04/23/2012 Live Update

[MrMiyagi]

Seems that Europe's bottomed as well...

[Mr. BachNut]

I need a help:

I designed a system, the back test shows, the bigger the stop loss, the higher the winning rate and for the 10K invested, the bigger stop loss even gets better performance, so it seems the bigger stop loss is not a problem here. However the stop loss reads ridiculous to me, if you think 6*ATR(10) is acceptable, how about 20*ATR(10)? Anyone has better idea to check besides the winning rate, the performance, anything else I can add to judge which parameter is the best to use? I really don't want to use a system with 20*ATR(10) as stop loss.

Hard to say without knowing what drives the system. Can a stop be engineered that gets you out when the system is probably wrong but keeps you in when it is probably right?
If the system is intrinsically volatile, a way to deal with it is to keep the big stop loss but adjust trade size so that drawdowns within the stop are tolerable. So, rather than fix the size at $10K, fix the maximum tolerable loss and back calculate into the trade size.

In reality trading, I always use fixed risks. You can see how I use it in my CIS. Say, the risk is $500, the stop loss is 3.1*ATR(10), so the position is always 500 / ( 3.1 * ATR(10) ), in this way, the loss is guaranteed to be around $500, no more surprise.

But in system design, to compare easily, apparently I cannot use fixed risk, because no matter what, a 2*ATR(10) system wins more than 6*ATR(10) system because 2*ATR(10) system would invest more money each trade, so as long as it's a winning system, the 2*ATR(10) system most likely earns more money.

So I'd like use a fixed 10K, to say which one performs better. If a 10K investment on 6*ATR(10) wins more than 2*ATR(10), then apparently, the 6*ATR(10) system is better.

Correct me, if I'm wrong. That's the purpose I'm asking here. If I'd know anything, I won't ask here, will I?

Got it. Yeah. Look into the CALMAR ratio (google ought to have something) and some of the related type calculations. You want something that measures the return you get for the risk you take as not all 15% returns are created equal.

[Cobra]

also the question is how to measure the performance.

The simplest way is you invest 10K, if at the end you get 6K, so your performance is 60%, right? But the question is, your 60% in 5 years or in 1 year, so time should be the factor, correct? Therefore in order to compare the performance, I need convert all the return to yearly based, then add them together, correct? The highest return one it the best performance one, correct?

So in this case, you got 60% in 5 years, but I got 15% in 1 year, 1 year performance 15% vs 60% / 5 = 12%, so apparently the 15% 1 year is better, correct?

You might want to look into the CALMAR ratio. It is basically a compounded rate of return over three years (usually based on monthly performance) divided by the maximum drawdown experienced by the system during the three year period. So, a system that generates a 15% annual rate of return but suffers a 5% drawdown somewhere along the way has a calmar of 3 and is superior to a system that generates a 15% annual rate of return but suffers a 15% drawdown. This sort of calc may also help you gauge the benefit of widening stop losses to get more return as it is a direct return/risk performance measure.

I'd consider this tip. but to me, the max draw down always is the stop loss, so maybe I should use my total annualized return divide by the stop loss?

[uempel]

Taggard, I like your reasoning.

[BullBear52x]

A day like today we should have another low. DB or Pos D to end the day, shock and awe, ambush!

[KENA]

Could be a small C&H forming.So lets watch.

[Petsamo]

UNG should rally now. Why? Because I sold my 40 shares of UNG.

[Cobra]

I need a help:

I designed a system, the back test shows, the bigger the stop loss, the higher the winning rate and for the 10K invested, the bigger stop loss even gets better performance, so it seems the bigger stop loss is not a problem here. However the stop loss reads ridiculous to me, if you think 6*ATR(10) is acceptable, how about 20*ATR(10)? Anyone has better idea to check besides the winning rate, the performance, anything else I can add to judge which parameter is the best to use? I really don't want to use a system with 20*ATR(10) as stop loss.

Hard to say without knowing what drives the system. Can a stop be engineered that gets you out when the system is probably wrong but keeps you in when it is probably right?
If the system is intrinsically volatile, a way to deal with it is to keep the big stop loss but adjust trade size so that drawdowns within the stop are tolerable. So, rather than fix the size at $10K, fix the maximum tolerable loss and back calculate into the trade size.

In reality trading, I always use fixed risks. You can see how I use it in my CIS. Say, the risk is $500, the stop loss is 3.1*ATR(10), so the position is always 500 / ( 3.1 * ATR(10) ), in this way, the loss is guaranteed to be around $500, no more surprise.

But in system design, to compare easily, apparently I cannot use fixed risk, because no matter what, a 2*ATR(10) system wins more than 6*ATR(10) system because 2*ATR(10) system would invest more money each trade, so as long as it's a winning system, the 2*ATR(10) system most likely earns more money.

So I'd like use a fixed 10K, to say which one performs better. If a 10K investment on 6*ATR(10) wins more than 2*ATR(10), then apparently, the 6*ATR(10) system is better.

Correct me, if I'm wrong. That's the purpose I'm asking here. If I'd know anything, I won't ask here, will I?

Got it. Yeah. Look into the CALMAR ratio (google ought to have something) and some of the related type calculations. You want something that measures the return you get for the risk you take as not all 15% returns are created equal.

google-ed it here: http://www.investopedia.com/terms/c/cal ... z1ssTYeQyr

looks easy to implement. I just need a total annualized return divide the max loss I suffered. Not exactly the CALMAR, but got the essence, didn't I?

[gappy]

Seems that Europe's bottomed as well...

Usually we approximate the damage there, there can be no true decoupling. They trade alot of US issues.

[Cobra]

still looks like a bear flag or pennant. the trend line support is here, let's see. NORMALLY, we should see more down, my target is still possible.

[TraderGirl]

Looks like there's potential for support in this area... Today and Wednesday are turn dates, so could be up into Wednesday and then down

[btran874]

If we close green today, that would setup bullish hammer (daily) on all indices. Of course, fat chance that will happen.

[Cobra]

thank you guys, remember, this is your stimulus package to me. Just click the link once a day if you cannot follow and vote (I'll always put the link here to make it easy to click, so really it's not very difficult job to do): http://stockcharts.com/public/1684859/tenpp

[Cobra]

let's see if EMA20 can put a dent on bulls.

[MrMiyagi]

Cobra,
I pride myself on being smart enough to open a bottle of Sake but for the life of me I can't see where to vote for you.

[Mr. BachNut]

Still short from last week. Retest of April 10 low has arrived. As this could be the start of a second push in a correction, I'll try to hold it. It is a shot at Cobra's 6.5% scenario, though uncertain.
The April 17 high was never really threatened late last week, especially on a closing basis, which is bear friendly. The flag broke to the downside, which makes sense.
Volume doesn't seem that heavy to me today though. So, no rush for the exits either. Perhaps just a stop run.
With Apple earnings and FOMC coming up, the week's trade is yet to be determined. So, just going to wait and see for now.

[Cobra]

Cobra,
I pride myself on being smart enough to open a bottle of Sake but for the life of me I can't see where to vote for you.

Only stockcharts members can vote. You can help clicking once a day.

[Cobra]

If we close green today, that would setup bullish hammer (daily) on all indices. Of course, fat chance that will happen.

I actually saw quite a few comments arguing for a green day today. How many of you guys now think we're to close in green today?

[MrMiyagi]

Cobra,
I pride myself on being smart enough to open a bottle of Sake but for the life of me I can't see where to vote for you.

Only stockcharts members can vote. You can help clicking once a day.

OK, I see, not a member 'cause I'm too cheap.

[Cobra]

just want non-members to know, that I added google as my featured stock report, you can always find the free chart here: http://stockcharts.com/public/1684859/c ... /265018963;