Ubzen System Development Process - page 3
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Now that the Time-2-Me checks out. I'm going to experiment with some Exit conditions which might improve the Emfe. The 1st thing which comes to mind is Trailing-Stop. Yep, it's time for that Break-Even stop to make way for something more dynamic. Then we'll also try BarrowBoy's Atr-Stoploss found here. Yep, BB I'm dragging you into this :) hope you don't mind. For those who don't know who, he's one of our Moderator. And Last but not least Zzuegg's Accelerated Ma found here. 8P Oh, and I taught about 2 more. Since the stop-logic started with previous 5-bar low why not keep that trend going. And, one of my own, Envelopes :) hope you like the mail man.
To preserve some of the Psychological easing and original logic, I'm going to trigger the trailing-stops after it set the break-even.
Wonder how you would use the Accelerated MA as stoploss indicator. For exit strategies i can recommend:
if CCi(14) > 300 exit all long's, of course cci<-300 exit all shorts.
Regarding trailing, i like trailing the last low/high, but that might be personal. I am also continuing my research. Not yet happy with the EMFE on lossing trades.
My responses seems too long for others to read out. Thats because I was thinking out loud. I'm going to try to keep them short moving forward.
@ Zzuegg: cool then, I'll use your recommendation of cci instead.
@ Phillip: Yes... I understood approach in your diagrams all along. However, my undermine is different from yours. Explained how I arrived at that value below. Using a scale of 100-Pips from mae-to-mfe below. I choose 100 because it relates percentages. I might be better off Averaging instead of Summing. And converting to % instead of Dividing but that's something for future implementations.
Added: Lol, and this was suppose to be short. I think it's worth noting that I tried but could not find a single trade in this system which toke the path of your diagram. (at least not a good one). It either goes down to stop-loss and close all orders. Or it goes to Profit and Set's Stop-loss at break-even. I find this interesting indeed.
Mfe Undermined by 37 pips. (Known)
Meaning wait until Ideal Entry - 37p later. Thats gaining 37p.
What I didn't know - and Attempted to create is Comparing this information with other systems.The example (left) stand best as 1-single trade.
However, in my attempt to standardize it treated it as a basket of trades. My ( Mfe - Mae = Undermine ) is mis-leading or mis-named as your Undermined is the value of the Mae itself.
My Undermine is something totally different and I was aware of that all along. Maybe I just wanted another matrix. Taking ideal entry & exits. Your undermine is 0-(or spreads)<--this value i already have from Mae, and my undermine is 100-(or 100-spreads).
That relationship makes sense now but gets fuzzy as values leave ideal entry-exits.
At this juncture, I'd like to point out a few things. Also, I'll like to thank everyone for helping so far.
1) The sample size for this is very small.
2) There going to be walk forward analysis.
3) I'd be very surprised @ similar results in walk-f.
4) No optimizer was... nor would be used for anything.
5) I have a Ton to gather b4 the walk-forward.
6) Zzuegg pointed out this period as sweet-spot for Sys.
I tried different trailing-stop-values well only 20, 30, 50, 100. I also tried the Hi-Lowest bar as Ts. (lets just say something interesting happened during that test).
And the winner is BB's Daily Atr. Was achieved using the ATR.Percent 100 as Trailing stop. Profit factor=9.26, Maximal Drawdown=350.07 (3.13%)....Draw back is during Ranges, But boy it doesn't hold any punches in Trending periods.
Next was my Envelops Profit factor=6.47, Maximal Drawdown=310.07 (2.74%)....More balanced but closes earlier than BB's during trend on the smallest consolidation or retracement.
Followed by Zzuegg's Cci. 3rd but definitely not least. It's not a trailing stop rather I used as he instructed. (more than likely I goofed up somewhere ). It's perfect for Topping Ranges. During trends however, the CCI goes to 300 not long after the order is opened. So, that's more of a logical issue which I'll need to resolve.
Below is the Default vs BB's Atr .... Equity Curves.
Next I'll get the M?E Scoring for BB and Envelope.
Default: in_Pips
Mfe 2700 - Profit 977= Excess Mfe 1723
Mae -1272 + Mfe 2700= Undermined 1428
Zen-Mxe 1.20
BB_Trail-Stop:
Mfe 3768 - Profit 1637= Excess Mfe 2131
Mae -1028 + Mfe 3768= Undermined 2740
Zen-Mxe 0.77
This is the first time the Zen-Mxe have gone below 1. My guess is that the Envelope results would be somewhere in the same ball park. This will mark the conclusion of Mae-Mfe. Next returning back to the article, I'm going to take on Standard Deviation, Variance and the Bell-Shape Curve. This would be important for answering questions like what's my chances of being 50% down within X number of trades with this system. And whats my Expectation within 1,2,3,4 (yeah 4 can happen) standard deviations if I run this system for the next 3 months to follow. The variance is used to answer ?s on Kelly Criterion aka Optimum F. IMO money management should begin and end with Kelly.
After doing some thinking, I've decided to use (Net Profit vs Mae) instead of Phillips RAROC for my Standard Deviation analysis. Reason is simply because of Ease. RAROC is a little subjective IMO as I'll have to define Risk-Free Assets, and I might re-evaluate it when I deal with Sharpe Ratio. I'm gonna calculate this using a method I somewhat familiar with, the one used in Bj. Using 1637(Profit) should be > than 1028(Mae) should be conservative enough. I'll use this as Benchmark for Systems : Net Profit > Mathabs(Mae). The data is below:
54
0
45
0
38
0
42
0
71
42
29
191
32
130
56
200
21
0
57
0
67
0
54
-1
-95
-96
55
0
56
0
79
0
44
0
28
0
37
41
27
333
-2
-3
-12
-11
-21
-22
-21
-20
-107
-106
-3
-2
-3
-2
-14
-15
-10
-9
-106
-105
-3
-2
-9
-10
-96
-97
-38
-37
-14
-13
-5
-6
-14
-13
-19
-18
-14
-13
-6
-7
From the Data I get the following information using Excel. Standard Dev: 65.5071, Variance: 4291.18, Main: 7.6125:
I created a Bell-Shaped curve using instructions here. Pretty cool as this is the first time I've made one :). Somehow I got 2 other lines and I'm not sure what they mean. However the Peak lands on a positive value thats a Good thing. That value is usually our Expectation. I'm used to seeing this picture with the positives on the Left hand side instead. Or maybe I'm thinking about the Kelly Curve. Anyways anyone please correct me if I'm wrong about my positive assessment.
Next, I believe I'll have to split the mountain into half for Expectation. And then those half into half for 1-Sd(66%), and those half into half for 2-Sd(95%)....etc. I'm going to base most of my forward test to be within 3-Sd or 99% confidence of the Expected value. If it goes to 4-Sd then we have a problem. This helps explains the probability of me being X-amount down within X-number of trades.
Ok, I'm just guessing here, but I think my Expectation per Trade is around 40-pips. Why 40, well because that's around where that green and red line crosses. When I multiply 40 time #of-trades 40, I get 1600 which is pretty close to the profit. One negative Sd is -65.5071. Meaning allot of the times when I lose, it's going to be around 66 pips.
Now I cannot remember if 2Sd-means Sd*2 or Sd*Square but I'll go research and be-right-back. Also, after that, I'm going to calculate the Kelly based on Sd and then accept 1/2 the value it returns because of Std-Errors. Well, I cannot find where I'd seen what Two Sd means in relationship to the One Sd. But I'll assume its 1-Sd*2 because 1-Sd squared would not fit on our X-axis there.
So then within our forward test, we'll expect to see negative trades up to 3-sd. Thats about -195 Pips. Anything above 200 Pips and it'll probably be back to the drawing board. Or I could just blame it on the small sample size :P. I think it'll be worth reminding my self here that Draw-downs increases infinitely the longer a system runs. Meaning there's no limit to how low things can go. That said, I can expect a loss over 200 Pips-Per-Trade, question is when and how often.
So how much should I bet ... I'm sorry Risk? ;) Lets have a little chat with my good-ole-friend Kelly. I'm going to use a simple formula for this one. k={ p(R+1) - 1} / R. Where R= Profit-to-Loss-Ratio, and p= Chance of Winning Trade. So k={ 0.75(1.6+1) - 1} / 1.6, k=59.38%. What!! 59%, I think Mr Kelly had 1 too many. No way I even go with half of that as intended. This highlights the problem with using small samples. But because I trust Mr Kelly, I'll go with 1/4 his recommendation. 14.75% of the equity would be risked per trade. Still high but this system's results is unrealistically good and that's why the high numbers. After looking at a year's result the numbers should look allot more realistic.
To help amplify how unrealistic these results are, below, I used one of my old Bj Sims to generate a Win-Loss Probability for 3-Months using the Sd# generated. Pretty much, it's saying within over 95% Confidence I should be somewhere between a profit of 532-->2668 Pips with the Mean=1600. This assuming I get 50 trades per Quarter. So it comes at no surprise that Kelly is going OC.
With these kinds of predications, I'm going to be running a Walk-Forward Analysis next. First with fixed 0.1 Lots then with Kelly. Of course if the next 3-Months fails, we're not even going to bother with Kelly. Instead, I'm going to include the data as part of the Statistical Analysis. I'll keep doing this until the system does Not Fail with the Forward/Backward Test Periods. By Fail, I'll define this as being outside the Lowest Expectation of the Quarter. In this case it's 532 Pips. The Rating would end up being Optimized Periods vs Forward/Backwards periods which were not Optimized. Here's an old article I ran into when I took up Forex / MetaTrader. The article has changed a bit but still same idea. I'll be trying to define something called walk forward efficiency ratio Next.
Well, It Passed for 4-2010-->6-2010. However it didn't do as well Profit factor wise. So, I'm surprised to see the Profit so closed to expectation. The bad news is now we have new Statistical Data to average with the old.
Because I don't have the time to post the pictures now. But I'll definitely tell you that I've ran the forward tests through December and it's passed. However July-September was the worse period and it barely passed that. Just looking at the curve I could see about 5 losing trades in a Row. If you betting high on that Kelly, this is definitely give back period. However, like I stated, new data means re-adjustment for the system. I'll hold off until I have performance for the maximum# of years Data I can get my hand on before programming the Money-Management into this. And then It's Demo-Testing to Penny Testing time ;)
Sakis the
funtametalist as usben call me i consider my self more like a techniciani derived from th 80's with the generetion of turtle even though i was not one of them.
The times that traders like Richard Denis, Tom Basso,Ivan Boesky(his life filmet by Oliver stone
in his 1987 film the wall street) making an epoch with they secret trading systems that was not
more than MA cross over,s or Doncian channels breack outs.
I am a trend folower..
The system expant by Usben is quite well i am imprest
i get 2 questions
1: the first system that you pose saki_0.2 it is 50,100 MACD?
2 question the 2nd system with MACD 50,100 sma and CCI with nett profits of 317690
why you dont like it? if the reason is the drwndown 43,50% you are wrong and let me explain why
Start look thinks not like a programmer but like a bussinessman you start the 04-01-2010
with 10000$ from your home you get a speculation by investing this ammount of money or you
can loose the 1/2 of it or you get the chance to grow your capital 30 times
do you take the risc or no??
and the answer is you will get the risc if you get more money at you account than the 10.000 so lets say if you get 10 times the 10000=100000 and then you can loose the 5000 so the REAL RISK IS 5% so by riscing 5% of your capital you can growit by 300%
is worth taking risk in isolated account!
Now technicaly the RSI i think it makes a great job but you can use another timer as well
when the entry happent you smoothen the price's with 3LWMA what we call the driver and then you put another MA (let say 30, 50 KAMA(against the voladility) or SMA)) lets call it driver
so!! when timer cross driver exit and thats it!!!
now by closing 1/2 of the position at breack even it protect your profits if you rise the SL at break even but you get less winers
you can do it another way close 1/2 of the position when you reach the ammount of sl and dont move the sl at breack even by this way you minimise your losses thats good as well
Can i have the code for the EA to test it as well?
you can contact me at any time at tranco@inbox.com for more questions and more systems that i curently use to make a living!!
thanks in advance
Sakis
Sure Sakis, I'll email you the codes I generate. The 1st system on page one was done by me. But tested by Zzuegg... you can download and test that EA from the location where you give the system link Here. That system has Ma 50 & 100 with MACD at Default. However the Second system MA50100+MACD is NOT done by me. Rather its done by Zzuegg. I believe he's using CCI Exits and Personal Filters and Opening new position like Sell even if he's got a Buy open. This is something my System is Not doing.
It's not that we don't like the systems or results. It's more that we're trying to make it better. I think anyone would accept any of those results, thats why we toke on this project in the first place. I like your Idea's on Timer and Drivers. I'll see what I can do about RSI, 3LWMA and KAMA. If you download and test the program, please let me know if it's Not doing what you intended. You can Private Message (PM) by clicking my name and select Write a private message. I'll email or pm you if I have questions.
Ps> I need a-lot of time to polish your system programming. I'll email you the EA when I complete it. What I did earlier is a rough draft for back-tester only. It need lots of stuff to work on live accounts.
Sure Sakis, I'll email you the codes I generate. The 1st system on page one was done by me. But tested by Zzuegg... you can download and test that EA from the location where you give the system link Here. That system has Ma 50 & 100 with MACD at Default. However the Second system MA50100+MACD is NOT done by me. Rather its done by Zzuegg. I believe he's using CCI Exits and Personal Filters and Opening new position like Sell even if he's got a Buy open. This is something my System is Not doing.
It's not that we don't like the systems or results. It's more that we're trying to make it better. I think anyone would accept any of those results, thats why we toke on this project in the first place. I like your Idea's on Timer and Drivers. I'll see what I can do about RSI, 3LWMA and KAMA. If you download and test the program, please let me know if it's Not doing what you intended. You can Private Message (PM) by clicking my name and select Write a private message. I'll email or pm you if I have questions.
Ps> I need a-lot of time to polish your system programming. I'll email you the EA when I complete it. What I did earlier is a rough draft for back-tester only. It need lots of stuff to work on live accounts.
i am interresting for both system it must sell when the market change direction
If you make a simple cross over between 3WLMA and 100SMA 30 min chart EUR/USD tp 150pips and exit when cross over happent not SL you will end up with extreme
profittable system which i use it for somme time look the results and write to me back with 13% drawndowm