Ed Mamula.com

Over the past couple of weeks, I’ve received some suggestions on how to decide when to turn a system off and on. 

Here are the options that I explored:

Equity Curve Timing :  The idea here is to track a system’s cumulative profits and losses and to turn the system off when the cumulative profit moves below some trailing moving average of cumulative profit.  Since I only run two systems, I can’t say whether or not this is useful in general.  I can only say that for my systems, this technique does not seem helpful.  The same issues with using moving averages on a price chart show up on an equity curve chart as well.  Namely, moving averages are a lagging indicator.  Recognizing the fact that I don’t use moving averages in my trading systems, but I do use MACD, I could choose to get fancy and chart a MACD of my equity curve, but this probably won’t work well either.  My goal is to keep the system on most of the time and turn it off when there’s reasonable doubt about the fact that it still works at all.  Using an equity curve timing method seems unreasonable for that purpose.

Control Charts :   I tried the methods detailed at Vertical Solutions:

“As an example, my automated systems turn themselves off if both the ratio of expected wins to actual wins and the twenty trade rolling P&L move 3 standard deviations below their average.
”

This method sounds great, and it probably is.  However, this method told me to keep trading both Euro Ranger and Cable Glider.  So what’s wrong with that?  Nothing…other than the fact that I WANTED to turn Cable Glider off and I was searching for a method that would support that decision.  Of course, in retrospect, I realize how silly that approach is…think of a trader who is stuck in a losing trade and keeps switching indicators until he finds one that tells him to stay in the trade…yeah, bad idea.

At the beginning of this week, I settled on a simpler criteria.  If the rolling cumulative 20 trade profit of either of my systems is negative, I will shut them off.  I chose this criteria because of my biased view that I needed to shut Cable Glider down.

In fact, in a truly out of character moment, I decided that since I’m shutting Cable Glider off, I will double the amount that I risk in Euro Ranger….well, since I’ve really had the anti-Midas touch for all of 2008, it’s not hard to guess what happened this week…

Euro Ranger immediately suffered its worst single trade pip loss EVER, and Cable Glider (off) had a fabulous week, netting something in the neighborhood of 250 pips. 

So this week, my equity is at a new low for 2008, and now I’m facing the decision about whether or not to turn Cable Glider back ON…well now, it’s always something…

Well, I’m back from my self-imposed two week break.  Within 48 hours of turning the systems back on, I’m down 3%, mostly on the back of two losers in the Cable Glider system.  It’s time for me to face the fact that Cable Glider has been a consistent loser all year, and it needs to switched off.  I am doing that immediately.  I will continue to monitor it in paper-trading mode, and if it shows two months of consecutive profitability, I will re-consider using it.

My plans had no provision for System Death.  I plan on becoming more studied on this subject, and I will incorporate it into future system deployments.    There is some good information over at Vertical Solutions, which I hope will point me in the right direction.

In the meantime, I will keep Euro Ranger active.  It was virtually break-even last month, but it has shown several consecutive months of profitability. 

The linear regression system that I was testing on the Japanese Yen totally fell apart in out of sample testing.  It’s a little bit disappointing, but then again, that’s why we test strategies before deploying them.

I’ll also be tinkering with tick charts a bit…this is something that I’ve never considered before, but I believe that it makes good sense to use tick charts instead of time based bar charts in a 24 hour market where most of the ticks are clustered into a small time frame.  For example, if we use a 500 tick bar on the Euro FX contract, we’ll see that sometimes a single bar can span a few hours (usually right after the session opens), and as little as 1 minute in a fast market.  Very interesting indeed…and that’s what I need right now…new ways of looking at the data.  We’ll see what it yields.