Trading the equity curve of the equity curve

[Gringotts Bank]

Say you have a simple system like a MACD cross.

Run it on 5 minute bars of a liquid instrument like a futures contract. The equity curve will be as choppy as anything, but within that will be some significant run ups.

Then apply a filter for the the equity curve, or manipulate the position size based on what the curve is doing. This is a common practice. Say for example you might turn off the system when the curve drops below a MA. Still, you don't trade this yet. Now apply a MA filter to the already filtered equity curve, and trade that.

Can this work? What about a 3rd or 4th layer of filtering?

 

[Lone Wolf]

Say you have a simple system like a MACD cross.

Run it on 5 minute bars of a liquid instrument like a futures contract. The equity curve will be as choppy as anything, but within that will be some significant run ups.

Then apply a filter for the the equity curve, or manipulate the position size based on what the curve is doing. This is a common practice. Say for example you might turn off the system when the curve drops below a MA. Still, you don't trade this yet. Now apply a MA filter to the already filtered equity curve, and trade that.

Can this work? What about a 3rd or 4th layer of filtering?

It’s an interesting idea that I’ve seen others use. But I think about it like this…

Imagine the ASX200 index is your equity curve, can you apply a filter to it that would minimise your drawdowns while maximising your profits? Then congratulations you have a system that beats the index. But many people can’t achieve this.

I think part of the problem is that with an equity curve filter, you are always in the market when it drops (until your filter switches your system off) and you are always on the sideline when your system would have made gains (before your filter switches your system back on). So you always have this equity smoothing that works against you.

I would like to smooth my equity curve by having more than one system, say a trend system and a mean reversion system. Then leave them on all the time so that you never miss the good run when it happens. The problem here though is that the system that is not in ideal conditions needs to drift sideways rather than go into deep drawdown and offset the profits from the other system. And of course sometimes both systems will do poorly and you get double drawdown.

My idea probably has more flaws in it than your equity curve idea. But I’ve just found over the years that applying a filter to my systems usually results in missing out on a good run, then jumping in just in time to hit a bad run. Poor choice of filters perhaps.

 

[Gringotts Bank]

I think part of the problem is that with an equity curve filter, you are always in the market when it drops (until your filter switches your system off) and you are always on the sideline when your system would have made gains (before your filter switches your system back on). So you always have this equity smoothing that works against you.

Maybe it switches on in drawdown, instead of off. This would be advantageous if there appears to be some dependency in the order of trades taken by the original system (winning trades follow losing ones).

So say the original system is a trend following. Then I apply a MA to the curve. Then I buy the dips of that curve and switch it off when it runs, like a mean reversion system. Since so much would get filtered out, this would require a LOT of original trades, which is why I suggested a 5 or even 1 minute time frame for the original system to run on.

 

[Trembling Hand]

Say you have a simple system like a MACD cross.

Run it on 5 minute bars of a liquid instrument like a futures contract. The equity curve will be as choppy as anything, but within that will be some significant run ups.

Then apply a filter for the the equity curve,

Trouble is there are so few trend days on futures that a simple 5 min chart can ride. Then if you are looking to ride a trend a MACD cross will have you out of the trend just as it takes its first break out or worst constantly fading it.

The idea of trading the equity curve is a good one but first you have to have a promising system that captures and hangs on to the very quick and powerful trends in the 5 min or less bar systems without getting cut to bits in the 90% of other times. You need a forward looking system that can recognise the few trending conditions and ride them. That aint a simple task.

 

[Gringotts Bank]

Trouble is there are so few trend days on futures that a simple 5 min chart can ride. Then if you are looking to ride a trend a MACD cross will have you out of the trend just as it takes its first break out or worst constantly fading it.

The idea of trading the equity curve is a good one but first you have to have a promising system that captures and hangs on to the very quick and powerful trends in the 5 min or less bar systems without getting cut to bits in the 90% of other times. You need a forward looking system that can recognise the few trending conditions and ride them. That aint a simple task.

The only approach I know which has forward looking success is neural networks, and these are 3 to 5 days ahead, assessing the overall direction (higher or lower) with a (?) 60% win rate. Can't remeber exactly.

Say you have a NN which says there's a high chance of the market being higher in the next 3-5 days. That tells you to be ready for some trending days.

Actually MACD is probably not the best example. Maybe supertrend indicator, or something like that. Filter and filter what's been filtered, then trade that.

 

[Trembling Hand]

My point is GB that intraday futs trends are very rapid. Most moves are over in a few 5 min bars. Even the real big ones that are very rare will print 3-4 15 min bars in the same direction and throw you off the trend. You need a promising system first then help it out with a filter. Step 1 is the hard bit.

 

[sinner]

Can this work?

Is it possible? Yes.

Is it workable/feasible/rational/sane? No.

 

[sinner]

Say you have a NN which says there's a high chance of the market being higher in the next 3-5 days. That tells you to be ready for some trending days.

No it doesn't.

 

[tech/a]

No it doesn't.

GB
Would work better on something that moves like a snail.
Like a portfolio and equity curve of a group of Blue chips.

 

[skc]

I would like to smooth my equity curve by having more than one system

If you are a part time trader who has a regular income, trading with "spare" capital looking to supplement income, I would suggest not trying to be too smart and just wear the drawdown.

If you are a full time trader needing income every week, then I'd agree with LoneWolf... having multiple, uncorrelated systems is probably a much better approach than a filter on a filter on a filter on an equity.

Remember, a key attribute of a good trading systems is robustness. If a system requires 3 layers of filter to make work (or work better), I'd question the underlying edge / robustness of the system. It would also become quite difficult to know what's going on.

 

[Trembling Hand]

GB right on cue here is a typical trend in futs land lately. 30 min into the cash session on the DAX and we have just done the same daily range as the previous 3 days range. Most of the move was in one 5 min bar. By the time your system is turned on its more than likely we are at the extremes for the day. You are going to need a strong day for the momentum to keep going. They are rare.

 

[Gringotts Bank]

GB right on cue here is a typical trend in futs land lately. 30 min into the cash session on the DAX and we have just done the same daily range as the previous 3 days range. Most of the move was in one 5 min bar. By the time your system is turned on its more than likely we are at the extremes for the day. You are going to need a strong day for the momentum to keep going. They are rare.

View attachment 64644

1 min chart with MACD with very low parameters will yield several entries. The entries will give rise to a curve, and the curve can be bought at dips. Then this produces a curve of it's own, and buy the dips again. Might only get a few signals over a week, I can't tell without intraday data. Actually I have got some for FX, I'll check.

 

[Trembling Hand]

, I can't tell without intraday data. Actually I have got some for FX, I'll check.

Good idea.

 

[Trembling Hand]

1 min chart with MACD with very low parameters will yield several entries. The entries will give rise to a curve, and the curve can be bought at dips. Then this produces a curve of it's own, and buy the dips again. Might only get a few signals over a week,

You test it?

 

[Gringotts Bank]

You test it?

As best I can. The limitation, as Howard has mentioned elsewhere, is that trading even a single equity curve is probably unlikely to improve profits. So doubling that process is unlikely to suddenly make it profitable. However the volatility of the curve can certainly be reduced. I need to try it with a system which actually works.

This is FMG daily with a MA cross system's equity in green, and the single filtered equity in blue. No point doing with intraday time frames.

 

[Trembling Hand]

This is FMG daily with a MA cross system's equity in green, and the single filtered equity in blue. No point doing with intraday time frames.

lol GB. Thats just curve fitted.

 

[Gringotts Bank]

lol GB. Thats just curve fitted.

DO you know what curve fitted means?

 

[Trembling Hand]

DO you know what curve fitted means?

You have just picked a stock and period that you know works. Take it for a spin on some out of sample data.

 

[Gringotts Bank]

You have just picked a stock and period that you know works. Take it for a spin on some out of sample data.

It didn't work though. I mean that in the sense that you'd never consider trading it, even though it ended up in positive territory. Also you might have mis-interpreted each of the lines.

The black line is the share price (FMG daily), the green line is the equity curve, the bright green line is an MA applied to the equity and the blue line is the equity once filtered by the MA.

I've tried multiple possibilities and the chance of a single filter actually helping the equity curve seems quite low (which had been mentioned elsewhere).

So anyway, worth a look. But that's the end of the thread for me, unless someone can add a new angle on this.

 

[Trembling Hand]

It didn't work though.

I think you need to approach it differently to get a result. It needs to be a portfolio curve to catch general bullish market conditions or go down to non time based charts to get faster equity curve feed back than end of min bars for intraday or a 1 minute system curve turning on a 5 minute system for positions etc.

In my thinking anything that can be easily tested must have been arb-ed out of profitably.