Invitation to BSD to prove MACD entry/exit alone can work on long time scales

[BSD]

Do we have a spec or not ???

Instrument: DAX

Time Frame: Weekly data

Indicator: MACD (Dunno the settings, think if I remember they had them in the print version along with more detail but unfortunately I threw that away, anyway can't find em online, but could always contact Hypo Bank who did the tests and ask)

Backtest period: 1989 - 2009 (A typical 20 year investing horizon and right up to when that article was published.)

http://www.capital.de/finanzen/100025900.html

As the 11 other strategies tested were long only - these are strategies for the unassuming public after all for many of whom shorting is presumably an unknown phenomenon - so I would presume this is also long only, so an upward MACD cross is a buy, but a downward cross is a flat, ie not short.

 

[megamuel]

Why not try something like this - If the histogram on the daily MACD is above 0 only consider longs. If histogram is below zero only consider shorts. Then on the hourly time frame, buy when the histogram closes above the signal line and sell when histogram closes below the signal line. Then maybe a trailing stop below or above the high or low of the previous 5 (?) bars. Doubt it will work but try it then we can tweak it!

Sam.

 

[DionysusToast]

I have the strategy but I don't have the DAX on my platform.

So - we need to put it on another market. E-mini data goes back 12 years, so not the 20 we need. Forex data goes back 20 years.

We need to pick a market or multiple markets and a stop loss strategy or 'always in'.

A stop loss strategy can be
- a fixed number of points - but that won't work on multiple markets (if you set the stop loss at 100 pips on forex, you wouldn't want a 100 ticks on the S&P).
- % change based
- ATR based

Also - position sizing can be fixed number of contracts, ATR based and/or based on account balance.

Of course, we can do more complex things such as putting exits n ATRs away from the last swing high/low and position sizing based on that but I think we should walk before we try running...

 

[timsk]

DT,
I'd suggest the YM or ES as they're the nearest proxy to the DAX (that I can think of). Also, my vote would be for an ATR stop as it's dynamic and adjusts to prevailing market conditions. I'd suggest a fixed number of contracts too as this keeps it really simple.
Tim.

 

[BSD]

and a stop loss strategy or 'always in'.

Hypo bank / Capitals version I take it only includes longs.

Stop loss / take profit is an MACD down cross.

No shorts.

http://www.capital.de/finanzen/100025900.html

Kauf = Buy

Verkauf = Sell

 

[bevok]

Shame about the lack of DAX data. I'm not sure whether the ES is an equivalent market or not (never traded DAX) but it is quite mean reverting in nature - certainly the US indices aren't markets I'd look to trade with a MACD system. This is BSD's challenge though, but I could suggest some markets & rules if desired.

Look forward to the outcome - well done for putting your code where your mouth is Toast!

 

[MrGecko]

teh DAx tends to be pretty volatile because:

* the weighting of the stocks is spread far and wide: I can't remember the details, but most of the index is made up of about 10 stocks IIRC.

* it is quoted in Total Return, as oppoosed to price return. That is, it includes expected dividends.

I don't understand why you can't backtest on the SPX instead of the rolling future? What about SPY?

 

[DionysusToast]

Mr G - SPX and SPY would be fine - we'd just need to figure out the position size or we could just do 10 lots.

We can have a switch to disallow shorts. No problem at all to make what we have a long only strategy.

In terms of it needing to be the DAX, if the strategy works because of the TA, then the TA should be robust enough to work in more than 1 market.

On the stop, can you guys agree on the stop method/market ? I want to remain impartial because I don't want to be accused of introducing something that skews the results in favour of my opinions.

We can bring in DAX at the end as a final test, I can just sign up for DAX data for a month, without knowing how long it will take us to get the rules agreed, I'd rather leave DAX till the end. This will also prevent us from curve fitting to the DAX.

 

[BSD]

In terms of it needing to be the DAX, if the strategy works because of the TA, then the TA should be robust enough to work in more than 1 market.

On the stop, can you guys agree on the stop method/market ? I want to remain impartial because I don't want to be accused of introducing something that skews the results in favour of my opinions.

It should be the Dax because that's on what the Commerzbank MACD fund runs, and that's what Hypo Bank / Capital based their 11 tests on including this MACD strategy, and as far as I know US indices are far more reversing to mean than the rather trending Dax.

Can't believe nobody on here has weekly DAX data, somebody should be able to provide that !

Re stops, I can't believe this, this is the THIRD time I am now posting this, what part exactly is proving so hard to understand about this very simple strategy that you're asking this question a third time !!!

http://www.capital.de/finanzen/100025900.html

For the last time, these were the specs of Hypo Bank / Capitals MACD test:

Instrument: DAX (they recommended buying ETF's for people wanting to emulate this strategy)

Time frame: Weekly Data

Testing period: 1989 - 2009

SOLE Indicator: MACD

LONG Only

From the pic above:

Kauf = Buy When MACD / signal line crosses up

Verkauf = Sell When MACD / signal line crosses down (Meaning stop loss < breakeven and take profit > breakeven, but NO shorts. In other words, lol, THIS is the stop loss AND take profit.

 

[BSD]

Re my point above about the trending DAX vs reverting to mean US indices, this guy here did a correlations study between DAX and Dow in 2008:

http://basili.wordpress.com/2008/05/26/dow-jones-dax-korrelation/

Findings where that over the

Last 4000 days: there is virtually no correlation

Last 1000 days: increasing correlation

Last 100 days: very weak correlation

Last 50 days: slightly increasing correlation

His conclusion: apart from a strong correlation phase in the last two years which has since diminished again correlation levels are once again reverting to their long term mean of between 0,2 and 0,5.

So not apples and apples, but apples and oranges the two.

 

[DionysusToast]

OK then - I just checked with Tradestation - $10/month for Europe - I think I can stretch to that.

In when they cross over and out when they cross under. Settings of MACD will be standard settings. Weekly charts, 10 years.

 

[redart]

OK then - I just checked with Tradestation - $10/month for Europe - I think I can stretch to that.

Nobody likes a show-off, big shot!
That 10 bucks represents about 50pips profit to me in my journal!

So do you backtest the method (once coded)in tradestation and see if it gives the same results as the Commerzbank MACD fund apparently got?

 

[BSD]

In when they cross over and out when they cross under. Settings of MACD will be standard settings. Weekly charts, 10 years.

Yup.

And it doesn't say explicitly in the online version but I'd also guess they used the standard settings.

But 20 years was the test.

Think we need to stretch to that, as they already had six down years in that.

Someone here must have weekly DAX data, I mean we're not talking about tick data for the last decades, but just weekly data...

 

[DionysusToast]

OK - Tradestation only has DAX data back to 1999.

Results are attached for the 11 years.

 

[BSD]

Right back from gym !

Well that is ten years missing tho, and as mentioned even the 20 year test had 6 down years, but even so those results don't really correspond too well with this chart with this chart I found here that someone on T2W did ??

 

[bevok]

Hmm doesn't look good does it - that largest winner skews things far too much. Thats the problem with weekly systems, impossible to get statistically significant results on a single market. I don't us tradestation but understand there are addons to let you do portfolio testing (just thinking more markets would give more trades i.e. greater statistical confidence) - do you have anything like this Toast? I realise thats taking us beyond the scope of BSD's spec . . . Alternatively if we could get another ten years that might help.

I see the Index began in 1987, would be fascinating to see the basis of this funds decision to use this technique - obviously specific backtesting wouldn't have played much of a part.

 

[DionysusToast]

Right back from gym !

Well that is ten years missing tho, and as mentioned even the 20 year test had 6 down years, but even so those results don't really correspond too well with this chart with this chart I found here that someone on T2W did ??

Same chart BSD - yours is monthly, mine was weekly.

My monthly & weekly are attached..

 

[BSD]

What springs to mind is the bursting of the web 1 bubble and ensuing bear market which probably took quite a bit away from the 20 year performance. (And even that had six down years already not to to forget.)

I guess that is why I personally am not the greatest fan of buying strength, of trend following systems, even though I absolutely accept that they do provide great net profitability given enough time, and why most trend following funds are also diversified across a range of markets as sthg somewhere will always be trending.

This guy here also trades a long term trend following system, but like Toasts backtest above also on only one market, oil, and an accordingly volatile performance that offers, I wrote about him earlier:

"Billionaire trader through 15 minutes work / day !

...

(take) "Billionaire T. Boone Pickens for whom trading was for a very long time also no more than a means to make money he needed for something else.

He started out as one of the big leveraged buyout artists from the eighties going after giants like Gulf Oil, that's where he made his first fortune.

Life just like trading doesn't just have ups or winners though, and eventually luck turned it's back on him, and he was left with his takeover vehicle that was producing usd 15 million in annual overhead costs alone, but without any takeovers that were succeeding to cover those costs.

So what does he do ?

He spends 15 minutes a day on speculative trading and actually manages to cover his firms 15 mill overhead from his trading profits for several years.

Hard work really never had anything to do with success.

Its not how much you do of something, it's simply about doing the right thing !

But Pickens story gets better.

At some point he is totally forced out of his firm by a hostile takeover - life writes better stories than any movie script sometimes - and he's left without anything to do.

To get out of the rut he decides to start up a commodity fund funded with usd 35 million from himself and some friends in 1997.

That fund starts out with a gut wrenching decline of 90% to less than usd 4 mill.

1999 is more or less flat, but with the advent of 2000 it goes on an exponential equity explosion taking it up to usd 252 mill, and, on top of those realised gains, up by another 148 mill in 2001, for a total of usd 400.

In the past couple of years he has headed TraderDaily's top earning traders lists several times with a personal income in the Billions !

The article from fortune.com is long but very definitely worth a read:..."

Continued:
http://www.trade2win.com/boards/gen...naire-trader-through-15-minutes-work-day.html