i********y 发帖数: 34 | 1 想做些Backtesting,不知道有什马软件比较好用,能灵活定义测试条件?
我目前知道的有Ninjatrader和AmiBroker。有没有其他更好的? |
a*****e 发帖数: 1717 | 2 openquant, wealth-lab, tradestation, etc
really depends on your frequency.
for minute to second level trading,
I found no commercial package really works.
【在 i********y 的大作中提到】 : 想做些Backtesting,不知道有什马软件比较好用,能灵活定义测试条件? : 我目前知道的有Ninjatrader和AmiBroker。有没有其他更好的?
|
w******s 发帖数: 16209 | 3 我有in house second/sub second的。
【在 a*****e 的大作中提到】 : openquant, wealth-lab, tradestation, etc : really depends on your frequency. : for minute to second level trading, : I found no commercial package really works.
|
f********n 发帖数: 1560 | 4 这个是不是要在NASDAQ或者NYSE租个服务器?好像亚马逊的也凑着。
【在 w******s 的大作中提到】 : 我有in house second/sub second的。
|
a*****e 发帖数: 1717 | 5 for these level, backtesting won't work.
Because you really have to consider slippage, market impact now.
I don't see how that happen with any simulator.
I mean even IB automatic trading paper account is debatable.
The whole mechanism is still black-box.
You don't know under what condition, your MKT/LMT order will be filled
by your paper account. And I don't think it's reasonable to make
any assumption.
【在 w******s 的大作中提到】 : 我有in house second/sub second的。
|
f********n 发帖数: 1560 | 6 小散什么样的可以?15分钟以上的?
【在 a*****e 的大作中提到】 : for these level, backtesting won't work. : Because you really have to consider slippage, market impact now. : I don't see how that happen with any simulator. : I mean even IB automatic trading paper account is debatable. : The whole mechanism is still black-box. : You don't know under what condition, your MKT/LMT order will be filled : by your paper account. And I don't think it's reasonable to make : any assumption.
|
w******s 发帖数: 16209 | 7 我知道你说的,我只能试图一些模拟来work around,书本读的少,逗是些土办法。
没法真正意义的完全backtesting,但还是很有用的。好机会在现实中总是难fill的。
【在 a*****e 的大作中提到】 : for these level, backtesting won't work. : Because you really have to consider slippage, market impact now. : I don't see how that happen with any simulator. : I mean even IB automatic trading paper account is debatable. : The whole mechanism is still black-box. : You don't know under what condition, your MKT/LMT order will be filled : by your paper account. And I don't think it's reasonable to make : any assumption.
|
a*****e 发帖数: 1717 | 8 depends what you trade,
I guess 2-5 minute will still be relatively ok for top liquid stuff.
【在 f********n 的大作中提到】 : 小散什么样的可以?15分钟以上的?
|
a*****e 发帖数: 1717 | 9 it's still not bad if it's just opportunity cost,
like I mentioned before, you could be stopped nowhere if you use stp,
and failed to fill if you use stplmt.
not sure, if one of the solution would be stplmt but keeps reducing(for long)
lmt if price gap down beyond the limit.
so, to backtest one minute level strategies, one really needs tick data,
one minute bar won't be sufficient.
【在 w******s 的大作中提到】 : 我知道你说的,我只能试图一些模拟来work around,书本读的少,逗是些土办法。 : 没法真正意义的完全backtesting,但还是很有用的。好机会在现实中总是难fill的。
|
g*****u 发帖数: 14294 | 10 Village Head is a high hand, indeed!
【在 w******s 的大作中提到】 : 我知道你说的,我只能试图一些模拟来work around,书本读的少,逗是些土办法。 : 没法真正意义的完全backtesting,但还是很有用的。好机会在现实中总是难fill的。
|
|
|
a*****e 发帖数: 1717 | 11 I think a lot of empirical stuff works better.
Trading has far less fun if it's a pure science.
pure science only exists conceptually.
【在 w******s 的大作中提到】 : 我知道你说的,我只能试图一些模拟来work around,书本读的少,逗是些土办法。 : 没法真正意义的完全backtesting,但还是很有用的。好机会在现实中总是难fill的。
|
w******s 发帖数: 16209 | 12 for stop, I think you need to use your own mechanism to stop.
the one you submit to the broker earlier is visible to others and could be
used to trigger..
my stop algorithm is quite complicated and always do much better job than
myself.
long)
的。
【在 a*****e 的大作中提到】 : it's still not bad if it's just opportunity cost, : like I mentioned before, you could be stopped nowhere if you use stp, : and failed to fill if you use stplmt. : not sure, if one of the solution would be stplmt but keeps reducing(for long) : lmt if price gap down beyond the limit. : so, to backtest one minute level strategies, one really needs tick data, : one minute bar won't be sufficient.
|
f********n 发帖数: 1560 | 13 看不懂,能不能用例子,动画或者录像解释一下?谢谢
long)
【在 a*****e 的大作中提到】 : it's still not bad if it's just opportunity cost, : like I mentioned before, you could be stopped nowhere if you use stp, : and failed to fill if you use stplmt. : not sure, if one of the solution would be stplmt but keeps reducing(for long) : lmt if price gap down beyond the limit. : so, to backtest one minute level strategies, one really needs tick data, : one minute bar won't be sufficient.
|
a*****e 发帖数: 1717 | 14 stop limit orders are probably candies for market maker.
I felt the same way, but haven't think it through.
I think your are probably right in most case, in a relative
low volatile market, stop by your own instead of submit a
child stop loss order is most likely better.
But for high volatile market, it's a trade-off, depends on
how fast your program is.
【在 w******s 的大作中提到】 : for stop, I think you need to use your own mechanism to stop. : the one you submit to the broker earlier is visible to others and could be : used to trigger.. : my stop algorithm is quite complicated and always do much better job than : myself. : : long) : 的。
|
w******s 发帖数: 16209 | 15 yeah.. bad opportunity is easier to fill..
market is always right..
after watch the movie of "source code", whenever I get good stuff not filled
when I was in the queue, I will just think that in another m universe, this
order is filled. and that village head is happily counting money already.
long)
【在 a*****e 的大作中提到】 : it's still not bad if it's just opportunity cost, : like I mentioned before, you could be stopped nowhere if you use stp, : and failed to fill if you use stplmt. : not sure, if one of the solution would be stplmt but keeps reducing(for long) : lmt if price gap down beyond the limit. : so, to backtest one minute level strategies, one really needs tick data, : one minute bar won't be sufficient.
|
a*****e 发帖数: 1717 | 16 that's a good point,
as I felt I was taken advantage of by market maker more frequent
than I thought.
Regarding ES, the spread is 99.5% one tick.
I really don't know how market makers profit unless majority orders
are from market makers.
【在 w******s 的大作中提到】 : for stop, I think you need to use your own mechanism to stop. : the one you submit to the broker earlier is visible to others and could be : used to trigger.. : my stop algorithm is quite complicated and always do much better job than : myself. : : long) : 的。
|
a*****e 发帖数: 1717 | 17 is that a good movie?
filled
this
【在 w******s 的大作中提到】 : yeah.. bad opportunity is easier to fill.. : market is always right.. : after watch the movie of "source code", whenever I get good stuff not filled : when I was in the queue, I will just think that in another m universe, this : order is filled. and that village head is happily counting money already. : : long)
|
w******s 发帖数: 16209 | 18 i like it a lot. mind uploading and parrellel universe.. jake did very well
in that movie.
【在 a*****e 的大作中提到】 : is that a good movie? : : filled : this
|
i********y 发帖数: 34 | 19 多谢,主要是中长线,不需要DT。
我看到往上说不少软件做Backtesting都有bug,几次运行结果不一致,所以想问一下有
没有人这方面有经验
【在 a*****e 的大作中提到】 : openquant, wealth-lab, tradestation, etc : really depends on your frequency. : for minute to second level trading, : I found no commercial package really works.
|
w******s 发帖数: 16209 | 20 tradestation is pretty good already.
if there is some probability involved, you may not see the same result every
time.
【在 i********y 的大作中提到】 : 多谢,主要是中长线,不需要DT。 : 我看到往上说不少软件做Backtesting都有bug,几次运行结果不一致,所以想问一下有 : 没有人这方面有经验
|