Magic of ADX

This system undoubtedly trades on margin. Am I mistaken?
Also, the risk adjusted rate of return is greater than the acutal return...hmm. Which tool does amibroker use to measure "risk adjusted return"?

Correct me if I am mistaken. If you trade using a margin of 20%, the untilized margin is 80% and hence, the exposure is 100% - 80% = 20%, right? So less the margin one uses, more inflated the RAR is?
By what means is this RAR?

Raja,
Seems you are still upset with me.However i am just floored with this system;
http://www.traderji.com/97981-post68.html
As mkt will open within 30 mins ,could not go thru your post in detail,there is a positive note with a Gr8 system.
Asish

cutting out your blah, let me tell you that you have missed the point AGAIN, as usual. the system shows that ADX works! and u can improve it vastly from here.

1. It is not about beating. any system that you post on PAST data can be beaten with a few lines of code, day in and day out with new so called theoretical systems. THAT DOES NOT MEAN A THING.

2. secondly for non traders, let me tell you that when you use the same system and your max DD (say 300000) reaches the second time instead of 80th time as in your simulated test, your a/c equity will shrink and u will be peeing in your pants.

3. without knowing (i am sure u will write that u knew these things since the day of your birth in the next post, but unfortunately they are missing in your superbrained posts, and talks only about returns) that systems comparison are a different ballgame and only fools will compare systems only on net return basis. hence because of your stupid premise of beating system only on return and not knowing how to compare them with appropriate parameters no exp person will post in this thread.

if u r sooooooooooo good at systems writing, then u must know that each instrument has its own characteristics and you have to use the tools accoringly on them, not write a few lines, put it in backtest and jump around with joy.

nobody is having a systems competition here. your claim that i have posted an improved system that you can beat anyday with ANY OTHER system is moronic and fufills your desire to stem your inferiority complex. u said adx is futile and gave a donchian test. i used the same adx on the similar instruments that u have used. i never claimed that i am an experienced and GREAT system developer as YOU CLAIM. But u dont know how to COMPARE SYSTEMS, and imagine a crappy developer hiding behind the name "beginner" WRITES 3 LINES OF CODE JUST TO SEE IF YOUR CLAIMS ARE CORRECT AND SEE THE RESULTS! there is nothing more to say.

as for winning/losing, there is no point in wasting time to win in beating theoretical (sometimes flawed) stuff. Go and win in the markets.

so stay in your ivory tower and be happy! my job is done. u said some stuff. i showed you otherwise. end of matter.

the same ADX based system where ADX is rising and above a certain level. (how did the returns change? anbody would know)

i know u will run and find some other system with greater returns and post it next. but since this thread is going nowhere, i wont be posting anymore systems. tho i'll be following the thread in case some other valid points come up.

Attached Images

ADX ALLINSTR.png (22.0 KB, 13 views)

Ah, thanks for the explanation. Had confused exposure with margin.

This question has often confused me . A few weeks back, I was working on a system, which surprisingly was based on 1 assumption EMH too. Excellent backtest reports but practical applicability tends to 0, which to my shock had not been disclosed in system performance report. Backtest of "The Holy Grail" (reasons disclosed),as I'd like to call it :

very very relevant question. backtesting like this doesn't mean anything.

1. backtest is on past data. and markets change every minute, everyday, every week. so similar results wont be possible.

2. as i said earlier, the results of the back test may not be replicated. suppose u have a very good backtested system that u now intend to use in real life. if instead of the max DD of say 10% after the 80th trade, where your system is already up 80%, you get two or three consecutive near max DDs at the beginning itself, the results change completely. here MC comes in, but there are lot of arguments against MC also.

3. MM also plays a huge role. in auto backtesting, u use a defined exit and loss point. this may not be the case in real life when u r actually using the system. research has shown that a bigger SL at the beginning when the trade is initiated increases the result even while u tighten it as market moves in your favor. this kind of thing is possible if u r writing your custom backtests instead of using the automated one that is supplied. for that it is extremely useful to learn the language of ur software.

4. many times unknowingly u use functions that look into the future. in real life thats not possible, but in backtest when u are 120 bars behind and the system actually has the future data stored, they do look and screw up results. this is possible in AMi, but not in Meoticker and TS as they use bar based system with no possbility to look into the future. many unscrupulous system sellers intentionally use these features to jack up results and claim how great their systems are.

5. any times the backtest results may be inappropriate e.g a system nay hold for 1200 bars to generate 80% annual. but in real life trading would u hold it and sit for 4 years and yet call yourself a trader? here u have just not attached profit target in the backtest. but ask urself. in real life if u get 200% absolute profit in one month, will you not exit?

6. many times you do not fill parameters like risk free interest rate, margin deposit etc correctly, which affects results. u dont look at entry and exit. your choice of whether to exit after or before the bar is completed affects results. so does stuff like entering on the same bar with 0 delay, which is practically impossible in real life. and if u r trading in india where u cant trade automatically thru integrated platforms, ur set up, emotions etc will prevent u to enter with 0 delay, that th ePC can enter instantly while backtesting.

hence traders like joe ross etc dont believe in backtesting.

so whats the way out?

What is happening to-day ,being floored 2nd time ,(which Ajay always does with me) let the mkt close to concentrate on these.This is Ajay's failed result.

quite a few! but none are perfect. i'll present each argument briefly

1. take that backtesting is the only way to see whether ur system was effective on past results. there is no other way. many argue that if ur system is not effective on past results, it wont be in the future. YES, this is true to some extent, though i dont subscribe to this fully, cos as i said markets change continuously. u can optimize to get good past results which will bomb in the future if ur system cant adapt to fut conditions. hence we see so many self adaptive "intelligent" systems today. (none are perfect enough)

2. having said that u have to realize that not only backtesting, u have have to forward test as well. here in sample and out of smpla and walk forward testing comes in.
A. u can divide ur existing data into two parts, test on one, correct and optisize and then test with this on the other part to see if the results are accurate. remember that it has its pitfalls. if u divide the data from 1994 to 2001 and 2001 to present, the results for a long only system will be terrific in your out of sample data (2nd lot) as the market was in a strong upward move. there's a solutio for this also in selecting your in sample and out of sample data.
B. u can use MC or generate random data with STDNORMINV within a certain range and then try out your walk forward to see the results.

u can also wait for a few cycles by paper trading it on actual markets (i would suggest trading 1 lot or small no of shares actually i the mkt).

and finally when u r back testing, remember that a system is a puzzle where all components have to come together. when u r actually using a system, for MM u have to put stuff like win/loss ratio, MFA/MFE etc which can be accurately obtained from real trading records that u have, not some ballpark figure that you can put.

finally i agree with what you have said, particularly about optimization. u should not over optimize, but should know when mkt conditions are changing, and then apply accordingly. fib smoothed stoc, fractal MA smoothed adx/macd etc can adapt quite well to the mkts. neither have i optimized, so lets not sit and optimize and dig out best results from the past which may not have any bearing in actual future trading.