"Scaling out" is inferior behavior

[Buy1Sell2]

Quote from thenewguy:

The profitability of scaling out solely depends on your probability of your profit targets AND the payout of those targets. - TNG

Bingo--

 

[Buy1Sell2]

Quote from thenewguy:


Now, if you were to argue that a system with a %30 probability of hitting it's targets is silly, then you have a case for not scaling out. But simply to say it is "inferior" to scale out is mathematically wrong.

- TNG

A system with only a 30 percent win ratio will either win more or lose less by not scaling out. The percentage of wins is not relevant to whether it's better to scale out or not. The math works the same irrespective of that.

 

[thenewguy]

Quote from Buy1Sell2:

A system with only a 30 percent win ratio will either win more or lose less by not scaling out. The percentage of wins is not relevant to whether it's better to scale out or not. The math works the same irrespective of that.

That is incorrect, and you don't understand EV.

- TNG

 

[Buy1Sell2]

Quote from thenewguy:

That is incorrect, and you don't understand EV.

- TNG

Not scaling out will outperform scaling out at any win percentage system except zero. At that level all trades will be stopped at the initial stop loss and the behaviors will have equal results.

 

[Buy1Sell2]

Quote from ForrestGump:

Very nicely said.

This is truly the heart of the matter. As Eckhardt points out, we all have a strong tendency to focus too much on the current trade we are in, and, anything that we do to keep the current trade from turning into a loss almost invariably acts to reduce the overall profitability of our system.

Cheers.

Thank you Forrest for the kind words. I feel certain that there are more than just a few that share our viewpoint. Good to hear from you!

 

[thenewguy]

Quote from Buy1Sell2:

Not scaling out will outperform scaling out at any win percentage system except zero. At that level all trades will be stopped at the initial stop loss and the behaviors will have equal results.

Funny, I posted mathematical proofs that this is incorrect, and you keep stating it like it's a fact.

Oh well, hard to debate when only one side is bringing facts to the table.

TNG

 

[Buy1Sell2]

Quote from thenewguy:

Funny, I posted mathematical proofs that this is incorrect, and you keep stating it like it's a fact.

Oh well, hard to debate when only one side is bringing facts to the table.

TNG

When you post mathematical proofs, please use the same parameters for both sides ie win percentage moving stops etc. --because both types of trading have access to that. Otherwise, you are comparing apples to oranges. When a proof is posted that has that involved, I will then analyze it.

 

[thenewguy]

Quote from Buy1Sell2:

When you post mathematical proofs, please use the same parameters for both sides ie win percentage moving stops etc. --because both types of trading have access to that. Otherwise, you are comparing apples to oranges. When a proof is posted that has that involved, I will then analyze it.

The example I posted has EXACTLY the same parameters on both sides.

TNG

 

[taowave]

Quote from Buy1Sell2:

New, if you are using profit targets and the probability is not high , then the wrong profit target is being used and should be lowered. Your backtesting should give you the correct profit target that hits with good probabiity. At that point, you don't scale out, you just let it run to the target. After all, you have tested and found that the target is your optimal target in terms of percentage-- If the target that hits 70 percent of the time is 6 points, then don't pull the plug at 5 points. That's all that is being said here.

Buy1Sell2,it appears you are entering into serious curvefitting territory,and I now see the ILL-Logic of your logic.Are you testing on out of sample data as well?

Of course if you backtest for the optimal profit target one should liquidate 100% of the position for the asset/time frame tested...We dont need computers to tell us that....

B1S2,I am pretty dam sure you are NOT testing on a sampling period and then walking it foward on "out of sample data"..

 

[Buy1Sell2]

Analysis from your example is as follows:

Scaling out 100 trades 2pt loss target(It can be any amount)
90 winners times 10 equals 900 points (1 contract)
1 winner times 11 equals 11 pts (1contract)
109 losers times 2 equals -218 pts (1 contract)
Total 693 pts


No scale out
90 winners times 10 equals 1800 points (2 contracts all out)
10 losers times 2 equals -40 pts (2 contracts all out)
Total 1760 pts

The example you have provided in your previous post concerning probability is simply a proof of why 10 pts is a better target than 11 pts. That is all. It has nothing to do with scaling out or not scaling out. A person would choose the 90 percent target as the place to exit all positions. As you can see from the example provided, "All out" is much more profitable.

Quote from thenewguy:

B1S2 I think you are forgetting to include probabilities in your calculations. Take a look at this:

scaling
contracts odds payout ev
1 0.9 10 9 %90 going to 10
1 0.01 11 0.11 %1 going to 11
9.11
all out
contracts odds payout ev
0 0.9 10 0 %90 going to 10
4 0.01 11 0.44 %1 going to 11
0.44


This is scaling out one contract at 10, letting the other run to 11. The second is letting the entire trade run to 11. I've used a %1 chance as an extreme measure to show the results. There is a point where your ev is higher letting the whole trade run, but it's a function of probabililty x payout.

TNG

EDIT: formatting sucks, looked good when i typed it in though...