I use range bars aset at 4. SL is 6 ticks, If SL is hit it's $120 loss.
I know what you will respond. You are risking $120 loss to make $150 profit. Why not buy 1 contract, at 5 tick profit buy 2nd contract and sell both at 10 tick profit. It's the some 15 tick profit, however it is reducing potential loss by half. Your point is very well taken and very valid.
I have been doing it my way since the beginning, and it wasn't until I wrote out "your respones" that I see a way to reduce potential loss. That's the great thing about trading, ask ten people how they do it and you'll get ten different answers and none of them wrong. Howeve we can all learn from one another.
Keep things as simple as possible, but no simplier. Albert Einstein
I have been reading this thread with great interest. As a purely systems trader, not looking for psychological comforts in each trade, but looking for over all 'edge" in a series of trades, I have been gravitating towards all in and all out approach, using the best possible risk to reward strategies over time.
However, that being said, I ran into this article by Dean Hoffman:
(I am not able to include the link as I am a new member of this Forum but you can Google "Improving Performance Results by Dean Hoffman" and this will take you directly to the pdf article.)
which finally seems to give a possible argument for scaling out (partial exits) based purely on reducing drawdown and increasing the Net Profit to Drawdown ratio. I find this possibility fascinating and wonder if any one has tested their system with this variation as shown. (As you can see no discussion on psychology or comfort levels, but just stats)
Would love to hear people's actual systems experience on this.
The following 2 users say Thank You to Systems trader for this post:
I'm a system trader myself, and I strongly believe that scaling In or Out is WRONG!
My argument is this:
You can split the scaling trade into two trades. One with a small profit and one with a bigger profit and a break even if you like. Now back test each trade. Without knowing the specific setup, I know that one of the trades will be much better than the other, so why do both?
In another thread I explained that the biggest profit you get by combining several strategies or instruments or time frames, but you need to have non correlated or negative correlated outcomes. Scaling out/in will be very strongly correlated.
The following 4 users say Thank You to baruchs for this post:
(1) Dean Hoffman presents a trend following system applied to a portfolio with the following key figures:
Net profit $ 673,307, Max. Drawdown $ 60,504, Winning percentage 36.1%, Avg $Win to Avg$Loss 4.44
(2) Then he suggests trading two contracts instead of one and comes up with the key figures
Net profit $ 1143,307, Max. Drawdown $ 69,015, Winning percentage 38.0%, Avg $Win to Avg$Loss 3.75
Are the figures correct?
At first glance, the figures are consistent. Winning percentage and Avg $Win to Avg $Loss and profit per contract are similar for both systems. The striking point is the reduced drawdown. And of course this is possible as well, so I think that the figures are correct and consistent.
Does the second system catch superior returns without increasing risk?
What is not obvious is that the max. drawdown is a proxy for risk-of-ruin and that the superior return of the second system is risk-free and you are therefore entitled to trade 2 contracts instead of one without increasing the risk.
2 contracts of one instrument traded
Let us imagine that we trade 2 contracts of a single instrument. In this case the drawdown is the result of a series of n consecutive trades with a combined loss, which is the drawdown. Now, if you scale out 1 contract by using a first profit target, some of these losses might have been reduced to less than half the size (as we have already scaled out half the position with a small winner and the loss is reduced to 50%), others would be double size (as they did not reach the scale out target).
Let us assume - for simplicity reasons - thath the max. drawdown was made up of 4 consecutive losers, out of which 2 never reached the first profit target and 2 others did. In this case the No-Scale-Out-System trading 2 contracts would have made a loss of 4*2R = 8R, while the Scale-Out-System would have made a loss of 2*2R + 2*R - 2*PT1 = 6R - 2PT1. If you further assume that the scale-out profit target sits at +R, the drawdown of the second system becomes indeed 4R, which is half of the drawdown of the first system.
So the dispersion of the results of the second system is indeed smaller. The standard deviation of the returns from the expectancy is smaller, and the Sharpe Ratio is better. Both allows for trading larger size.
Add the Portfolio Effect
If you only trade 1 instrument and increase size, the results are 100% correlated with themselves. The portfolio effect -trading different instruments that are less than 100% correlated - further reduces risk. This again allows to increase size.
I do not know whether the presented system trades only 1 contract at the time or whether it allows to open several positions over different instrument with a low correlation of returns. But I believe that the drawdown should be further reduced by the portfolio effect, if several positions are held at the same time.
The improved risk profile is due to the effect that negative and positive trades can cancel out
- if you trade 2 different system with a sufficiently low correlation
- if you trade a portfolio of different instruments
The catch with the described system is that it does have a low winning percentage and will therefore likely have a series of 6, 7 or 8 consecutive losers. So the scaling-out will have a positive impact on risk even without the portfolio effect. This would not necessarily be the case for a scalping system with a high winning percentage.
But system with a higher winning percentage have lower drawdowns anyhow, so that you can increase the size without scaling out.
Yes, I agree with the author.
The following 4 users say Thank You to Fat Tails for this post:
The question here is indeed correlation. The performance maybe greatly enhanced, if not correlated. I think I showed with my previous post, that scaling-out / drawdown can be led back to the following two cases
(1) prior to drawdown 1st profit target is reached -> 100% negative correlation, drawdown is reduced
(2) prior to drawdown 1st proft target is not reached -> 100% positive correlation, drawdown is identical
I think that the positive effect on risk cannot be derived from a single trade, but a series of trades. It is the series correlation that changes. The correlation is further reduced, if a portfolio of instruments is traded.
But your conclusion is correct. Rather than scaling out, you can find an entirely different strategy, which is less correlated with the original one than the scale out strategy and this will improve the Sharpe Ratio even more!
But in case you are trading a portfolio, the scaling out is already good enough, to reduce your risk by 50%, if Dean Hoffmann's calculations are correct.
The following user says Thank You to Fat Tails for this post:
Hi Fat Tales,
This is not how I look @ correlation.
Lets examine a scaling out trade in which after taking the first target you move your stop to break even (just because this is what most people do, although its wrong too)
In this trade we can have 3 outcomes:
1. We were stopped out - correlation = 1
2. We reached the high target - correlation = 1
3. We reached first target, moved stop to break even and were stopped out - correlation = 0.
So the cumulative correlation is close to one.
But I don't measure the correlation this way, because I don't do scaling. I measure the correlation of daily PnL of different strategies / TF / instruments. Thats what is important.
So if in theory you have two strategies with correlation = -1, that means that on each day if you are in profit in strat#1 you lose in strat#2 and visa versa.
In this situation even if one of the strategies is a losing strategy you are much better off by trading it together with the winning one. But if the strategies are strongly correlated, as scaling is, then don't trade the losing strategy!
The following 2 users say Thank You to baruchs for this post:
You made a slightly different assumption, with the adjustment to breakeven, which is probably a better idea.
But in my opinion you missed the principal point, which is the correlation not between strategy one (original) and strategy two (taking profits early) but
- the correlation between different consecutive trades, leading to the interruption of a chain of losers
- the correlation on the portfolio level, when the 2 strategies are applied to various instruments at the same time
Both effects contribute to reducing the dispersion (standard deviation of the results from the mean expectancy) and therefore reduce the risk of a major drawdown and the risk of ruin.
The following user says Thank You to Fat Tails for this post:
I havent got hard statistical data to back this up but based on my experience and learning, my opinion is that all in / scale out can work strategically for advanced traders but mostly it is used by learning traders to ease the short term psycholgical stress of being in a trade. In this case I would argue that it is a counter productive strategy because a learner most likely wont have above 1:1 risk/reward on a set of trades. A couple of full stops will wipe out several small wins. I think if a trader has achieved a very high win % to the first profit target level ( say around 80% or better) then a scale out approach can work quite well by taking lots off at predefined SR levels and catching the odd big runner...but for those trading a lower win % method ( like 50%) I doubt it is of any use except to feel good intratrade while you watch your account balance fall.
The following 3 users say Thank You to Linds for this post: