|
PureBytes Links
Trading Reference Links
|
Yuki,
Run only these two lines, I mean (explore):
Filter = 1;
AddColumn( TimeNum(), "TimeNum" );
You should see lots of output this way.
Later you can run:
Filter = TimeNum() < 120000;
AddColumn( TimeNum(), "TimeNum" );
Best regards,
Tomasz Janeczko
amibroker.com
----- Original Message -----
From: "Yuki Taga" <yukitaga@xxxxxxxxxxxxx>
To: "Tomasz Janeczko" <amibroker@xxxxxxxxxxxxxxx>
Sent: Thursday, February 08, 2007 10:39 AM
Subject: Re: [amibroker] How stupid can I be???
> Hi Tomasz,
>
> Nothing is returned. Nothing for any N number of days, nothing for
> any selected date range, nothing for "all quotations".
>
> Periodicity is *definitely* 1 minute, and I am *definitely* in the
> eSignal database.
>
> Now, if I go and change Periodicity to Daily and run explore, all
> proper trades (symbols and dates are returned). But all have a
> TimeNum of 150,000.00 and all are "timestamped" at 3 PM on the date
> of entry.
>
> Yuki
>
> Thursday, February 8, 2007, 6:18:40 PM, you wrote:
>
> TJ> Yuki,
>
> TJ> As usual, EXPLORATION mode is for you to help.
>
> TJ> Add these lines:
>
> TJ> Filter = Buy;
>
> TJ> AddColumn( TimeNum(), "TimeNum" )
>
> TJ> And remove temporarily 'AND Timenum() <= 120000" from the Buy rule.
>
> TJ> Then run "Explore" it will display what TimeNum() values you really have.
> TJ> Maybe you are running with Periodicity set to daily (in the Settings),
> TJ> anyway Exploration will tell you.
>
> TJ> Best regards,
> TJ> Tomasz Janeczko
> TJ> amibroker.com
> TJ> ----- Original Message -----
> TJ> From: "Yuki Taga" <yukitaga@xxxxxxxxxxxxx>
> TJ> To: "Tomasz Janeczko" <amibroker@xxxxxxxxxxxxxxx>
> TJ> Sent: Thursday, February 08, 2007 9:48 AM
> TJ> Subject: Re: [amibroker] How stupid can I be???
>
>
>>> Hi Tomasz,
>>>
>>> I don't know what to say. It doesn't work.
>>>
>>> Your assumption 1 is correct, I believe: see the
>>> IntradaySettings.png.
>>>
>>> Your assumption 2 is *absolutely* correct. In fact, I do not exit
>>> same day, anyway (unless I override, which I rarely do). Here is my
>>> exit:
>>>
>>> ApplyStop( stopTypeNBar, stopModeBars, delay, True, False );
>>>
>>> I have had this code a long time, and RT results and trading results
>>> are never off by even one single yen -- other than fast-market trade
>>> entry miss or something like that. But it is accurate, realistic,
>>> and works -- the code I mean. But I cannot, using my eSignal
>>> one-minute database, isolate AM and PM entries. I have tried, on my
>>> own, to do this FOREVER, and I cannot do it.
>>>
>>> All I do is two things (other than what I would do with my master EOD
>>> database):
>>>
>>> 1) Change periodicity to 1 minute (see file)
>>>
>>> 2) Append 'AND Timenum() <= 120000;' to the end of the Buy statement.
>>>
>>> Then I backtest.
>>>
>>> But I generally get *no* trades as a result when I backtest after
>>> doing this. For example, using either last n days = 2, or using From
>>> 2/7/2007 To 2/8/2007, I get no trades. That's wrong. There were
>>> four signals, three yesterday and one today. One of the trades on
>>> 2/7 (yesterday) was at about 9:30 AM. I didn't dream it.
>>>
>>> Looking back, say, 100 bars, where I would have dozens and dozens of
>>> trades, I get maybe 4 to show up. That's not realistic. It's flat
>>> out way wrong. I know from years of experience most of my entries
>>> come in the AM session. Off hand, I'd say it's 2-1 or higher.
>>> Naturally, now that I am accumulating a longer and longer intraday
>>> database, I'd like to isolate these instances and test them.
>>>
>>> I cannot.
>>>
>>> Using the RT database in daily mode (with periodicity set at daily),
>>> there is no problem backtesting. But of course I'm *not* able to
>>> isolate signals by using Timenum() that way. Not that I can isolate
>>> them in any case, mind you.
>>>
>>> This is, to say the least, excruciatingly frustrating for me. I
>>> don't believe I am completely stupid, obviously, and I cannot see why
>>> this (apparently) simple little thing will not work for me.
>>>
>>> This is the *reason* I bought the RT version of AB years ago, and why
>>> I started subscribing to eSignal immediately when it became available
>>> in Japan. And on top of that, I knew the first few years of
>>> subscription would only serve to build up a database, and that I
>>> could not do realistic intraday testing until I had sufficient
>>> instances and data to draw reasonably valid conclusions from.
>>>
>>> I want to throw up now. ^_^
>>>
>>> Yuki
>>>
>>> Thursday, February 8, 2007, 4:45:56 PM, you wrote:
>>>
>>> TJ> Yuki,
>>>
>>> TJ> You code is correct assuming that
>>> TJ> 1. You don't use time shift (File->Database Settings->Intraday Settings)
>>> TJ> 2. You mean that your ENTRY is limited to AM session
>>> TJ> (the code cares only about Buy signal, it does not limit you from
>>> TJ> exiting later in the PM session, you would need to write
>>> TJ> condition for EXIT to close positions before 12 PM.
>>>
>>> TJ> Best regards,
>>> TJ> Tomasz Janeczko
>>> TJ> amibroker.com
>>> TJ> ----- Original Message -----
>>> TJ> From: "Yuki Taga" <yukitaga@xxxxxxxxxxxxx>
>>> TJ> To: <amibroker@xxxxxxxxxxxxxxx>
>>> TJ> Sent: Thursday, February 08, 2007 5:13 AM
>>> TJ> Subject: [amibroker] How stupid can I be???
>>>
>>>
>>>>> This program really makes me feel like an idiot sometimes. But this
>>>>> idiot mops up tens of millions of yen annually from the local equity
>>>>> market, so she can't be *that* stupid. Right?
>>>>>
>>>>> Nonetheless:
>>>>>
>>>>> I am trying to add what -- I (probably stupidly) think -- should be a
>>>>> simple qualifier to existing, known-good code.
>>>>>
>>>>> Simply, we have two sessions in Tokyo, AM & PM. Since *nothing* ever
>>>>> trades at exactly 12 PM (market is closed), I used that for the
>>>>> divider.
>>>>>
>>>>> Buy = (all my secret Rocky The Flying Squirrel stuff) AND Timenum()
>>>>> <= 120000;
>>>>>
>>>>> Backtesting with an interval setting of one minute *must* show *only*
>>>>> AM trades. Right? Wrong? Do I need neurosurgery? Do I need a whap
>>>>> on the head with a Whack-A-Mole mallet?
>>>>>
>>>>> (BTW, Whack-A-Mole is *extremely* popular in Tokyo, but we call it
>>>>> something else, of course. I'm trying to get it called
>>>>> "Whack-A-Politician" -- little Abe-san heads would pop up -- but I'm
>>>>> not having much success, despite his plummeting popularity.)
>>>>>
>>>>> Can anyone fix my personal, intra-cranial neural network? If anyone
>>>>> could help "girl genius" here, she'd be very appreciative.
>>>>>
>>>>> My best Bullwinkle The Moose voice: "This time for *sure*!"
>>>>>
>>>>> Yuki
>>>>>
>>>>>
>>>>>
>>>>> Please note that this group is for discussion between users only.
>>>>>
>>>>> To get support from AmiBroker please send an e-mail directly to
>>>>> SUPPORT {at} amibroker.com
>>>>>
>>>>> For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
>>>>> http://www.amibroker.com/devlog/
>>>>>
>>>>> For other support material please check also:
>>>>> http://www.amibroker.com/support.html
>>>>>
>>>>> Yahoo! Groups Links
>>>>>
>>>>>
>>>>>
>>>>>
>>>>>
>>>
>>>
>>> Best,
>>>
>>> Yuki
>>>
>>> Please note that this group is for discussion between users only.
>>>
>>> To get support from AmiBroker please send an e-mail directly to
>>> SUPPORT {at} amibroker.com
>>>
>>> For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
>>> http://www.amibroker.com/devlog/
>>>
>>> For other support material please check also:
>>> http://www.amibroker.com/support.html
>>>
>>> Yahoo! Groups Links
>
>
>
>
> Please note that this group is for discussion between users only.
>
> To get support from AmiBroker please send an e-mail directly to
> SUPPORT {at} amibroker.com
>
> For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
> http://www.amibroker.com/devlog/
>
> For other support material please check also:
> http://www.amibroker.com/support.html
>
> Yahoo! Groups Links
>
>
>
>
>
Content-Description: "AVG certification"
No virus found in this incoming message.
Checked by AVG Free Edition.
Version: 7.5.432 / Virus Database: 268.17.30/674 - Release Date: 2/7/2007 3:33 PM
|