Viable trading system statistics

[Roller_1]

Curious to know when you guys build a system, what are the stats you look for ie. returns vs drawdowns etc. What are the minimums and maximums that you would look for to make it a viable system?

 

[VSntchr]

Re: viable trading system stats

Curious to know when you guys build a system, what are the stats you look for ie. returns vs drawdowns etc. What are the minimums and maximums that you would look for to make it a viable system?

Everyone will have a different answer but some of the obvious ones:
expectancy, max win, max loss, avg win, avg loss, win %, avg trade length, max/min excursions.

Then there are things like:
number of signals/trades, correlation with other strategies, capital efficiency etc.

 

[Roller_1]

Re: viable trading system stats

Everyone will have a different answer but some of the obvious ones:
expectancy, max win, max loss, avg win, avg loss, win %, avg trade length, max/min excursions.

Then there are things like:
number of signals/trades, correlation with other strategies, capital efficiency etc.

what would be the values though for yourself or your general opinion. ie. for a trend following system, win rate greater than 50%, <20% DD etc

 

[minwa]

Re: viable trading system stats

what would be the values though for yourself or your general opinion. ie. for a trend following system, win rate greater than 50%, <20% DD etc

If it's a breakout system around 30% winrate, R:R 4+. Mean reversion 40-80% R:R 1.5+. Draw down can be tailored to personal taste based on position sizing. Personally I don't like 15%+ DD but many can tolerate upto 30%.

 

[ThirtysixD]

Curious to know when you guys build a system, what are the stats you look for ie. returns vs drawdowns etc. What are the minimums and maximums that you would look for to make it a viable system?

Through a bull, bear and sideways market I want:
Annual returns > 15%
Drawdown < 50%
Sharpe > 1.5

Ill also remove the effect of large winners (not losers) but im really not that nit picky about statistics as long as the system is profitable and makes sense (not data mined).

 

[sinner]

Curious to know when you guys build a system, what are the stats you look for ie. returns vs drawdowns etc. What are the minimums and maximums that you would look for to make it a viable system?

For most systems I think it's irrelevant because the returns will largely be a result of the market regimes in which you test it.

You might think your trend following system only has a 30% maxDD but then you go and test it against the Nikkei 225 and find out it's 40. Or against the Great Depression and suddenly it's 50.

So what is the point of evaluating stuff in that perspective? There isn't one, at least my ..

What I try to test for is how well the system captures a particular market regime. I don't need numbers to know that trend following systems are gonna eat it during choppy range or whatever.

But I do need to know, does the Close > 200SMA capture trends better than buying the 200 day high? Or, whatever. In most cases, a simple Sharpe ratio of returns divided by volatility is good enough for me to understand.

 

[howardbandy]

Greetings --

I believe traders have a near-universal goal: To have confidence that the signals generated by their trading system precede trades that provide reward adequate to compensate for the risk. The key word is confidence. The primary limitation is risk.

Every trader has a personal risk tolerance. Mine is: I am trading a $100,000 account and looking forward two years. I am willing to accept a 5% risk of a 20% drawdown, measured from highest equity to date.

Every set of trades (actual, backtest, or hypothetical) defines a distribution. Assuming the system is stationary (that the future will resemble the past), that distribution can be used to determine the risk associated with trading it for the next two years. By adjusting position size, the risk of the trading system can be brought into alignment with the trader's risk tolerance. That position size is called safe-f. At safe-f, the trading system is defined to be "risk-normalized."

The position size that normalizes risk determines the distribution of profit potential.

All alternative uses of money can be normalized in this manner. Trade the one that has the highest risk-normalized profit potential.

Best regards,
Howard

 

[Gringotts Bank]

The position size that normalizes risk determines the distribution of profit potential.

All alternative uses of money can be normalized in this manner. Trade the one that has the highest risk-normalized profit potential.

Best regards,
Howard

Howard, could you give an outline of how dynamic position sizing works and what sort of difference it makes?

Thanks.

 

[Roller_1]

You might think your trend following system only has a 30% maxDD but then you go and test it against the Nikkei 225 and find out it's 40. Or against the Great Depression and suddenly it's 50.

So what is the point of evaluating stuff in that perspective? There isn't one, at least my ...

That is why a system should be tested across a range of market conditions imo. for stocks the tech boom, gfc, bull markets etc so you have a better understanding of what could happen under certain conditions.

 

[howardbandy]

Hi GB, and all --

My latest book has details and computer code. You can read a lot of it on the book's webpage:
http://www.quantitativetechnicalanalysis.com/
And using Look Inside on the Amazon page:
http://www.amazon.com/Quantitative-...eywords=bandy+quantitative+technical+analysis

There are several videos of my recent (within the past two years) presentations that show the process in operation. Those from the ATAA, MTA, and IFTA are good places to start:
http://www.screencast.com/t/Vli0B4oJr5
https://www.youtube.com/channel/UCcPOv0K7zMQNOkaMFpou31g

Broadly, dynamic position sizing uses Bayesian Analysis (contrast with frequentist statistical analysis) of recent trading performance to estimate the health of the system.

The flowchart on the cover of the book, and on page 18 of the text (you can download that chapter), show the two systems that together make up a trading operation.

The trading system is on the left. The trading system generates buy and sell signals (but not position size) from processing the price and volume data.

The trading management system is on the right. The trading management system uses dynamic position sizing to analyze recent trades to determine how large a position to take on the next trade. The goal is to increase position size when the system is working well and decrease position size when it is working poorly. When the system is broken, the dynamic position sizing will have reduced the position size to zero before a large portion of the account has been lost.

Best,
Howard

 

[sinner]

That is why a system should be tested across a range of market conditions imo. for stocks the tech boom, gfc, bull markets etc so you have a better understanding of what could happen under certain conditions.

But can you not see that the outcome is only a result of what the market did, and not any magic on the part of any system?

Which means that the returns of any system once again boils down to ones ability to predict the next market regime, or at the very least correctly classify which regime the market is currently in, and deploy an appropriate system for that regime.

I think the idea suggested by Dr Bandy is not a bad one, assuming that the calculation is undertaken continuously as new trades completed, this will provide a momentum effect to systems, as the market regime changes some systems will become safe (and can be adopted) and some will become unsafe (and can be discarded).

 

[Gringotts Bank]

Hi GB, and all --

My latest book has details and computer code. You can read a lot of it on the book's webpage:
http://www.quantitativetechnicalanalysis.com/
And using Look Inside on the Amazon page:
http://www.amazon.com/Quantitative-...eywords=bandy+quantitative+technical+analysis

There are several videos of my recent (within the past two years) presentations that show the process in operation. Those from the ATAA, MTA, and IFTA are good places to start:
http://www.screencast.com/t/Vli0B4oJr5
https://www.youtube.com/channel/UCcPOv0K7zMQNOkaMFpou31g

Broadly, dynamic position sizing uses Bayesian Analysis (contrast with frequentist statistical analysis) of recent trading performance to estimate the health of the system.

The flowchart on the cover of the book, and on page 18 of the text (you can download that chapter), show the two systems that together make up a trading operation.

The trading system is on the left. The trading system generates buy and sell signals (but not position size) from processing the price and volume data.

The trading management system is on the right. The trading management system uses dynamic position sizing to analyze recent trades to determine how large a position to take on the next trade. The goal is to increase position size when the system is working well and decrease position size when it is working poorly. When the system is broken, the dynamic position sizing will have reduced the position size to zero before a large portion of the account has been lost.

Best,
Howard

Thanks Howard. I might need to buy it. I wonder if you'd mind posting some screenshots of how a system's equity curve changes with the addition of dynamic position sizing. Just a MA cross or whatever.

 

[tech/a]

Thanks Howard. I might need to buy it. I wonder if you'd mind posting some screenshots of how a system's equity curve changes with the addition of dynamic position sizing. Just a MA cross or whatever.

And if you wouldn't mind placing a few trades for me over the next 6 mths---that would be helpful.

 

[Gringotts Bank]

And if you wouldn't mind placing a few trades for me over the next 6 mths---that would be helpful.

... and some duck repellent. Just a quick spray around the place would clean things up nicely.

Thanks Howard.

 

[tech/a]

... and some duck repellent. Just a quick spray around the place would clean things up nicely.

Thanks Howard.

Think about it GB.

Its minimising loss and maximising profit.
Without seeing a curve, you'd expect shallower drawdowns and steeper rises.

Some systems I know of use their equity curve as a filter.
Either turning new trades off at X downturn or exiting the whole lot at Y.
Lots of variations.

Winchester put out a duck repellent.

 

[Gringotts Bank]

Think about it GB.

Its minimising loss and maximising profit.
Without seeing a curve, you'd expect shallower drawdowns and steeper rises.

Some systems I know of use their equity curve as a filter.
Either turning new trades off at X downturn or exiting the whole lot at Y.
Lots of variations.

Winchester put out a duck repellent.

I think for Howard selling a book, a 'before' and 'after' shot of the equity curve would be a nice selling point to see just how much difference it makes. It looks like implementing dynamic sizing would require one to learn Python, so I'm not going to jump in without seeing the beneifts. If it's a big difference, I'd go ahead. If it's minimal, I wouldn't.

Also, if one's existing equity curves aren't too affected by drawdown, such an approach might even be counterproductive. Those frequent little drawdowns of a smooth equity curve are followed by frequent little gains, and you wouldn't want to reduce position sizing just as it's ready to move back up. You might even want to increase the size during drawdown... but I guess this would also be a possibility of his dynamic approach.

 

[tech/a]

I think for Howard selling a book, a 'before' and 'after' shot of the equity curve would be a nice selling point to see just how much difference it makes. It looks like implementing dynamic sizing would require one to learn Python, so I'm not going to jump in without seeing the beneifts. If it's a big difference, I'd go ahead. If it's minimal, I wouldn't.

Also, if one's existing equity curves aren't too affected by drawdown, such an approach might even be counterproductive. Those frequent little drawdowns of a smooth equity curve are followed by frequent little gains, and you wouldn't want to reduce position sizing just as it's ready to move back up. You might even want to increase the size during drawdown... but I guess this would also be a possibility of his dynamic approach.

That's why you systems test.

Howards books are excellent I have them all.
Even a duck can follow them.\
I do attempt to avoid Python--'s

 

[Trembling Hand]

It looks like implementing dynamic sizing would require one to learn Python, so I'm not going to jump in without seeing the beneifts. If it's a big difference, I'd go ahead. If it's minimal, I wouldn't.

GB you are an interesting character. Seems you will go to any lengths not to do extra work or increase your general knowledge. I personally wonder how an approach would work in a field that requires self learning and a broad skill set? just my .

 

[sinner]

GB you are an interesting character. Seems you will go to any lengths not to do extra work or increase your general knowledge. I personally wonder how an approach would work in a field that requires self learning and a broad skill set? just my .

It's also wrong, can be done in any computer tool that supports basic statistics and mathematics functions, even Excel.

I use R to do all of this stuff now.

 

[Gringotts Bank]

GB you are an interesting character. Seems you will go to any lengths not to do extra work or increase your general knowledge. I personally wonder how an approach would work in a field that requires self learning and a broad skill set? just my .

Close. I will go to any lengths to avoid unecessary work. I have done far too much learning of unecessary crap in other fields outside of trading (stuff fed to me by experts, lecturers, academics) only to find out it's all bull****. The only way I found out it's bull**** is by questioning everyhting like crazy. That's my nature, but it can be exhausting. You get to see who's really on the ball.

So I'm extremely wary of time spent on things that won't be likely to yield a result. The result doesnt have to be money, it could be something else. Aside from that, life is extremely short and we'll all be dead before you know what's up, so I sure as hell am not going to learn just for the sake of it. Leave that for the academics and geeks.