail traders.
: Pair trading 应该在 horizontal chart TA 的范畴里。Just remember,
: whenever you forget fundamental and look into historical price for
: answers, you are in chart TA.
: Risk management is another story, not related to what I am trying to
|
M*M
发帖数: 1993 | 109 损人不带脏字，赞
【在 m********0 的大作中提到】
: 大牛发话了。。。
|
m******t
发帖数: 635 | 110 I re-read your part2 post and found out that you did have one paragraph
talking about 'advanced TA' (you called it horizonal chart TA) and I totally
missed it the first time. Now I got it. Thanks for clarification.
【在 g****u 的大作中提到】
: Book TA is not what you have mentioned, not at all. I don't know why
: you get that impression... please read again. If you are interested
: what it's really about, you have to look at the market depth, Nasdaq
: Level II quote, or ECN books. If you don't really know what these
: things are, it's okay to forget them altogether. As I said, it's not
: a field for retail traders.
: Pair trading 应该在 horizontal chart TA 的范畴里。Just remember,
: whenever you forget fundamental and look into historical price for
: answers, you are in chart TA.
: Risk management is another story, not related to what I am trying to
|
|
|
g****u
发帖数: 695 | 111 I try to not make this personal against anybody. It's a general idea that TA
doesn't work for
long-term profitability. As I said in the first part, good results can come
from bad method
for an extended period. I was expecting this kind of question, that's why I
wrote the first
part before anything else.
I didn't track Lexian systematically, so I don't want to make comment
specifically for him
or the method he is using.
【在 g*****u 的大作中提到】
: 如何解释蒙古大夫的成功率？去年到最近。
: 用的不外乎pattern等等，算chart TA吧。
|
M*M
发帖数: 1993 | 112 大帅，in your opinion, how to get long-term profitability?
Is there an easy way that a frog like me can follow? Thanks.
TA
come
I
【在 g****u 的大作中提到】
: I try to not make this personal against anybody. It's a general idea that TA
: doesn't work for
: long-term profitability. As I said in the first part, good results can come
: from bad method
: for an extended period. I was expecting this kind of question, that's why I
: wrote the first
: part before anything else.
: I didn't track Lexian systematically, so I don't want to make comment
: specifically for him
: or the method he is using.
|
z****e
发帖数: 54598 | 113 这些东西用java来做还可以结合php和cpp的优点
还可以包装成产品出售
【在 p*******0 的大作中提到】
: 以前做了一个项目，用php+mysql，根据成交量选择了进2000只股票，每15分钟扫描一
: 次，包括pre和
: after market，加上option的监视和分析。数据收集了半年左右吧，发现数据量太大了
: ，mysql受不
: 了，一个表就几十G，随便查个什么数据都要很长时间。还有就是php运算比较慢。
: 现在准备用c++和mysql重新做(做好以后可能升级到oracle,不是很难)，理论上保守估
: 计新系统要比以前
: 快几十倍(c++ vs php)。新系统我准备每天扫描所有股票3到10遍，从所有股票中找到
: 100只以内值得做
: 的做5分钟的跟踪。其他都是按天分析。
|
S*L
发帖数: 1746 | 114 嗯,Java 不错，很多基金现在也用了。
【在 z****e 的大作中提到】
: 这些东西用java来做还可以结合php和cpp的优点
: 还可以包装成产品出售
|
