RIP Mechanical Trading Systems?

[meanreversion]

True and its getting bloody tedious staying right up between the 2 and 3ATR on my Keltner Channels for so long. I don't like entering trades when its up at that level so I've missed a large part of the move.

All hail the Bernanke put - which has some kind of ratcheting feature as well perhaps.

Do you know your annualised daily vol? Without that, or max drawdown, the percentage return is meaningless.

Also, whilst I appreciate you've gotten good returns for several years, do you know with certainty the counterfactual, i.e. did you consistently outperform the model (by achieving higher Sharpe or Calmar ratio)?

 

[Hotch]

Not sure that I agree with that point. I generally trade pullbacks to trends but my discretionary element will only considering entering trades when the pulback is orderly for axample. Its very difficult to define what "orderly" is but if I look at a chart I can decide instantly.

Understanding the overall market conditions is an important discretionary part of my trading too. A long trade that I would have entered 3 months ago I won't do today as I think most of the risk to the market is on the downside at the moment.

Difficult != impossible. Your brain after all, is only a rather complex machine.

Look at it this way, if I show you a chart, you can apply your method and say whether it's an entry or not. If I show you the same chart tomorrow, and ask you to apply your method, you're still going to come up with the same answer.

Beckham can't model the trajectory of his free kicks, doesn't mean it's impossible.

 

[Hotch]

If there was a good EA, there would only be one....

What is this? Highlander?

There's a significant difference between the EAs which are mass marketed and the ones which people produce and use privately.

 

[x4x]

What is this? Highlander?

There's a significant difference between the EAs which are mass marketed and the ones which people produce and use privately.

If you think there's an EA that would work in the long term without any discretion you should be in One flew over the cukoo's nest, never mind highlander. LOL

 

[Hotch]

lol

 

[meanreversion]

Excuse my ignorance, is an EA a complete trading system, or just something that tells you when to get into a trade?

 

[Hotch]

complete system.

 

[meanreversion]

If you think there's an EA that would work in the long term without any discretion you should be in One flew over the cukoo's nest, never mind highlander. LOL

This statement needs clarifying, otherwise it makes no sense.

It's not hard to think up a system that would have made money over the last 2, 5 and 10 years.. a little optimisation will do it.

Question is, will it work going forward? But it depends on your definition of success. Is a system that returns +100% in 3 years and -5% in 7 years a success or not? (for example)

Or is a system that returns +10% for 9 years then -50% in the 10th year a success? etc.

 

[the hare]

If you think there's an EA that would work in the long term without any discretion you should be in One flew over the cukoo's nest, never mind highlander. LOL

How long do you call long term ? mine's been going for 8 years.

 

[counter_violent]

How long do you call long term ? mine's been going for 8 years.

Genuine questions.

Did you write this EA or have it written?
Are we talking about a MT4 EA or some equivalent in another package?

Ta

 

[meanreversion]

I think this kind of goes to the heart of the issue as to why people automatically distrust systems. They have no metric for comparison. Thus, no system can be any good, otherwise a) everyone would trade it or b) the market would become more efficient etc.

You hear comments that systems are not "robust", or "consistently" profitable etc. without anyone actually defining what they want in a system.

If you want an EA which makes money every single day, it doesn't exist, move on.

If you want an EA which has a very high probability of making money over 10 years, that does exist, and potentially with decent return/vol ratio as well.

Actually, on that note, to anyone reading this who thinks systems are bunk, what WOULD a good system look like? What parameters or metric would you want to see? What kind of annual return, or max drawdown, or Sharpe ratio, or win rate, anything?

 

[meanreversion]

If you think there's an EA that would work in the long term without any discretion you should be in One flew over the cukoo's nest, never mind highlander. LOL

This question to x4x - define in concrete terms (win rate, annual return, annualised volatility, max drawdown) a GOOD system.

 

[x4x]

I am a full time 'system trader' - so I have no problems with systems. However, I have to adjust my strategy according to market conditions. No point trying to trend trade when the market is flat or choppy, or range trade if it's trending.

Can an EA look at the market and decide various strategies according to what it's 'seeing'?

If it can then I'll accept it could work, if it can't then the market will ultimately go into a phase that it won't handle.

Do you really want me to post charts from pre-2008 and the 6 months after Lehmans went pop?

I'll be very surprised if an EA could handle the different market dynamics......

 

[meanreversion]

Trend systems all cleaned up in 2008, and did fairly well in 2009 as well. Is this what you mean by EA?

 

[the hare]

Genuine questions.

Did you write this EA or have it written?
Are we talking about a MT4 EA or some equivalent in another package?

Ta

I originally wrote it. I do have an MT4 version that I use for ongoing research. I've a few different versions, written in Java, C# depending on the API its conncecting too.

 

[meanreversion]

I originally wrote it. I do have an MT4 version that I use for ongoing research. I've a few different versions, written in Java, C# depending on the API its conncecting too.

oooh professor, you're my hero ..

 

[counter_violent]

I originally wrote it. I do have an MT4 version that I use for ongoing research. I've a few different versions, written in Java, C# depending on the API its connecting too.

Yep fair do's cheers.

 

[Hotch]

Can an EA look at the market and decide various strategies according to what it's 'seeing'?

Yes.

BOO!

 

[DashRiprock]

Actually, on that note, to anyone reading this who thinks systems are bunk, what WOULD a good system look like? What parameters or metric would you want to see? What kind of annual return, or max drawdown, or Sharpe ratio, or win rate, anything?

I'm not in the "systems are bunk" camp, but surely a system is only "good" if it's better than what you've already got?

 

[x4x]

This question to x4x - define in concrete terms (win rate, annual return, annualised volatility, max drawdown) a GOOD system.

I can only define what is a good system for me, but that is not a good system for everyone.

All I meant was it's not possible to set an EA running without monitoring market conditions. If it was set to run 24 hours 7 days per week I don't believe it would return a profit over the long term - long term is until the next big event....

Let's say tomorrow there was an absolutely catastrophic world event - the trader would need to make a discretionary call in regards to his strategy.

I believe the person running his EA would make a discretionary call on whether to run or not to run it on that occasion.

The EA hasn't blown up the account solely due to the human intervention - which to me invalidates the purpose of an EA.

I have taken this to extremes, but this happens on a day to day basis, only drastically scaled down. A trader has to make lots of discretionary decisions on a day to day basis as news is released which can have a dramatic effect on the market.