DMUNCASTER
34 posts
msg #115776
- Ignore DMUNCASTER |
10/9/2013 11:39:16 AM
oldsmar52
Just reread yor posts and see that you are using 2% and not the 1% that TRO set up his code for. So at 80% you could get 100% in a year.
But if you get in at the open, you get more like 50% reliability. No offense intended, I just couldn't see why you'd be "peeved" by my post.
|
Kevin_in_GA
4,599 posts
msg #115779
- Ignore Kevin_in_GA |
10/9/2013 1:12:58 PM
Just to inject some basic mathematics into this discussion, let's assume that you do the following:
1. Take 5 trades per week (could be more over time but for now we'll just use five). All trades are the same size.
2. You make your 2% target 80 percent of the time (4/5).
3. Your losing trade costs you on average 5% (could be less, but this is a relatively stiff penalty and these stocks are relatively volatile).
4. For the moment, slippage and commissions are ignored.
You would end up making (2+2+2+2-5)/5 = 0.6% per week. Using compounding that works out to be (1.006)^52 = 36.5% per year.
If you assume that the losing trade only costs you 3% instead of 5% you net out at 1% per week or 67.7% per annum. Factoring in commissions and slippage you might get a 50% realized gain - still exceptionally good for most traders.
If you are trading 10 "units" and 9 of them hit the 2% target and the losing trade is still 3% your numbers get a lot better - (2+2+2+2+2+2+2+2+2-3)/10 = 1.5% per week, or with compounding 116% per annum before commissions and slippage.
What I see in this system as most attractive is that you are highly likely to have a winning week, almost every week. The exact annualized returns aside, very few systems can do that, and as an investor it is hard to find fault in a system that makes you money every week with little or no time spent trying to time the market.
|
mahkoh
1,065 posts
msg #115783
- Ignore mahkoh
modified |
10/9/2013 2:56:41 PM
What I meant with DUST is the following statistic
2% gain from open 48 times out of the last 52 weeks
2.5% gain 47 times
3% gain 45 times
3.5 % gain 44 times
4 % gain 42 times
What I'm trying to point out is that you could aim a bit higher without substantially diminishing your odds.
|
Kevin_in_GA
4,599 posts
msg #115785
- Ignore Kevin_in_GA |
10/9/2013 3:21:25 PM
Not sure that I would trade the improvement in upside for the hits you would take - in the misses that you would suffer through some of the drops are 15% or more. Ouch.
I think Frank has the right idea.
BTW, here is my "streamlined" version of TRO's filter. Just cut out some of the distractions and kept the focus on the 1% and 2% gains. I also raised the price to above 5 to absorb some of the bid/ask spread issues that might hurt your returns (IMHO).
|
mahkoh
1,065 posts
msg #115787
- Ignore mahkoh |
10/9/2013 4:36:12 PM
Surely you'll increase your chances of taking a loss by aiming higher, but if you make 1 % extra the other 45 weeks you can afford 3 hits of -15% (without compounding). What it comes down to is finding the right reward/risk ratio. But I understand that Frank does not exit unless the target is reached so in his case the risk is a 20% capital loss in the case of bankruptcy.
The larger risk is having a part of the capital locked up for a longer period, and if you're long a leveraged ETF it could mean trouble, especially for inversed ones.
But as Frank mentioned they usually close out profitable within a couple of weeks.
|
DMUNCASTER
34 posts
msg #115788
- Ignore DMUNCASTER |
10/9/2013 4:43:26 PM
No argument with you Kevin. You the man. However, I'm not sure where this 80% figure comes from. Summing up the 6 weeks I've now looked at, it's nearer 50%. Also, I've always had a problem with getting in at the open. I do a lot of swing trading and I've found that getting in once the price goes your way a bit, gives better results (high above previous close, or previous high). You stay out of a lot of trades where you get in at the open and it heads rapidly south. Is there a reason to always get in at the open other that it's easier to backtest? Any way to backtest Crock Pot in SF or SS?
|
Kevin_in_GA
4,599 posts
msg #115789
- Ignore Kevin_in_GA |
10/9/2013 8:45:21 PM
I'm trying to figure out the right code for SS. Using SF you need to do it manually.
A quick manual back-test since 8/2 using only the top 5 from the filter I posted (only difference is close above 5 instead of 1) shows the following:
Number of trades: 50 (ten weeks, 5 trades per week)
Number of wins: 45 (90% win rate on the 2% limit)
The 5 losing trades were not small, at -10.7% (EVC week of 8/5), -10.2% (CLNT week of 8/26), -2.55% (GMCR week of 9/2), -1.9% (EVC week of 9/16) and -10.8% (GMCR week of 9/23)
You took a few large hits, that functionally erased all of the other profits those weeks, but you basically crushed the SPY over the same period with little or no effort.
|
duke56468
683 posts
msg #115790
- Ignore duke56468 |
10/9/2013 9:26:19 PM
What would be a reasonable way to exit the bad trades?
|
Kevin_in_GA
4,599 posts
msg #115791
- Ignore Kevin_in_GA |
10/9/2013 10:10:06 PM
I would need to confirm this, but it looks like most of the successful trades were completed by Wednesday. Don't know if that helps or not.
|
dknoonan
27 posts
msg #115810
- Ignore dknoonan |
10/10/2013 11:12:47 AM
One way to backtest this semi-automatically would be dump the backtest data to Excel, then add a column for day of the week. Insert a new column "C", and give it a formula of =WEEKDAY(B2,2), and fill down column C. You'd want a trigger date of Friday I think, so filter out any rows that aren't a 5 in column C.
|