Amibroker Backtest presents it as
Recovery Factor (=Total Gains Made/Max DD)
Happy
As we can have different profit points for different years, i was suggesting average profit points/max DD (max DD is the biggest DD across the timeframe and average points is the 'average'
)
If Amibroker is giving this measure, great. If not, we can do it in couple of seconds. Am not sure if recovery factor is same as what i was suggesting (as it mentions the word 'total' instead of 'average')
@madank yes, this sounds good. ratio will definitely be better since we reset the returns computation for every year. also addresses the issue of random distribution of winners and losers.
MG - we have gone too much into the technicalities of the numbers but i want you guys to see the reasoning behind that pointer. Lower max DD can enable us to compound in an efficient way as the system will recover faster (in that sense, i like the 'recovery factor' term of Amibroker). Let's say we have another system with
average profit points for 10 years as 5000 NF points and
max DD as 1200 points. I would trade the 1200 points NF with 200 maxDD over this 'super-duper' 5000 points system on any given day and can guarantee that i can turn a small account to a decent sized one with 1200/200 system.
Basically, I want the readers of this thread to start thinking in maxDD perspective not from the profit points viewpoint. There is no real magnificence in profit points and it's time to focus on drawdown. Mind you - 1200 points per year is just 100 points per month in NF. Am sure some folks are vying to make 150 points per week in NF
@madank i remember seeing a plan/list of topics you intend to cover but cant find it now. would you also cover something on the lines of - at what stage one must stop optimizing the backtests?
for instance --> let's say our system is a simple 20 MA crossover based. next, we use 'X' MA to filter long and short entries. then we slap another indicator on the chart......when does one stop and move on to work on other factors such as position sizing?
Here's the post.
http://www.traderji.com/community/threads/general-trading-chat.96368/page-6348#post-1260741
Even though am not a big fan of optimizing even a basic system, there are certain things a trader can do to come out as a marginal winner in optimization.
If one has couple of indicators in the system, optimizing can eliminate useless rules and parameters. Optimization is also useful to determine whether certain kind of stops are useful. For example, a trader can determine if a protective stop can add or deter a system performance. Often, the distance of trailing stops is too close to the last price, causing premature exits. These determinations can be analyzed more closely with optimization.
But the biggest bane of optimization is that we can custom-make any indicator to match the past data perfectly. Especially, after the advent of software that can do this BT task in couple of minutes. Because the future is what we are trying to control, most optimization is useless and dangerous as it gives a false sense of confidence.
So, it is up to the trader’s penchant for the optimization and its results. In my opinion, one has to stop optimizing at some point as the negatives override the positives of optimization by a huge margin. If one gets stuck in optimization (mostly in the search of holygrail), it is very difficult to come out of it. Long story short - there is no silver bullet solution for this problem. Once we start making money in our system and realize that we can positively control the outcome by MM and position sizing predominatly , this search will stop (hopefully).
Another piece of suggestion I want to lay down here is that to see if the system works on multiple instruments. It should work reasonably well across basket of instruments (with varying returns/risk) but if it makes losses in other instruments but works only in NF, then most probably, it has been curve fitted for Nifty.
Madan,
Can you just point different ways of learning minute details of market?
I know
1) Chart.
2) backtest ( Checking logic results)
Backtest helped me to see somethings which I couldnt notice from chart. Charts help me to make some kind of logic for backtesting. Tweaking logic (curve fitting) many consider harmful but it has revealed few things for me. Any other things you guys use?
Lemondew,
When it comes to chart reading, am a bigtime novice/amateur. I have just figured out a way or two to control my basic pivot system with MM and position sizing..thats all. As am a lazy person, I stick with the same system for the past several years and have not ventured into anything new.
Well, lot of screen time can help a discretionary trader to learn about market behaviour but wont help a mechanical trader that much. As a matter of fact, we dont need to understand market dynamics that much to make money in trading. Being a mechanical trader, one can devise new systems to start trading once his/her account grows bigger than what the existing single system can handle.
Curve fitting can be detrimental to our account for various reasons mentioned above but please refer to my answer for MG. There is no end to this curve-fitting saga. It is a mild form of searching for holygrail. It is like a frivolous man keeping on looking for the best partner and because of his proclivity to get the best (compared to the current partner), he will keep looking until he dies – somewhere the partner search has to stop. Am glad that curve-fitting has taught you certain things but sooner or later, it will pull you deep to keep optimising the system.So, we gotta take that tough call (of stopping optimisation) sometime.
I think there is a confusion here, the ratio suggested by madan are to measure the systems max DD: Total Return for each selected time period, i.e if 1 considers 1 yr as measurement period then the system Max DD: Total return should be 1:7 i.e 50:350/100:700/200:1400 etc to be considered as a good system to ensure compounding happens faster.
Yes ncube – my interest lies with
average profit points and max DD for any given timeframe. Ofcourse, total profit points is needed to calculate the average profit points for a given time frame