AI & Trading

Discussion in 'Trading Software' started by JS11374, Oct 28, 2001.

  1. JS11374

    JS11374

    Thanks for the link. I'm looking at it.

    The one advantage I do have, however, is to inspect the AI system. I get to read the source code and do out-of-sample test myself. So I won't get burned there.

    Still... black-box trading - kind of takes the human-factor out of it. Which I suppose is good. but then again, what good are we if we are only here so that we don't violate the human-origin program trading rule.
     
    #11     Oct 28, 2001
  2. tntneo

    tntneo Moderator

    shrine or traderx wannabe, have you ever tried AI for yourself, do you use automated trading systems ? I do both.
    I said that using past data to forecast the future is not the way to do it. This is because past does not imply future.
    Of course I am aware the past is all you have and I am a dedicated technical analyst who believes everything is in the price.. well.. but everything is in, in the PRESENT at each tick. This does not tell you anything about the future.
    The whole game is to understand that you need to find out what is the current direction, not the future direction. And even then, that's only a probability. And that's enough to make a trading plan and make money.

    From what I hear from the AI system in question, it seems (we know NOTHING about it, by the way) to concentrate on general direction. But is it lagging or forecasting ?

    Anyway, I pointed out the real issues regarding drawdown, volatility and risks and the like. Even if the system is good it can blow up with wrong trade management, and we learned it does not specify entry (maybe not exit either).. Therefore I double my recommendation to be very cautious.
    And while you are polluting this board, Shrine, I will ignore you and let my robot (since you know already machine are better suited for this) so let my robot take out your money in the market.

    tntneo
     
    #12     Oct 28, 2001
  3. JS11374

    JS11374

    tnt,

    Okay, so here's a bit more details about the AI. You seem to have experience with it, so I definately want to hear what you think.

    1. It's NOT a neural net.
    2. It's forecasting. Not lagging. Everything can be done in real time.
    3. It's based on learning historical data (non-technical analysis)
    4. There's no risk management. No entry/exit signals (it's just a forecasting AI. what to do with it is another issue).
    5. Both in-sample and out-of-sample tests are very good.
    6. It doesn't give definitive forecast every month. It gives an associated % chance of correctness with each forecast. There will be months (maybe 1 in 5) when it will tell you it doesn't know.
    7. It can be adapted to any index. possibly even stocks.
     
    #13     Oct 28, 2001
  4. tntneo

    tntneo Moderator

    OK JS,
    So I am affraid you don't have enough to make a decision.
    If this system works (and out of sample data is a starter) then it is interesting to investigate some more. There are all kind of ways to trade and indeed I build several systems for people or myself with very different strategies.
    Forecasting is not possible (you know cristal balls...) that's my belief, OK. But there are things which are closed to that. If you take a long enough time frame, the analysis of the current direction (or change in that direction, speed etc..) will give you the impression it does forecast the future. But here, the debate is not really important imo.

    You still lack crucial information. However, if this systems looks so good and use a new way of programming (it is not neural net you said) I believe you. I left neural nets too and use kind of AI. that's called AI rules based (it models the way the trader works). So there are ways, that's no doubt.
    The only reasonable thing you can do, if you want to proceed, would be to allocate a very small amount of money relative to your risk tolerance. Be ready to loose all that money indeed. And build with actual trading the account history (drawdown, volatility etc).. if the system works you make money, if not you don't loose too much.

    Now, the problem is : you don't have any trading plan.
    You and your buddy need to decide when to enter and under what condition, enter all at once or in parts, how to exit, take profits etc.. etc...
    Since you can't really backtest. the amount you will invest will be used to answer all these questions. It will take some time. maybe a very long time, and you better get help (maybe you know already thanks to him knowledgeable people in the business who can help define the trading plan).

    Without trading plan, you really have nothing at this point. since you can't look at the machine and tell your broker (even online direct access broker) what to do.
    So you are just at the beginning of the process. but that's OK.

    tntneo
     
    #14     Oct 28, 2001
  5. tymjr

    tymjr

    JS11374: “In fact, it only tells you whether the S&P will rise, remain neutral, or fall next month.”

    I’m wondering what specifically defines a neutral determination. I assume this is measured against the close of the previous month. Is there any overlap of the definition boundaries? How often does the indicator assign a higher percentage to a neutral reading? The indicator could easily generate a high percentage of accuracy, depending upon the relationship of current volatility to the extent of the neutral “area” and the degree to which it is capable of duplicity.

    “It doesn't give definitive forecast every month. It gives an associated % chance of correctness with each forecast.”

    You’ve stated that “90% of the time” the indicator is “accurate”. What constitutes being “accurate”? Is it when the actual outcome is identical to the outcome that was assigned the highest “% chance of correctness” by the indicator or some other method?

    “There will be months (maybe 1 in 5) when it will tell you it doesn't know.”

    Does that mean the indicator assigns a higher percentage to a neutral reading or that it doesn’t generate any percentages?

    “if you know which direction the market will go next month, but dont' know by how much or when, how would you trade with it?”

    If I felt that a tool was giving me a fairly reliable idea of the market’s bias I’d simply incorporate its conclusions into a more specific method of entry and trade management.

    Being that I’ve been correct about the long-term direction but couldn’t hold onto a position because of trade management, I’d be more interested in a system that could reliably indicate the type of market environment I was likely to find myself in. That’d help me to alter my trade management to maximize my profitability.

    Again, no need to respond. Just thoughts for you to consider when contemplating your involvement with your friend.
     
    #15     Oct 28, 2001