Atuomatin Failsafes

Discussion in 'Automated Trading' started by Norm, Jan 12, 2006.

  1. Norm



    I expect to go live with my automated strategy in about a month. I am considering building some failsafes into my strategy, as follows:

    1. Go flat and shut down if losses exceed a preset threshold.
    2. Go flat and shut down if the stocks that I am trading go below a preset price.
    3. Go flat and shut down if preselected indices drop abruptly.
    4. Go flat and shut down if my account balance (equity) drops below a predetermined value.

    Doe anyone have any experience with automation failsafes or any comments on these?

  2. I think the failsafes depend on the type of strategy, but here are my thoughts (high frequency pairs trading)

    1. strict stop loss calculated on volatility estimate (NOT hard coded percentage, dont want to get stopped out by white noise)
    2. Dont use arbitrary consecucitve wins/losses, plot the distributon of returns and estimate tail indices, there are statistical methods that will tell you the probability of losses arising due to small sample size, or real structurual change.
    3. Need some way to handle data stream interrupts, for me this is not so bad because i always trade pairs so the risk is minimal compard to pure directional trading
    4. Predetermined equity value might be dangerous and cause you to prematurely jump out of a position that could turn profitible, again, this is minimized by using strict stop loss policies so your maximum risk is known before even entering the position
    5. "Abrupt" index value drops need to be quantified, is this a good idea? sometimes it can be the most profitible time to trade
    6. Might want to build in "blackout" periods around macroeconomic announcements

  3. Generally, building some kind of failsafe is a good idea, however, it should be based what the strategy is, and what the expected behavior should be. For my system, the failsafe is tightly integrated into a completely separate risk management component, one that can go flat, cancel all working orders, even shutdown in some unforeseen event.

    For me, the rules for failsafe are quite stringent, generally it is a measure of return volatility, as compared to an internal risk model of the market (one that is different from the trading strategy), only if the models exhibit significant deviation would the stop command be sent (all of my components listen on an administrative channel for remote commands).

    So far, in about a year, such a command has never been generated, except twice in testing where I explicitly sent the command manually.
  4. slacker


    5. Do not open more than n number of contracts at the same time.

    Another note. All of these 'failsafes' are 'client based' you might consider 'broker based' failsafes as well.

    For example, I use IB's API and they will not open a new position if your account drops below $2000. So a 'automated test account' only slightly larger than that amount could be used to add a little failsafe on the broker side. It would not protect me from existing open positions going negative however (Rule #1 above).

    Are there other 'broker side' failsafe measures that could be implemented?
    Good luck
  5. Based upon a continuing examination of markets swings on announcements and/or news, it appears to me that a lot of the swings are the kicking in of fail safe programs.

    The major symptom that I believe is being exhibited is how "protection" turns into a market order that has no chance of being filled in the neighborhood of where the "protection" is desired.

    Naturally, there are strategies other than the fail safe class that can be deployed to avoid this condition, situation or circumstance.

    For a first look at the range of possibilities, look at the two lists presented so far.

    Nothing on either list is focussed on making money. Mind you, people do make money during these periods but they are not the sort of people who are making these kind of lists.

    Lets say you are a reader of the thread and, further, you may be able to decide to consider how markets enable you to make money.

    Perhaps the first step is to use your programming to figure out how many other like minded people are doing what you are doing and, further, if possible find out how many others are going to take your fail safe trade to help you out.

    You may find out two things: there isn't a helper for you and more likely, you do not have the capability to do the above because of lack of information available to you because of how you reason and use resourses.

    So you defintiely have to work at this a lot further. Two avenues are available and quite different. Don't get in this situation in the first place. (A major consideration that is telling you something very big and important about how you have reasoned to get to where you are presently in your knowledge, skills and experience.) Or, second, how can you make money in a situation that you have determined is a place to escape from the market and at the same time give the market your capital.

    It looks like fail safe systems are abundant and do increase market volatility and at these times someone is making money at these high velocities.

    Front runners are the ones making the money and mostly "uninformed traders" are supplying the market with capital. Fail safe strategies originate with uninformed traders. Front running strategies originate with informed traders who use market information to front run.

    So we have to get to the bottom line here to save space in ET and to provide the guys who are going to be improving their trading with more means to capture it form the fail safe crowd.

    we can see traders are asymmetrically informed; that is some traders are better informed than others.

    Knowledge of fundamental values allow persons to be better informed those they face are examples of adverse selection.

    You see this "effect" in the market a lot. It has a form that is well known to informed traders.

    This means that as it appears some persons know they are going to be making more money (higher money velocities at this time) at least temporarily.

    What replaces fail safe then? Well, the business school version of what it is is called the adverse selection spread component.

    What group in ET can grasp this stuff and then program it? Does it have to be done even? Not likely.

    The way to deal with fail safe is to make money on it. Naturally, it comes down to being opposite those whose market actions are to exercise fail safe stuff. How can you be opposite the "cascading" of fail safes each time it happens to this abundant group of people?

    It is not a money management issue. Fail safe stuff is the supposed answer to money management expressed needs. These expressed money management needs are a direct consequence of a trading strategy design failure. When the trading focus is on automation, then there is a degree of unnecessary dificulty being added due to the cost of "being informed". More and more of being uninformed is being introduced by the automation thinking orientation. most often a person "pushes" his skills in automating into the picture and pushes being informed out of the picture.

    It would be interesting to see how a person who gets trapped by fail safes manages to have a strategy to reenter the markets. Probably this is not a common kind of connection that is possible.
  6. auto