The Burning Question: Do I Need to Re-Optimize my System?

The Burning Question: Do I Need to Re-Optimize my System?

July 11th, 2018

If you’ve ever optimized a system and then traded it live, you’re familiar with the question that never quite leaves the back of your mind:

Do I need to re-optimize?

Of course, when things are going wonderfully, this question never seems to come up. When money is pouring in, the only question that comes up is whether you’re a genius or a super genius.

(Hint: the answer is always super genius.)

But when things go poorly, or weirdly, then we’re right back where we started.

Is my system acting this way because it needs new inputs?

The first scenario to bring this doubt into play is losing. Losing always causes the worst of everything. When the system loses, obviously something is wrong. And that something needs to be fixed.

That’s where the max drawdown number comes in.

If the system has gone through a losing streak that’s greater than the max drawdown, then something bad is going on. The market might have changed or the testing wasn’t thorough enough. Either way, it’s time to stop trading live and look again at the settings.

In short, if a system has blasted through the previous max drawdown number, it’s probably time to re-optimize.

But that’s not the only time the question comes up.

Sometimes it comes up when things aren’t necessarily going badly. Sometimes it comes up when the system appears to be acting strangely.

Case in point: my USDJPY Hornet robot.

On paper, nothing really bad is happening. It had a great start to the year and since has tapered off. As I write this, it’s at about break-even for the year. Clearly, that’s nowhere near the max drawdown number.

(In addition, my Hornet took a full loss on a Sunday open when the spread widened to borderline criminal levels. Some of my Lifetime Members never got that stoploss, so they’re doing better this year than I am.)

But the USDJPY has been doing something odd for the past few months: it hasn’t been trading very much.

And when it doesn’t trade very much, it’s not going to make very much money–hence the ho-hum performance.

So, if it’s not trading as much as I’d expect, we’re back again to the magic question.

Should I re-optimize it?

The emotional answer is: Yes, re-optimize it! Re-optimize it every day of the week and twice on Tuesday until profit starts flowing again!

But we probably need to look a bit closer first.

For one, my USDJPY Hornet was trading quite often not that long ago. In November 2017 it traded about every other day and in December 2017 it did the same.

And in January 2018, it traded nearly every day.

That’s three straight months of active action, the most it had seen since 2016.

So it’s not like it’s been dormant for a long time.

In addition, there have been periods like this in the past. November through April 2017 was similar. May 2013 through September 2013 was even more similar.

And there have been many three-month periods that have seen bigger losses than have been seen recently.

Last, I don’t think that inactivity is necessarily a sign of danger for a system.

Value investors can go years and not see levels tasty enough for them to trade.

Trend followers can also go several months without seeing a single entry.

If prices aren’t dropping low enough, no one would say we need to stop being value investors or change our ways.

If prices aren’t breaking out, we wouldn’t say trend following is dead and we need to change everything.

Well, actually, people say trend following is dead all the time, but they’ve yet to be correct.

So what’s the bottom line?

If a system is racking up a massive drawdown, it’s time to take a look.

If a system is a little more inactive than is preferable, then it’s time to stay calm and watch it trade for several more months.

Of course, no one likes to sit around and watch their system not trade.

And I’m one of those people.

If the lack of trading continues throughout the rest of the year, it might warrant some attention.

In the meantime, I’ll be watching closely. Very closely.