Kevin_in_GA
4,599 posts
msg #96744
- Ignore Kevin_in_GA |
10/5/2010 1:34:26 PM
I was referring to the filters used in this thread. Also, it makes sense to go dollar-neutral if possible (except for today, when everything went up).
As to the Stratasearch filters, I have focused on developing "all-long" filters to date. The ability for the program to backtest massive numbers of indicator combinations is really impressive.
Here's one I pulled from a recent optimization run
Exit when either the MA(7) crosses above the MA(40) or when the rmi(4,14) crosses above 45.
Backtest results for the last 4 months (arguably some of the hardest months to trade profitably given the ups and downs seen):
There were 149 total stocks entered. Of those, 133 or 89.26% were complete and 16 or 10.74% were open.
Of the 133 completed trades, 123 trades or 92.48% resulted in a net gain.
Your average net change for completed trades was: 4.60%.
The average draw down of your approach was: -4.40%.
The average max profit of your approach was: 6.19%
The Reward/Risk ratio for this approach is: 13.06
Annualized Return on Investment (ROI): 94.86%, the ROI of ^SPX was: 10.33%.
The stats for this simple filter are really impressive, and there's no way I would have come up with this specific combination.
I'll post a series of these for everyone to evaluate (at least those that can be readily translated into SF code).
Kevin
|
einok
msg #100440
- Ignore einok |
4/27/2011 12:41:23 AM
Hi Kevin,
How exactly is this a long filter?
market is S&P 500
close above 1
average volume(20) above 200000
MA(7) below MA(40)
rmi(4,14) crossed below 89
Exit when either the MA(7) crosses above the MA(40) or when the rmi(4,14) crosses above 45.
I see the exit, but what is then entry? When MA(7) below MA(40) and
rmi(4,14) crossed below 89? That seems short to me.
Are you working on any other filters with stratasearch?
|
duke56468
683 posts
msg #100444
- Ignore duke56468 |
4/27/2011 10:20:05 AM
Kevin these 2 filters of yours produce impressive backtest results, do you feel one is superior to the other in most markets? Thanks for sharing your work, I would have never come up with those exit indicators, and like ewhiteman said even the entry seems counter intuitive.
|
Kevin_in_GA
4,599 posts
msg #100614
- Ignore Kevin_in_GA |
5/6/2011 12:27:16 PM
These are a few filters that StrataSearch identified. I ran over 2 million different indicators settings/combinations on the S&P 500 for about a full year's worth of data (2009 - 2010). Long only filters, which weren't hard to find given the market uptrend during that time window.
I agree that both the entries and exits seem to defy common logic, but as you pointed out you can't argue with the results. As to which of the two RMI filters is better for a given period, I did not really look. These are part of a 10 filter system I put together late last year, which has done quite well. Unfortunately, most of them are not easily coded using SF as they require sector data and indictors that are not commonly shared between the two programs.
My suggestion - give StrataSearch a try for a few months and see if you like it (fee is $30/month, which you'll most likely make back in your first trade or two).
|
einok
msg #100617
- Ignore einok |
5/6/2011 1:10:49 PM
Hey Kevin,
Are you familiar much with excel charting? Do you use skype? I
|
RocketFetcher
59 posts
msg #115328
- Ignore RocketFetcher |
9/12/2013 6:42:55 PM
Didn't know whether to start a new thread, as this one is dated, but, it's on the topic and a lot of posts here, so...
Does anyone have the ConnorsRSI filter for StockFetcher?
At TradingMarkets, they have started publishing some code for TOS, and a couple of others,
Does anyone have it working here for StockFetcher?
|
DMUNCASTER
34 posts
msg #115356
- Ignore DMUNCASTER |
9/13/2013 1:34:47 PM
Kevin_in_GA:
i also have been getting into StratySearch. Aren't you concerned that running all combinations of variables and arriving at something that works gangbusters for a particular period, is just a form of data mining? Doesn't it do a great job of predicting the past with minimal relevance to the future?
|
DMUNCASTER
34 posts
msg #115357
- Ignore DMUNCASTER |
9/13/2013 1:38:28 PM
Kevin_in_GA:
i also have been getting into StratySearch. Aren't you concerned that running all combinations of variables and arriving at something that works gangbusters for a particular period, is just a form of data mining? Doesn't it do a great job of predicting the past with minimal relevance to the future?
|
DMUNCASTER
34 posts
msg #115358
- Ignore DMUNCASTER |
9/13/2013 1:39:19 PM
OOPs. I might have sent this twice.
|
Kevin_in_GA
4,599 posts
msg #115359
- Ignore Kevin_in_GA |
9/13/2013 1:46:02 PM
Possibly, but several of the systems I have developed using SS are still doing quite well years after the system was optimized.
There is a risk of overfitting on any system, especially one that can do brute force calculations as fast as SS can. The key (for me at least) is two keep the complexity of the system as low as possible, and develop it from basic principles that inherently make sense. The Z-score system and the "BUY THE DIP" threads here at SF are examples of what I mean. Both started from a single basic premise, and were tweaked minimally to get to the versions I posted here.
|