mahkoh
1,065 posts
msg #113878
- Ignore mahkoh
modified |
6/3/2013 5:24:30 PM
"Just a comment - it is human nature (or at least the nature of folks here on SF) to tweak things, thinking they are improving them. I still do this, but am moving away from this idea and now try to write filters that are as simple as possible, and then leave them that way.
This filter is an example of showing how a basic concept like buying a dip can be quantified against a large set of data (the S&P 500 over 5 years). I am almost certainly cherry picking the best combination of settings by backtesting 7200 variations - probably already overfitting the data. More tweaking is not likely to improve the results. "
I for one certainly wasted quite some time in the past trying to tweak filters you posted here. Due to lessons from the past I haven't spent much time on tweaking the filter itself but I do believe the drawdowns point towards possible gains to be made from trade management. Read doubling down, although I am not a big supporter of that.
If one would decide to incorporate that in the strategy it would mean that instead of dividing your trading capital over 5 positions you would now invest 10 % in each position, possibly leaving 50 % of your capital unused.
But if you could calculate the statistical possibility for having to double down it would be possible to shift these numbers.
In my example from 5/20: Originally 282 trades from 5/18/2012 until 5/18/2013. Of these 98 experienced a drawdown of more than 4 %, which accounts for about 1 out of 3. Average gain if these were opened as 2nd positions 2.31 %, 72 profitable.
So one could allocate 14% of capital to each position and have room for two positions to double down upon. Obviously this model doesn't take into account that it is likely that if you happen to be in a bad market environment all 5 positions may go 4 % underwater.
|
Kasmodean
2 posts
msg #114121
- Ignore Kasmodean |
6/17/2013 11:43:32 AM
Kevin,
As a new trader trying to learn the ins and outs of making a good system, I tried plugging your strategy in stratasearch, and I didn't get anything like your outcome. I did get several that did ok ~10-20% return, none hit your level of performance. One thing i noticed is almost always my results took a huge hit when the S&P hit the bear market. I also notice the average length of my trades was almost always greater than 100 days.
Did you have any other exit criteria other than ROC > 2? Also did you use PROC in stratasearch?
I appreciate all you help you have given so far to this community.
Thanks.
|
Kevin_in_GA
4,599 posts
msg #114123
- Ignore Kevin_in_GA |
6/17/2013 2:33:07 PM
The stratasearch code for this is simple:
ENTRY:
proc(c,7) < -2 and
proc(c,80) > 20 and
close > mov(c,200,s)
EXIT:
proc(c,7) > 2
RANK:
proc(c,7) ascending
Nothing fancy here. Run this against the S&P 500 for the dates I showed in the first post and you should get the same answer.
|
dknoonan
27 posts
msg #115523
- Ignore dknoonan |
9/24/2013 9:24:40 PM
I believe the backtests are very promising. But whenever I try to do any kind of stop loss, the results suffer. Is it realistic to trade without setting a stop loss? Does the fact that this is only using S&P 500 stocks mitigate the risk somewhat?
|
miketranz
969 posts
msg #115527
- Ignore miketranz |
9/24/2013 10:49:07 PM
Never ever take a trade without setting a stop loss.Remember,the downside is having your account wiped out.This may sound a bit extreme,but it can happen.How about wiping out 50 percent of your account? There's got to be a limit to how much you're going to lose on any particular trade and you should know what that number is in dollar terms going in.Are you setting your stops to tight? Is the stock your trading volatile? SPY stocks are less risky to trade than tech stocks,biotech especially.Hope that helps.Mike......
|
jimmyjazz
102 posts
msg #115529
- Ignore jimmyjazz |
9/25/2013 1:46:39 AM
Why have any one trade, or any one strategy subject your account to 50% wipeout? That's just not good risk management.
Plenty of studies show stop losses reduce returns. This is not to say you shouldn't have an exit plan, but a hard stop on price might not be the best choice. Getting stopped out early sucks, even if you don't really track what happens down the road.
|
abelduarte
9 posts
msg #115530
- Ignore abelduarte |
9/25/2013 2:17:03 PM
Kevin, how do you exit this strategy if rate of change dipped below 2% because a change of trend is occurring ? This is my biggest concern with your filter, other than that those seem like amazing returns.
|
dknoonan
27 posts
msg #115534
- Ignore dknoonan |
9/26/2013 2:10:25 AM
In pondering the question of stop losses, I had planned to run backtests on this Buying the Dip filter several times, with no stop loss, then stop loss of 2%, 4%, etc. Since my concern about stop losses has to do with avoiding a really catastrophic loss, I picked the worst timeframe I could find: Dec 31 07 to Feb 27 09.
So I ran with no stop loss first, expecting a horrible result as a baseline. What do I get? My $100,000 account grows to $125,620 !! Once again I am amazed at how counterintuitive the market is at times. I can see that whatever the final wisdom is on stop losses, my current line of inquiry isn’t going to reveal it!
If anyone wants to verify my results, here are the particulars.
Filter
S&P 500
--close above open
ROC(7,1) crossed below -2
ROC(80,1) above 20
close above MA(200)
add column ROC(7,1)
sort on column 5 ascending
draw ROC(7,1) on plot ROC(80,1)
Exit Trigger: ROC(7,1) > 2
Min Days to Hold: 1
Fees Per Trade: 8.95
Dollars Per Trade: 20,000
Allow Neg Balance & Fractional Shares: Yes
On Advanced Tab, all NO. No Selection Criteria entered. Max trades per day: 3 Max Open Pos: 5
|
Kevin_in_GA
4,599 posts
msg #115535
- Ignore Kevin_in_GA |
9/26/2013 9:14:54 AM
That is mostly because of the requirement that you only take positions in stocks above their MA(200) and that have a ROC(80,1) above 20. That keeps you out of trouble during a lot of that time.
|
jimmyjazz
102 posts
msg #115539
- Ignore jimmyjazz |
9/26/2013 12:42:33 PM
dknoonan, did you intend to comment this line out?
-- close above open
|