mmaurice
51 posts
msg #105084
- Ignore mmaurice |
2/21/2012 9:52:57 AM
Just an FYI, if you are using a Google doc to find your entries, they were slow this morning on their updates. Took 20 minutes to get GOOD data. Usually they are much faster with the data stream but. . .
Always verify what you are seeing before you trade. Measure twice, cut one.
|
mmaurice
51 posts
msg #105085
- Ignore mmaurice |
2/21/2012 10:03:25 AM
Kevin,
Are the filters posted on the 1st page of this thread consistant with the current version??
I remember reading somewhere you had some criteria you had forgotten but in review I can not find that post.
Thanks
|
Kevin_in_GA
4,599 posts
msg #105086
- Ignore Kevin_in_GA |
2/21/2012 10:38:57 AM
Stockfetcher is giving me the following results (caveat - the opening prices they quote have been revised far too frequently for my taste).
Gap Down:
No stocks selected for this week.
Gap Up:
HDY - open price at 1.42, target of 1.38
AMSC - open price of 4.87, target of 4.70, (GAP ALREADY CLOSED FOR A GAIN OF 3.49%)
VHC - open price of 25.20, target of 24.61 (GAP ALREADY CLOSED FOR A GAIN OF 2.34%)
DYN - open price of 1.50, target of 1.41 (GAP ALREADY CLOSED FOR A GAIN OF 6.00%)
CTIC - open price of 1.50, target of 1.35
These are the top five based on PCT_GAPFILL.
Are people seing the same thing?
|
xzajic
14 posts
msg #105087
- Ignore xzajic |
2/21/2012 10:50:11 AM
Olny VHC; others has too low volume and "no shares available for short sale" ;-)
|
mmaurice
51 posts
msg #105088
- Ignore mmaurice |
2/21/2012 11:00:05 AM
I ran the Gap Down filter this morning and found 2
MBLX
CGNX
|
Kevin_in_GA
4,599 posts
msg #105089
- Ignore Kevin_in_GA |
2/21/2012 11:26:31 AM
Volume too low?
Gap Up volumes as of 11:15 AM:
HDY - 1.58M
AMSC - 98,315
VHC - 190,713
DYN - 283,585
CTIC - 3.797M
These seem reasonable, especially given that we are less than half way through the day. I can't comment on whether or not shares are available for shorting, since this is not known for any filter that I have seen.
I appreciate your feedback on this, as long as it is intended to improve the result for everyone here. This is a work in progress - I am testing this out moving forward and trying to modify/adapt to make it better for real use.
|
xzajic
14 posts
msg #105093
- Ignore xzajic |
2/21/2012 2:54:28 PM
Kevin, I appreciate your idea, but at the moment just can not reasonably be used in real trading.There are huge problems in "real use" of short side: liquidity and dividends. I think we cannot backtest short trades precisely.
And there is another problem on both sides - one simply cannot enter 30 or more "LIMIT ON OPEN" orders, because ALL orders may be executed and you can get margin call (for example 2011-09-12 generated many candidates, etc.)
At the market open, I can't say "these five are the best candidates", or "those are not".
Worse, you can't open position in 9:45, because on that time gap may be already filled. So, from my point of view, LIMIT-ON-OPEN is the only order type suitable for this strategy.
I've one idea and going to backtest it: Generate FIXED AMOUNT of candidates for LONG and SHORT every week. Say 10 LONG candidates and 10 SHORT candidates. And I'm going to write a small software utility (based on IB API here: http://www.dinosaurtech.com) that will allow me to place orders programatically into IB platform. If I succeed, I will try to describe the process or provide some source code here.
|
Kevin_in_GA
4,599 posts
msg #105094
- Ignore Kevin_in_GA
modified |
2/21/2012 3:41:58 PM
The issues on short side trading systems are endemic - every strategy is subject to them, mine included. The only way I can see to do this here is to add "optionable" as the first line of code (this might help to address the limit on shares available to short, as you had pointed out in an earlier post).
As to this system's applicability in trading long side, I envision this being handled in Ninjascript or equivalent and run as a basket of stocks. The code is currently non-existent since I do not program in C or its variations. This is on my list for 2012 but who knows ...
I'd suggest that if you are going to write code for automated trading it be given its own thread even if based off of these concepts. That way others can add or help in that without too much cross-chatter on this SF version.
|
Kevin_in_GA
4,599 posts
msg #105118
- Ignore Kevin_in_GA |
2/23/2012 11:38:03 AM
All gaps this week were filled. There was (again) some jockeying of opening prices that took a day or two to get finalized. This is a BIG issue if you are using SF data at the open - which is not really possible since it is delayed by 30+ minutes on average and not finalized till the end of the day.
My thoughts on this are that one would use an automated rading program like NinjaTrader to enter these based on real-time data feeds. I'm trying to learn NT and set up a simulated trading strategy for this to test it out under more realistic conditions. The coding here should be fairly straightforward, but I'm speaking without any experience.
Anyone here have experience coding in Ninjascript and would like to collaborate? If so, let me know here or send me a note at statisticalinvesting@gmail.com.
Kevin
|
xzajic
14 posts
msg #105135
- Ignore xzajic |
2/25/2012 10:07:17 AM
I backtested this strategy and discover following:
If I try to trade (hypotetically) EVERY gap on EVERY RUSSEL 2000 stock from January 2000 till today, results are really amazing:
Long side: 136000 trades, 87.6% winning, NO unprofitable year
Short side: 33243 trades, 76.0% winning, NO unprofitable year
But every possible attempt to LIMIT number of trades to some "candidates" leads to significant losses.
For example, if I try to trade only 50 "best candidates" every week (close > 1, 70% or more gapfills in past 80 weeks, 3 or more gaps in past 80 weeks), results are:
Long side: 2147 trades, 83.74% winning, 2 of 10 years unprofitable
Short side: 1644 trades, 81.07=% winning, 2 of 10 years unprofitable
Double checked results on two independent backtesting software (RightEdge an my own) and testing with WEEKLY AND/OR DAILY DATA and still no luck.
Kevin, you must have some "secret formula" for "candidate picking", or my historical data must be wrong, or I'm mising something really important.
Can you please post some of your backtest result with symbols and entry dates? Something like "recent trades" on your site, but for a longer period (year 2011, maybe)? If so, I'l try to mimic your trades one by one (and maybe find some bugs in my calculations)...
Thanks in advace.
|