When to shut down your automated trading system?

[pepper_john]

We all know that it is not a good idea to override your automated trading system (ATS) if it has been thoroughly tested (both backward and forward). However, for contingency planning, are there scenarios that you want to shut down your ATS temporarily?

I have the following scenarios in mind

-- a nuclear attack

-- a large scale terrorist attack

-- a major war

In these cases, I will probably shut down my ATS and buy US bonds instead.

What do you think?

 

[abattia]

Quote from pepper_john:

... However, for contingency planning, are there scenarios that you want to shut down your ATS temporarily? ...

IMO, you want to be careful with this sort of thing:
1. If you’re trading systematically on the basis of backtests/optimizations/etc, this sort of reasoning introduces a discretionary element (e.g. how will you differentiate between a major and a minor war? A large-scale and a small-scale terrorist attack? Etc ..) which may make all your testing useless.
2. Such events can represent maxima in fear/greed, which might be exactly what is needed to make your strategy work (i.e. it could depend on your strategy).

 

[rwk]

I believe all systems are ultimately discretionary. We have to decide how much money to put at risk, though some of the newer, more complex systems may consider that. But whether to continue running the system is always discretionary.

If conditions occur that are clearly outside the range of our testing, then we're in effect trading an untested system. That would seem to be unnecessarily risky.

At what point do we pull the plug on a system that underperforms our tests? It could be that our tests are flawed, or the market has changed. If we pull the plug too soon, would could be dumping a valid system. Too late and we damage our capital.

My own thinking is that all testing has the inherent weakness that it cannot fortell the future. That's why the regulators mandate a disclaimer on marketing materials. Of course, if a system looks bad in testing, there's not much point in using it to trade.

 

[southall]

Quote from pepper_john:

We all know that it is not a good idea to override your automated trading system (ATS) if it has been thoroughly tested (both backward and forward). However, for contingency planning, are there scenarios that you want to shut down your ATS temporarily?

I have the following scenarios in mind

-- a nuclear attack

-- a large scale terrorist attack

-- a major war

In these cases, I will probably shut down my ATS and buy US bonds instead.

What do you think?

If the markets/exchanges are still open and have good liquidity then you should continue to trade the system, volatility will spike in the scenarios you have listed and so you would cut down the number of contracts/shares traded.

 

[vikana]

I will override my system when the underlying assumptions no longer hold. For instance, if I detect unusually large network latency, when NASDAQ went off the air the other day, or if I detect missing data, I go "exit only".

I test my systems on "big news days" such as the Lehman/Bear collapse, flash-crash, and opening after 9/11 to make sure I understand how "bad" it could be. It is less likely that I'll panic if I understand how my systems behaved in adverse environments.

 

[jack hershey]

Traders grow more sophisticated over time.

At some point a trader recognizes that events are opportunities.

The above posts are recognizing the trader's progress and mention events.

For some approaches and markets, it could be recognized how an avent is an opportunity. Most of the above mentioned events represent a shorting sentiment. A good "anticipating" approach that deals with the system of operation of the market will always take advantage of these opportunities.

The opportunities can be less extreme than those mentioned, however. This week on Wadnesday around 1400 hours is a good example. As I remember, 777 mentioned the very end of the day where he took a trade (in the presence of three actual trades).

As a trader becomes most sophisitcated in the use of complete ATS's, he sees two measures that are valid for shutting down. they both have to do with only one thing: the market is not making an offer.

An ATS, grows money continually. Sweeping is used on a periodic basis to keep the utility of the capital at optimum. (There are always other app's of cap for the swept capital.)

Since any optimized ATS more than satisfies any lifestyle; there is no problem of shutting down when no business is there to be done. Even if a person accepts responsibilities of supporting community service functions, these too can bridge the shut down gaps. (Usually the objective is to grow endowments as a substitute for the inefficient raising of funds via a donatiion route.) Floats are invariably longer than shut downs.

so shutdown are a function of two market variables: capacity and liquidity. the former affects trading size and the latter affects the functioning of the ATS.

To use wednesday last again, before 1400 the problem was capacity and liquidity; after 1400 there was no problem.

So with two variables to consider, a gradient can be used for capacity. This is not possible for liquidity, however. Liquidity is a function of relative overlap of data and the volatility of data (lack of to be more precise).

What I personally notice with my system of operation of markets is that my failsafe component is in heavy use as the market becomes more and more idle during RTH. This means that a person can do testing in off hours to determine where the potential application of gradient to liquidity may pan out.

I always have a shutdown regime in effect.

It materializes on two levels, Speaking in an anualized measure, I make half my money before and after Labor Day. So Summer is where the shutdown focus comes into the picture most obviously.

By looking at a coarse measure (weekly pattern), it turns out the sandwich bread is on the ends and the meat is in the middle. It is a three on and two off and Summer cruising used to be a four day event for me. A less sophisitcated measure would be the ATR seen as a synchronized wave on a week.

I probably should post my 5 annotated charts for last week since the gaussian for the week was centered on 1400 of Wednesday.

From a positive point of view, it is a good idea to have your ATS running for all calendar events. It you can't handle the CW emotion system for events; then use Behavioral Finance rules to "fix" your ATS. (See "BF or BS" on home page of BF).

 

[tiddlywinks]

Quote from rwk:



...

But whether to continue running the system is always discretionary.

If conditions occur that are clearly outside the range of our testing, then we're in effect trading an untested system. That would seem to be unnecessarily risky.

...

I did not develop "the" system, I developed "a" system for specific instruments intended to free my focus onto other instruments. Not as a replacement of me!! As the architect of and the historical and live tester of such, I "know" that I think I know what the limitations are, where maintenance may be needed, and under what conditions to expect outsized or under performance. Since my intent is not personal replacement, I use my explicit working system knowledge to determine when ATS usage may not be advantageous. Not using this knowledge would be extremely foolish.

 

[mcgene4xpro]

Quote from pepper_john:

We all know that it is not a good idea to override your automated

I have the following scenarios in mind

-- a nuclear attack

-- a large scale terrorist attack

-- a major war

In these cases, I will probably shut down my ATS and buy US bonds instead.

What do you think?

Those mentioned conditions might be highly favorable for other systems. So, i think it is relative and there is a broad spectrum responses here.

 

[pepper_john]

Quote from abattia:

IMO, you want to be careful with this sort of thing:
1. If you’re trading systematically on the basis of backtests/optimizations/etc, this sort of reasoning introduces a discretionary element (e.g. how will you differentiate between a major and a minor war? A large-scale and a small-scale terrorist attack? Etc ..) which may make all your testing useless.
2. Such events can represent maxima in fear/greed, which might be exactly what is needed to make your strategy work (i.e. it could depend on your strategy).

I understand what you are saying, but the back-testing I did only had one terrorist attack-9.11 and my ATS lost money on that one. Hence the caution.

 

[pepper_john]

Quote from rwk:

I believe all systems are ultimately discretionary. We have to decide how much money to put at risk, though some of the newer, more complex systems may consider that. But whether to continue running the system is always discretionary.

If conditions occur that are clearly outside the range of our testing, then we're in effect trading an untested system. That would seem to be unnecessarily risky.

At what point do we pull the plug on a system that underperforms our tests? It could be that our tests are flawed, or the market has changed. If we pull the plug too soon, would could be dumping a valid system. Too late and we damage our capital.

My own thinking is that all testing has the inherent weakness that it cannot fortell the future. That's why the regulators mandate a disclaimer on marketing materials. Of course, if a system looks bad in testing, there's not much point in using it to trade.

Exactly. Back-testing has its own limitation and I am trying to think some scenarios that will not work for my ATS. The more I think about it, the more I realize that it's not designed to work under these environments.