Combining multiple systems

Corey · Dec 8, 2008

Quote from TSGannGalt:

First, there really isn't a point of talking about how people pick signals out. Few reasons to this:

1. The reason people pick signals from a system is because there is a discretionary aspect involved with the trading one does. Obviously, a Black Box (fully automated) systems and Grey Box (discretion-added) systems are developed differently.

When developing a Grey Box model, you put into consideration what the trader inputs on top of the coded signal. This can be different things... it can be execution, macro market direction, timing, contract size, news event, momentum... basically it can anything. In simply terms, you delegate a trading process between the code and the trader. Which part of the process you delegate depends on the developer.

With all the possibilities a trader can input into the Grey Box, it's useless to have a generalized discussion over how to combine multiple grey boxes.

2. How someone deals with an automated system is going to be different. Just read MGJ's post. Common sense stuff.

3. Most of you lack clarity in how you develop and approach systematic trading. Even more, what's involved in trading as a whole, discretion included. You guys should start thinking about how a trade is processed in detail, step-by-step. You just don't get a signal and do what you feel. Even more, there's more to a system than having a "setup", "signal", "order entry"...

Even if you have a throrough process defined, you have to dig in deeper, considering the risks of each of the steps within the process.

Finally. The point of running system is not because you get don't emotionally involved or some psycho-shit. If that's the reason why you trade signals or systems, you're going to fail.

Now... I'm going back to my real life...

Peace out... bendejos...

Oh come now, TSGannGalt. You know that an ad hominem argument doesn't disprove what I wrote.

While the onus probandi isn't on me (it's on you to DISPROVE what I wrote, not DISPROVE my credibility), I will make it a little clearer, just to point out that your ad hominem attack is pure rubbish.

Nothing in what I wrote indicated that I was taking discretionary entries and exits. In fact, what I wrote is quite easy to implement in a black-box.

Imagine a simple trading system that has two components: a trend following mechanism and a reversion to the mean system. Each system subscribes to the data feed, and after each data-feed update takes some action and posts it to the order execution system. The order execution system waits for all the systems to make their decisions before firing off its orders.

In the case where the trend system and the reversion system enter truly conflicting signals, they can be netted out.

In the case where we have a trend signal that has been processed and just received a new reversion signal, we should take our profit target on the reversion signal (we are reverting to some MEAN, right?), and determine whether it violates our assumption of the trend (aka, simulate an immediate reversal of that magnitude -- does our trend indicator reverse?). If it does, get out of our trend position. Either way, enter the reversion position.

The same is true for the situation where we have a reversion position and get a trend signal.

It gets a bit more complicated when you have more than two systems, but you get the idea.

Ball is in your court big guy. Ad hominem attacks are going to work here, so either swallow your pride and admit you over looked something, or prove where my logic is flawed.

TSGannGalt · Dec 8, 2008

Quote from Corey:

Oh come now, TSGannGalt. You know that an ad hominem argument doesn't disprove what I wrote.

While the onus probandi isn't on me (it's on you to DISPROVE what I wrote, not DISPROVE my credibility), I will make it a little clearer, just to point out that your ad hominem attack is pure rubbish.

Nothing in what I wrote indicated that I was taking discretionary entries and exits. In fact, what I wrote is quite easy to implement in a black-box.

Imagine a simple trading system that has two components: a trend following mechanism and a reversion to the mean system. Each system subscribes to the data feed, and after each data-feed update takes some action and posts it to the order execution system. The order execution system waits for all the systems to make their decisions before firing off its orders.

In the case where the trend system and the reversion system enter truly conflicting signals, they can be netted out.

In the case where we have a trend signal that has been processed and just received a new reversion signal, we should take our profit target on the reversion signal (we are reverting to some MEAN, right?), and determine whether it violates our assumption of the trend (aka, simulate an immediate reversal of that magnitude -- does our trend indicator reverse?). If it does, get out of our trend position. Either way, enter the reversion position.

The same is true for the situation where we have a reversion position and get a trend signal.

It gets a bit more complicated when you have more than two systems, but you get the idea.

Ball is in your court big guy. Ad hominem attacks are going to work here, so either swallow your pride and admit you over looked something, or prove where my logic is flawed.

For starters.... I didn't know what "ad hominem" and "onus probandi" so I wasted time Googling what it meant...

Next... I know what you are talking about. You don't have to reword the things you wrote but to put things into perspective:

1. (Copy& Pasting) Each system subscribes to the data feed, and after each data-feed update takes some action and posts it to the order execution system. The order execution system waits for all the systems to make their decisions before firing off its orders.

2. Before the orders are sent to the exchange you have a conflict signals from 2 systems. Within this example, a trend-following system and a mean-reversion. Trend-following signal is suggesting the trend will continue. But the mean-reversion is suggesting that the market has over-extended and highly possible for it to revert.

Now...

I still don't get why your so determined to prove something that is useless. You seem to suggest that you're supposed to override a system's action by another system's criteria. So is that action part of the system that has been coded or tested?

If it is, then there's no discussion to be made. You just have a system that considers both a trend-following and mean-rev. signal.

If it is not, then where does this action come from. Isn't it discretion? Rationalism? What is it?

Corey · Dec 8, 2008

Quote from TSGannGalt:
I still don't get why your so determined to prove something that is useless. You seem to suggest that you're supposed to override a system's action by another system's criteria. So is that action part of the system that has been coded or tested?

If it is, then there's no discussion to be made. You just have a system that considers both a trend-following and mean-rev. signal.[/B]

Of course it is part of the system that has been coded and tested. That is the point of a mechanical system.

Why do you insist on calling it useless? I just gave an example where it may be extremely useful.

How would you handle such a situation?

TSGannGalt · Dec 8, 2008

Quote from Corey:

Of course it is part of the system that has been coded and tested. That is the point of a mechanical system.

Why do you insist on calling it useless? I just gave an example where it may be extremely useful.

How would you handle such a situation?

OK...

Seemingly, you've gotta major problem comprehending English so I'm going to conclude that you're a retard or just plain ol' non-American (In that case, I'll take back the retard).

Which part of "useless" don't you get? Why would I waste time "handling shit" that I find useless?

Corey... goto sleep. Have a good night rest and reply back when you're pea-brain starts working better.

Corey · Dec 8, 2008

Quote from TSGannGalt:

OK...

Seemingly, you've gotta major problem comprehending English so I'm going to conclude that you're a retard or just plain ol' non-American (In that case, I'll take back the retard).

Which part of "useless" don't you get? Why would I waste time "handling shit" that I find useless?

Corey... goto sleep. Have a good night rest and reply back when you're pea-brain starts working better.

Again with the ad hominem arguments. They are unnecessary, so let's stay on topic. By the way, it would be "your pea brain," not "you're pea brain." Ironic, considering you are questioning my English comprehension, isn't it? I think so. As well, there are other people world-wide who speak English as their native tongue (Canadians, English, Scottish, and the Irish, for example) not just Americans.

Obviously you have trouble comprehending that just because you find it 'useless' doesn't mean it can't be 'useful' for other people. We are indeed discussing a similar situation, but I am trying to point out that there may be cases you are ignoring. Trend and reversion are NOT mutually exclusive. You have have a reversion WITHIN a trend. It all depends on the time frame each system is running on. If my trend system is a 60-minute system, then it is reasonable that I would trade a 15-minute reversion system on top of it. Now an over-extended signal in the 15-minute system does not indicate an end-of-trend for the trend system. It works the other way around as well. A reversion on a 60-minute system looks like a trend on a 5-minute system.

On the other hand, if I have a 60 minute reversion system and a 60 minute trend system that both fire a signal simultaneously (or within a small enough time delta), it might be indicative that each trade is too low probability and I should take it off (once again, this must be mechanically developed and tested before implementing -- but it makes fundamental sense).

Order management (including money management) is just as important to a trading system's success as the entry and exit signals, in my humble opinion.

So I wait with bated breath, still. Perhaps you can explain why using one system to override the other is 'useless' in all circumstances. I seem to have presented cases where it could indeed be useful.

Normally I appreciate reading your posts. I find that you have some good insight. Let's refrain from the pointless insults, please. They add nothing to debate and discussion.

TSGannGalt · Dec 8, 2008

Quote from Corey:

Again with the ad hominem arguments. They are unnecessary, so let's stay on topic. By the way, it would be "your pea brain," not "you're pea brain." Ironic, considering you are questioning my English comprehension, isn't it? I think so. As well, there are other people world-wide who speak English as their native tongue (Canadians, English, Scottish, and the Irish, for example) not just Americans.

Normally I appreciate reading your posts. I find that you have some good insight. Let's refrain from the pointless insults, please. They add nothing to debate and discussion.

You have 2 base systems:

1. Trend-following.
2. Mean Rev.

Plus. Each system overrides itself, under specific circumstances.

So you'll end up testing:

System 1.
System 2.
System 1 + 2. (aka. System 3)

All the shit you've been writing about doesn't matter.

You still don't find this a useless topic? Even more (I'll feed you with some more ideas and things to write back about), you're increasing the vulnerability of the portfolio by intentionally increasing the exposure of a tendency of one system.

That's all I'll add to this topic. Anyone else wanna bite? Any "polite" person want to shed some light into Corey's retarded brain?

TSGannGalt · Dec 8, 2008

Quote from 1 in 100:

TSganngalt

I've been watching you for a while now

you are a geek well versed in technicals and math

YOU JUST NEED TO GROW SOME BALLS NOW and become a real trader

stop wasting time will ya

Time to drop da ET

JaiSreeram · Dec 9, 2008

Quote from MGJ:

As you could in 1995 within Trading Recipes using a "SIMULTEST". Other testing engines that will run multiple systems simultaneously include Stratasearch, Trading Blox, PowerST, Ultra-10, and Mechanica.

Thanks MGJ for the reply.

phattails · Dec 10, 2008

Quote from Corey:

Oh come now, TSGannGalt. You know that an ad hominem argument doesn't disprove what I wrote.

While the onus probandi isn't on me (it's on you to DISPROVE what I wrote, not DISPROVE my credibility), I will make it a little clearer, just to point out that your ad hominem attack is pure rubbish.

Nothing in what I wrote indicated that I was taking discretionary entries and exits. In fact, what I wrote is quite easy to implement in a black-box.

Imagine a simple trading system that has two components: a trend following mechanism and a reversion to the mean system. Each system subscribes to the data feed, and after each data-feed update takes some action and posts it to the order execution system. The order execution system waits for all the systems to make their decisions before firing off its orders.

In the case where the trend system and the reversion system enter truly conflicting signals, they can be netted out.

In the case where we have a trend signal that has been processed and just received a new reversion signal, we should take our profit target on the reversion signal (we are reverting to some MEAN, right?), and determine whether it violates our assumption of the trend (aka, simulate an immediate reversal of that magnitude -- does our trend indicator reverse?). If it does, get out of our trend position. Either way, enter the reversion position.

The same is true for the situation where we have a reversion position and get a trend signal.

It gets a bit more complicated when you have more than two systems, but you get the idea.

Ball is in your court big guy. Ad hominem attacks are going to work here, so either swallow your pride and admit you over looked something, or prove where my logic is flawed.

The intersection of trading systems should be already planned before deciding order execution. If system A behaves differently in the presence of B then that should be accounted for in the model building. I would call it a third system (and deal with the other two factors) or if was some sort of synergistic relationship i would deal with that in the portfolio design. I think what you are suggesting is a scalability issue and should be determined by the portfolio design.

This discussion would be more fruitful if you assume that the strategies generate the same amount of money over a significant period of time. Consider the extreme example: if my trending system typically lasts for a year and my counter trend trade lasts a few seconds, then I need to figure out the point of adding this system and the scalability of it. In a perfect world (no slippage) you scale the positions so the risk:rewards are equal. Otherwise, when trades overlap you scale back on the trend trade when the shorter term mean reversion trade signals.

Perhaps specifying the argument to either order execution or system design would benefit the discussion.

TSGannGalt · Dec 10, 2008

Quote from phattails:

The intersection of trading systems should be already planned before deciding order execution. If system A behaves differently in the presence of B then that should be accounted for in the model building. I would call it a third system (and deal with the other two factors) or if was some sort of synergistic relationship i would deal with that in the portfolio design. I think what you are suggesting is a scalability issue and should be determined by the portfolio design.

This discussion would be more fruitful if you assume that the strategies generate the same amount of money over a significant period of time. Consider the extreme example: if my trending system typically lasts for a year and my counter trend trade lasts a few seconds, then I need to figure out the point of adding this system and the scalability of it. In a perfect world (no slippage) you scale the positions so the risk:rewards are equal. Otherwise, when trades overlap you scale back on the trend trade when the shorter term mean reversion trade signals.

Perhaps specifying the argument to either order execution or system design would benefit the discussion.

Thanks to a Fund Manager in HK, who got a laugh at reading this stupid shit. I'm making my final word... to provide him with more laughs...

...

It's a lot more simple.

I have two systems as what Corey mentions. Though unrealistic, but sticking to his examples under a single lot example, under a time series, you'll have:

"Trend system" - "Counter-trend system"
("L" is for Long and "S" is for Short. "F" is for flat)

L - F
L - F
L - S - Usually, you'll be flat here.
L - S
L - S
L - F - And back to being long here because Counter-Trend closed out.
L - F

Now what Corey is suggesting is having another system influence over the other, so:

L - F - F
L - F - F
L - S - S - You have 1 long + 2 short leading to having an open short position.
L - S - S
L - S - S
L - S - S
L - F - F
L - F - F
L - S - F - The counter trend hits but it doesn't over-ride it.
L - S - F
L - S - F

...

So you have 3 systems. Simple.

Obviously, rather than whining talking about all the what-ifs and dreaming about some shit I can't imagine. You test out how the 3rd system acts as itself.

If the 3rd system performs well, then that's a "step". If not, why waste your resources.

Of course, testing the performance is a step. The next thing to do is test how vulnerable is it. Will the 3rd system fail if either System 1 or 2 fails? If yes, and most likely it's because of the design of the model, you're dealing with wasted resources and increasing leverage on a single tendency.

You make a 3rd criteria (which is actually a system, which the retard never acknowledges because he's got a pea-brain).... As time passes and a system fails. Pre-brain may think that your vulnerability is 66.66% rather than 50%. But if come to think of it, because your system is based on both 2 of your systems, if any one of 2 fails... you'll be dealing with more problems... this leads to a 75% chance (IN GENERAL)...

X - X - X
X - O - X
O - X - X
O - O - O

50% to a 75% chance of being vulnerable... Who doesn't understand why it's useless SHIT?

So... with Corey...

Yeah... you're a freagin' retard bendejo. So... you've been asking me to answer and challenge your useless logic using some Latin words which I still don't understand... Plus all the weird whines about credibility, proof, etc. etc. etc. Seriously, you deal with too much newbies kissing your ass.

You're a piece of shit. Quit ET and trading when you have the time. You're a disgrace to all people who trade for a living. Being book smart and the ability to actually "do and think" is different.

Yeah. Ballz in yo court. Challenge my logic....

BWAHAHAHAHAHAHAHAHAHAHHAHAHAHA

LOOOOOHHHHOOOOOOHOOOOOZZZZEEERRRRR

PS. I quoted phattails but no offense. You're a kewl dood.