Are there any systems on Elliott?

Discussion in 'Strategy Building' started by travis, Aug 8, 2003.

  1. travis


    I ask because many people I know are convinced of its validity, but I haven't seen any systems backtested on it and with what results.
  2. GIG


    I can't comment as to whether or not there are any backtested Elliot systems, because I don't know any myself.

    I think one of the problems with the Elliot theory (from a backtesting point of view) is that the structure of the wave is relatively abstract.

    What I mean by this is that you need to define what an Elliot wave 'looks like' in order to identify them in the first place. This is subject to personal interpretation to an extent; I've seen some instances of 'Elliot Waves' that I personally don't think meet the requirement of an Elliot Wave.

    The other hurdle is to actually identify the wave before it's complete, otherwise there would be no use in looking for them. In essence, you wouldn't be looking for the complete cycle, but only the beginning of the cycle.

    If the cycle doesn't end up being an Elliot Wave, then what? You end up identifying patterns that aren't Elliot Waves, and some that are...

    If you were to tackle the Elliot Wave backtest, it would probably be best if an Artificial Neural Network was trained to recognize 'structures' indicitave of a wave. Developing a proprietary stragegy would be difficult if you didn't have ANN programming experience, and I don't know if there are any software packages that could do this for you (maybe if you looked hard enough...).

    Interesting story on the subject...

    I just finished an undergraduate research project involving pattern analysis in the stock market.

    I wrote an application that looked for patterns in stock market historical data. I analyzed 500 companies in total.

    Whenever the application found a pattern it liked, it would 'flag' the date it found the pattern (for a given company).

    I designed some charting software, and I set it up to highlight days where the system flagged patterns it liked. Green bars are 'good' buy signals, red bars are 'bad' buy signals.

    As I looked at the results, I noticed that 'buy signals' sometimes came in 'chunks of three'. I've attached an example for you to look at - it's a chart of of Agilent Technologies (over the past 40 days or so).

    I saw the pattern often enough to notice it, but I didn't really understand the significance of the 'three chunks'. At the time (about a week and a half ago), I didn't know about the Elliot Wave theory.

    Looking at the charts now, these three 'chunks', from what I can gather, are located in Elliot Waves...maybe there is a signficance to this, maybe not...I have to look into it further.


  3. travis


    The reason I asked that is that I use my own system to trade futures, and I realized that the levels and projections that a certain Elliott analyst hands out every day really help my trading, but I can't find a way to, first of all, generate those daily levels myself, and, second, integrate those calculations into my system to backtest it, gain confidence in it, and be self-sufficient.

    I know it can be done, as I have a friend who created a system that generates daily levels and backtested it with excellent results, yet he won't give it to me. So now I am proceeding with my system used along with the analyst's report in a discretional way, which doesn't satisfy me, as the analyst could stop the report at any time.
  4. Hrmm...

    Yes, coded Elliott Theory is there. I've done it and there is one critical problem.

    The setup of the Elliott patterns are actually pretty simple if you stick to what the book says. It's got it's ups and downs but generally, it shows "some" valid edge. But how you test and trade with these Elliott setups and anticipation is the problem. Another problem with the setup (I write setup because Elliott in itself is an analysis not trading method) is the range of freedom with the numbers. For example, would you consider a tick beyond the Fib Retracement Number valid or not, no tick, 1 tick... 3 ticks... or Bar Low crossing but Close above number...

    Anyways... there's the problem and the degree of freedom you give the computer alters the validity of Elliott in their own terms just how each individual gets different wave counts for their analysis. I've got my tested figure but I'm sure few will have the same result.

    Well, statistics... statistics... testing... testing... testing...


    With the backtesting with a system... there a major problem. What kind of signal would you use? Countertrend? Trend-follow? to... Limit order anticipating a change in trend at a Fib Number. Stop-Order of a breakout of Wave 1... Trailing stops at 38.2 or 61.8 of the Top of the current wave... etc. etc....

    Getting a definite Elliott System really comes down to how you percieve Elliott and your coding skill. Just like discretionary traders...:D
  5. see Trading Systems and Methods by Perry Kaufman. He coded the Tom Joseph approach (Advanced Get) in EasyLanguage.
  6. travis


    Thanks, programtrader, I will look for that system.