jackmack
334 posts
msg #113261
- Ignore jackmack |
5/10/2013 9:23:06 AM
Kevin - good morning
I was wondering if you wouldn't mind running this filter set in SS as I do not have and SF only seems to pick up SSO and blows out back test data.
Your current set up is (5,1,5) (5,1,10) but I was wanting to see how (5,1,5) (5,1,14) did in comparison on a side by side back test as you have listed on the first page.
Very sorry to have to ask for this.
Thank you in advance if you complete - if not I also understand.
|
Kevin_in_GA
4,599 posts
msg #113274
- Ignore Kevin_in_GA |
5/10/2013 11:23:42 AM
You know that this wins only 53% of the time, right? And that the divergence filters (Pangolin D on Collective2) are much more reliable?
I'll run the comparison this weekend, but would still advise you to look at consistent win rates like those seen with the divergence filters.
|
jackmack
334 posts
msg #113280
- Ignore jackmack |
5/10/2013 12:47:25 PM
Kevin
I agree COMPLETELY and I am in NO way challenging the win rate - I am merely curious as I cannot test the data here in SF.
Thank you for doing this.
|
Kevin_in_GA
4,599 posts
msg #113281
- Ignore Kevin_in_GA |
5/10/2013 1:30:54 PM
Ok, just making sure!
|
jimmyjazz
102 posts
msg #113313
- Ignore jimmyjazz |
5/11/2013 4:43:57 PM
Kevin, I'm not sure what the adjective is -- maybe "robustness"? -- but how do you assess the effect of a slight shift in win rate for these methods that are very close to 50/50? Perhaps it's an issue with all systems, but if I go from 53/47 to 52/48, that seems like it could lead to a HUGE drop in returns.
|
fortyfour
189 posts
msg #113319
- Ignore fortyfour
modified |
5/11/2013 8:25:03 PM
Jimmy jazz ,
Simple... The best thing you can do here is to learn the basics of the stockfetcher backtester
and Excel. Download the TRADES from the SF backtester into Excel and learn to do these
calculations in Excel.
Example is.....if avg win is $60 and avg loss is $40 and system is 50/50% and there are 100 trades totally
then you have a gross of $3000 won , gross $2000 lost for a net profit of $1000. Now if 51/49% and all else stays the
same the gross profit is $3060 and gross loss is $1960 for a net profit of $1100.
If 1000 trades total then net profit would change from $10,000 to $11,000 if system change from 50/50 to 51/49
I asked Kevin something elsewhere but the bottom line is I can do all the work myself too. I agree it's just easier to
Ask hen you are watching an NBA double header. Good luck. Gotta love iPads.
Ps: barring any simple arithmetic errors I learned something here as I wouldn't give a 50/50 compared to
51/49 system too much thought in my testing but damn a 10% profit difference is indeed significant.
The simple assumption of a 50/50 system with a 60/40 avg win/ avg loss is not extreme but maybe 55/45 better.
Definitely a lot of value getting dirty with simple data iin Excel.
|
Kevin_in_GA
4,599 posts
msg #113320
- Ignore Kevin_in_GA |
5/11/2013 9:10:58 PM
From a previous post I did a while back - relevant here:
What is important is not win % but the expectancy of the filter, meaning the average profit per trade one can anticipate based on using this approach versus others.
An example to help clarify this -
We have two systems that we can use, one which results in a winning trade 40% of the time and one which results in a winning trade 80% of the time. If this is all you know about each system, then the obvious choice for people is to go with the system that wins more often.
But now we learn that when it wins, the 40% system returns 15% on every winning trade, and loses 3% on every losing trade. In contrast, the 80% system returns 3% on every winning trade but loses 15% on each losing trade.
Now which system would you go with? Intuitively, you might still stick with the 80% system, but let's now look at the numbers more closely:
$10000 x (0.8*0.03) - $10000 x (0.2*0.15) = 240 - 300 = -$60 per trade on average
And let's look at the 40% system as well:
$10000 x (0.4*.15) - $10000 x (0.6*.03) = 600 - 180 = +$420 per trade on average
Are things becoming a little clearer? At this point it becomes very obvious that you are probably going to lose money on the 80% winning system with every trade you place.
I'd usually now talk about frequency of trades as the next key point (obviously with two profitable systems you need to look at how often one has the option to make money and multiply the average profit per trade by the frequency of trades), but since one is negative expectancy there is no real need.
Money management is different from the above - that usually is about trade sizing, stops, etc. Expectancy is what one needs to understand before placing ANY trade.
|
jimmyjazz
102 posts
msg #113321
- Ignore jimmyjazz |
5/11/2013 11:31:46 PM
All these points are valid, but they're not what I was (trying to) ask. Of course I understand the arithmetic combinations involved. It's why I used the 53/47 VS 52/48 example. It's a 33% reduction in profit, all things being equal.
What I'm trying to get at is how does one assess the likelihood of retaining that slight edge in a system that depends on that edge to put money in the bank?
Maybe it all comes out in the wash. I'll think my way through it. I'm just concerned that a system so close to (or below) 50/50 would be more sensitive to market conditions than one which goes 80/20, but obviously, if market conditions cause that 80/20 system to suddenly become less profitable (per win), then we have a similar problem on our hands.
|
bkhurana43
103 posts
msg #113324
- Ignore bkhurana43
modified |
5/12/2013 9:17:34 AM
Using SPY, SSD & SSO charts with RSI Divergence Filter, it appears Entry and Exit points are in reverse order. Am I missing something?
|
ashokar5
4 posts
msg #113350
- Ignore ashokar5 |
5/13/2013 4:15:27 PM
hi
can you make a system for buy and sell for macd and klinger volume oscillator.
thanks
asok
|