Turning on and off a strategy ...

M

mizhael

Suppose you are running multiple strategies, some of the strategies are dropping like a rock these days... do you shut down these strategies? And if you shut them down, when do you turn them back on? (of course, all these mechanisms need to be backtested)

Here is what I came up with, any thoughts?

The stop logic:

The input is a data stream of strategy cumulative pnl; and the output is another data stream, of binary 0-1 numbers.

The output is essentially a switch which switches on and off.

Initially the strategy is on.

1) Any time when the strategy cumulative pnl number is below 75% from the running max of the cumulative pnl, the strategy is shut down and it turns into paper-trading. And we start to keep track of the running min of the paper-trading cumulative pnl.

This is not the running min of the overall cumulative pnl, but is the running min of the cumulative pnl stream starting from the very point where the strategy is shut down.

2) Any time when the paper-trading cumulative pnl is above 125% of running min from that paper-trading, the strategy is turned back on, and now we start to keep track
of the running max from that point and we go back to step 1).

Does this sound like a good idea to you?

Any thoughts are appreciated... thank you!
 
Quote from abattia:

See...

http://www.elitetrader.com/vb/showthread.php?threadid=198742&perpage=6&pagenumber=1

... and the comments about joekrut diff.
More...

Thanks for bringing that back.

After these days, I have some new thoughts regarding your reply to that thread.

I think:

1. Shutting down a strategy is in no conflicts with analyzing the problem of the strategy and improve it.

2. Using 30 day SMA crossover to turn off and turn on a strategy might not be a good idea.

Think about it: what's the use of the MA? It's for smoothing out noise, and it's lagging. It's good to apply this to a raw price series, and use it as a stop loss (stop out and then back in) control, because raw price series are mostly like random walks.

But when you apply it to a strategy equity curve, most strategy equity curves don't move like random walks. They should be smooth enough, especially if they are the portfolio level equity curves (i.e. each strategy runs on a whole portfolio of securities).

Therefore, you don't need SMA in this case. You should try my method - percentage off from the peak as well as percentage up from the bottom.
 
Quote from outsource:

Look at any SMA EMA crossover historical data. It`s almost no losing trade at all:D
More...

What do you mean?

Could you please elaborate? Are you saying that you found SMA/EMA crossover to be a good trend-following strategy?

-------------------------

My post above was addressing stop-out-and-back-in. He was complaining that his SMA doesn't really work in controlling stop-out-and-back-in for his strategy. I was suggesting maybe instead of using SMA, percentage-off-from-peak-and-up-from-bottom is a better control method.
 
No hope.
I did some experiments.
Lets suppose your strategy is buy-and-hold a trend-following strategy.

And then now we want to smooth the equity curve, using the SMA30 as abattia has mentioned in another thread.

I compared the buy-and-hold strategy vs. the SMA30 turning the buy-and-hold on and off strategy.

The SMA30 based strategy always underperforms the simple buy-and-hold strategy, both in terms of Sharpe ratio and Sortino ratio.

This is weird... and disappointing, because it means that even though I currently am experiencing large drawdown, I cannot shutdown my strategy because turning strategy off and back on degrades both Sharpe ratio and Sortino ratio.
 
Quote from mizhael:

No hope.
I did some experiments.
Lets suppose your strategy is buy-and-hold a trend-following strategy.

And then now we want to smooth the equity curve, using the SMA30 as abattia has mentioned in another thread.

I compared the buy-and-hold strategy vs. the SMA30 turning the buy-and-hold on and off strategy.

The SMA30 based strategy always underperforms the simple buy-and-hold strategy, both in terms of Sharpe ratio and Sortino ratio.

This is weird... and disappointing, because it means that even though I currently am experiencing large drawdown, I cannot shutdown my strategy because turning strategy off and back on degrades both Sharpe ratio and Sortino ratio.
More...

Even more weird is the following observations:

In the buy-and-hold strategy vs. the SMA30 turning the buy-and-hold on and off strategy, the underlying is the equity curve of a trend-following system.

Even though the SMA30 turning the buy-and-hold on and off strategy has a worse 15-yr Sharpe ratio as well as a worse 15-yr Sortino ratio than the buy-and-hold strategy, it has a very positive 2010 year-to-date PNL..., where is the buy-and-hold strategy has a very negative 2010 year-to-date PNL...

Any thoughts?
 
Quote from mizhael:

Thanks for bringing that back.

After these days, I have some new thoughts regarding your reply to that thread.

I think:

1. Shutting down a strategy is in no conflicts with analyzing the problem of the strategy and improve it.

2. Using 30 day SMA crossover to turn off and turn on a strategy might not be a good idea.

Think about it: what's the use of the MA? It's for smoothing out noise, and it's lagging. It's good to apply this to a raw price series, and use it as a stop loss (stop out and then back in) control, because raw price series are mostly like random walks.

But when you apply it to a strategy equity curve, most strategy equity curves don't move like random walks. They should be smooth enough, especially if they are the portfolio level equity curves (i.e. each strategy runs on a whole portfolio of securities).

Therefore, you don't need SMA in this case. You should try my method - percentage off from the peak as well as percentage up from the bottom.
More...

No one makes money using that technical analysis garbage
 
Quote from outsource:

you are completely on the wrong track.i don`t understand you.are you defending a thesis or you wanna day trading?
More...

Why am I on the wrong tack completely?
 
You must log in or register to reply here.
Back
Top