Big R multiple wins and their impact on expectancy

[pavilion103]

As many of you would know I've been creating/testing a discretionary system for the past 6 months or so. As well as running a live simulator I am also testing historical data throughout my days (obviously I understand there was a strong bull market from 2003-2008 and conditions then were much better than they are now).

I am trading a hypothetical $20,000 account with each trade risk of 1% ($200). Most of my losses are between -0.5R and -1R after trailing my stop with the average win being around 2R (this isn't over a huge number of trades yet though).

Now I completely understand that this is different for different systems but how often do people tend to get huge wins e.g. 10R+? As you can see in the screenshot below from my excel file, I had a win of 12.2R and this caught my attention because my previous biggest win was only 3.5R which I had a few of (over around 80 trades).

Is it these types of trades that really increase the expecatancy? and even often the difference between a profitable and non-profitable system?

Also what about a big loss. For example in one of my trades price gapped on the open and I lost 3R. How often do people encounter this sort of thing?

What are people's experiences/opinions?

Thanks

 

[wayneL]

IME it is the outliers (and a system capable of riding them) that make a trend trading system in the end.

 

[tech/a]

The prefered basis for any trading method is to cut losses and allow profits to run.
Those systems that put you in a position to be on an outlier move and I have had 50+ R returns in the 2003/8 days-- are what your after.
Your equity curve on closed trades wont look that good but will smooth out as you include open trades and of course time.

So look for systems which place you in this position.
We never know for sure when another 2003 will come along and when it does--and it will--- trading with a system which j
Has you ready and waiting will do amazingly well.
Don't forget though that an exit which allows you to ride outliers WILL give back a fair % of profit during the course of a longer term trend.
Don't expect to pick up those short term small caps which + 100 % in a few days with the sort of system I'm suggesting.
Another topic all together.

Cheers from Monte Carlo.

 

[TabJockey]

I always have a few 10R trades in my backtests but the important thing to look at is how long it took to get 10R. I am primarily trying to get 2.5R in about 10 days. So if I see a trade that achieved 10R but took 100 days that is actually under performance, with that same capital the system could have made an additional 9 trades at 2.5R = 22.5 R.

This is why I now implement a nbar stop in allot of my prospective systems which improves performance over time even if it drags average win down a tini bit.

 

[pavilion103]

I always have a few 10R trades in my backtests but the important thing to look at is how long it took to get 10R. I am primarily trying to get 2.5R in about 10 days. So if I see a trade that achieved 10R but took 100 days that is actually under performance, with that same capital the system could have made an additional 9 trades at 2.5R = 22.5 R.

This is why I now implement a nbar stop in allot of my prospective systems which improves performance over time even if it drags average win down a tini bit.

This is a very good point. I was thinking about this today. It may not be wise to tie up capital in those situations for 150+ days if I am making gains of 3-4R in a month.

 

[IFocus]

This is a very good point. I was thinking about this today. It may not be wise to tie up capital in those situations for 150+ days if I am making gains of 3-4R in a month.

The market determines your return. This market currently is difficult but there are still outliers ILU case in point (I missed the trade) I have been saved more than once by such outlier moves and generally when looking for bigger moves it involves a certain amount of draw down.

 

[sinner]

As many of you would know I've been creating/testing a discretionary system for the past 6 months or so. As well as running a live simulator I am also testing historical data throughout my days (obviously I understand there was a strong bull market from 2003-2008 and conditions then were much better than they are now).

I am trading a hypothetical $20,000 account with each trade risk of 1% ($200). Most of my losses are between -0.5R and -1R after trailing my stop with the average win being around 2R (this isn't over a huge number of trades yet though).

Now I completely understand that this is different for different systems but how often do people tend to get huge wins e.g. 10R+? As you can see in the screenshot below from my excel file, I had a win of 12.2R and this caught my attention because my previous biggest win was only 3.5R which I had a few of (over around 80 trades).

Is it these types of trades that really increase the expecatancy? and even often the difference between a profitable and non-profitable system?

Also what about a big loss. For example in one of my trades price gapped on the open and I lost 3R. How often do people encounter this sort of thing?

What are people's experiences/opinions?

Thanks

View attachment 43232

The question is, what sort of system is it?

If you are trading a system that needs outliers to work i.e. it doesn't need to be right very often because statistical loss is so much smaller than statistical gain, then outliers will be what you need to be profitable. In this case the outliers are the winners.

If you are trading a system that relies on being right most of the time usually this means statistical losses are larger than gains (usually, because otherwise this would be the only way worth trading), these systems hate outliers. In this case the outliers are the losers.

wayneL is probably (as usual) right. Trend following systems are where you need those big outliers to really bring home the returns (even or especially if you pyramid IMHO).

Think about it in terms of breakouts.

You can test it. More than half the time, breakouts fail. In some markets, significantly more than half the time.

If your system relies on outliers to be profitable, you are betting on the lesser probability being realised (breakout) and your reward is commensurately larger (in the less likely event you are in fact rewarded).

If your system relies on being right more often than not, you are betting on the more regular probability being realised (breakout failure) and your reward is commensurately smaller - but often more consistent.

In the first case your risks are usually easily to define and manage. In the second case the risks can often be subject to "outlier volatility", where the unexpected happens (massive breakout) and the person on the other side of your trade gets paid extra for betting on the lesser probability being realised.

 

[Wysiwyg]

What is 'expectancy'? Van Tharp has this ...

By definition, expectancy is how much you can expect to make, on the average, over many trades. Expectancy is best stated in terms of how much you can make per dollar that you risk.

But what is blatantly never mentioned is how 'many' is "many trades" or how long is 'time' in the oft mentioned "over time"? Back testing or walk forward back testing is only a rough guide in my opinion. What was before, may or may not be similar this time around. You will hit some winners and you will hit some losers but that won't be known till afterwords. Knowledge/experience will cut them short or let them run.

Eyes and ears work 100% better when tuned to a 'positive expectancy'. For example as Tech/A did, "sell in May and go away".

 

[pavilion103]

I've got a question that may be a little off topic. I was thinking about the implications of these large R-multiple wins increasing expectancy and this got me thinking about commissions and fees decreasing the expectancy.

Presently the simulator I am trading is a $20,000 hypothetical account. I am risking 1% per trade ($200). If I use IB cost per completed trade will be $12. That represents 6% of the $200. Is this far too much? Is it almost impossible for me to make money this way?

Would it be wiser when I actually go live to begin with a $25,000 account risking 2% per trade ($500). Cost per trade will then be 2.4%. This seems much much more viable.

What are people's thoughts?

 

[TabJockey]

I've got a question that may be a little off topic. I was thinking about the implications of these large R-multiple wins increasing expectancy and this got me thinking about commissions and fees decreasing the expectancy.

Presently the simulator I am trading is a $20,000 hypothetical account. I am risking 1% per trade ($200). If I use IB cost per completed trade will be $12. That represents 6% of the $200. Is this far too much? Is it almost impossible for me to make money this way?

Would it be wiser when I actually go live to begin with a $25,000 account risking 2% per trade ($500). Cost per trade will then be 2.4%. This seems much much more viable.

What are people's thoughts?

I think that if your account size is 25k, the position size has to increase as a percentage of your total. 2.4% commission is absolutely jinormous and if you come up with a system that is clearly profitable with such commission allowance its either curve fitted or bloody amazing.

I actually simulate at a similar size level because that is what I have to work with, and usually I have a position size of 10% and a risk of 1.6% of total account. The commission at this level is 0.1% for the brokerage and 0.1% for slippage allowance, so that eats 0.4% of my expectancy per trade. The mdd's on permutations of this system range between 4% and 18% which are acceptable to me.

A smaller position size may reduce risk on paper but at that level it decreases the chance of coming up with a system that's going to work in real markets.

 

[skc]

The idea of Monte Carlo analsis is to take out the impact of big Rs. Your system return will be a bell curve when you run a Monte Carlo analysis. The high returns on the far right of the curve (which may be 120%) are boosted by big R wins, while those on the left (which may be -10%) are impacted by big R losses.

It will be up to you to determine what is acceptable. E.g. You may only trade systems that have a 66% probability of producing 20%+ return per year.

One thing for certain is that your actually is uncertain, but will fall somewhere between the highs and the lows. Over many trades however it will revert to the mean...

I always have a few 10R trades in my backtests but the important thing to look at is how long it took to get 10R. I am primarily trying to get 2.5R in about 10 days. So if I see a trade that achieved 10R but took 100 days that is actually under performance, with that same capital the system could have made an additional 9 trades at 2.5R = 22.5 R.

This is not how one estimates the opportunity cost. Your opportunity cost is your expectancy, not the next 9 trades all being winners and conveniently lined up for you at the end of every 10-day period.

If say you win 40% @ 2.5R and lose 60% @ 1R (Expectancy 0.4R), and say you average 60 trades in a year holding on average 3 positions at a time... you can roughly say that your opportunity cost for holding 100 days instead of 10 days is ~2R (90/360 * 60 /3 * 0.4R). So if holding 100 days get you additional 7.5R your capital is working very efficiently.

I think that if your account size is 25k, the position size has to increase as a percentage of your total. 2.4% commission is absolutely jinormous and if you come up with a system that is clearly profitable with such commission allowance its either curve fitted or bloody amazing.

It's not 2.4% commision. It's 2.4% of the risk per trade. $12 commission on $200 risk is 0.06R. If your edge is so thin that it can't absorb 0.06R then it may not be a valid system. With $200 risk per trade, the position size would probably be $2-5K. So the effective commission is ~0.3-0.4% round trip.

Yes it eats into your profits (60 trades @ 0.06R is 3.6R) but it doesn't have to be unviable. In my pairs trading my commission is >20% of my net profit Now that is bloody amazing

 

[pavilion103]

I'm looking over the statistics for my most recent testing.

This is a very minor question, but when I'm calculating the accuracy, do breakeven trades count as % of winners or % of losses.

So if I have, say, 24 trades and I have 11 winners, 11 losers and 2 breakeven, does my system have
11/24 = 46% accuracy
Is that how it is determined?

 

[peter2]

I remember asking that question many many years ago. You can include them as losers or discard them from your results giving you a W% of 11/22 (50%). I've noticed that people/organisations who promote their results do not include the BE results as it reduces their W% and public appeal.

I count BE results as losers as my money was exposed to risk in the market but didn't produce any profits. This reduces my W% quite significantly as I defend my capital aggressively and have many BE results (~20%). The reduced W% looks bad and my ego is dented. The advantage however is that each BE result reduces my average loss. Seeing this makes me feel much better. I'm comfortable including BE results as losers as it curtails my ego and keeps me focused on improving my AW/AL.