Quantcast
  1. This site uses cookies. By continuing to use this site, you are agreeing to our use of cookies. Learn More.

ONE and Option Vue Model changes and testing

Discussion in 'OptionVue Forum' started by John Locke 1, Dec 10, 2015.

  1. John Locke 1

    John Locke 1 Guest

    Hello Traders,

    Everyone income trading needs to be aware that model changes and setting changes affect ALL strategies utilizing any sort of Greeks number OR T+0 line as a trigger for entries, adjustments and exits. For example, even if you are only using 30 Delta of a single option for a criteria for entry, the software and settings you use will change that Delta and therefore change the trade. The question is how will the differences change the trade and is it significant enough to bother the system. The challenges most traders have are they they not consistent enough or aware enough to even notice this. We are! So kudos to us.

    Just to be clear we can trade our strategies on TOS, ONE or Option Vue both how it was and how it is currently with the same parameters, the systems will work fine. Sure we may end up making some minor tweeks to deal with a particular model and we may have individual months or specific conditions where results may be better or worse but that's about it.

    The challenge we've had with the TOS and ONE and OV EIOIO models in the past is NOT that they are different but rather that they are much more inconsistent than the OLD Option Vue variable model as to how they reacted over time depending on volatility and skews across the options chain.

    This doesn't mean we cannot use the other models nor that one is more accurate than the other nor that the OV model will provide better results over time.

    BUT it does mean that the old model on the suggested settings produced the most consistent and predictable results over the widest variety of conditions, over the longest periods of time, regardless if you're position had puts or calls at any specific strike price and that is why it was used.

    The inability for the TOS, ONE and EIOIO models to do this resulted in periods of time where the T+0 lines and Greeks numbers were very unrealistic. Is this a problem? Well yes and no. In some conditions it may have turned out to be a benefit and others a problem but the point is when those models were crazy for an extended period of time such as in 2013, and you didn't pick up on it, and you did not adapt your trade parameters, and the market happened to be irrational and push the trade limits. If all these things aligned, then with those models you may have had a losing streak where in OV you did not. Generally when virtually anything else happens, you're fine. And this was not only true for M3 but clients trading other positions such as a weidor or any trade that took on upside risk. This was because the market environment was irrationally bullish AND the analytics of TOS/ONE could not deal properly with the volatility skews resulting in a misinterpretation of risk with price movement. Once the market flatten out it was not longer an issue, however, those conditions do pop up sporadically from time to time.

    So as it stands now we may have lost our most consistent model.... bummer but certainly not the end of the world. We can fall back on the others and still do spectacular over time.... however it may affect the smoothness of profits a bit.

    I know many of you are testing the new OV model and some the ONE Beta model. When doing so please focus on consistency over a wide variety of conditions rather than accuracy. We can easily adjust to accuracy however it is much more difficult to adjust to inconsistency.

    Of course we'd like both accuracy and consistency but if worst comes to worst we can deal with what we have and we will still do spectacular!!!!

    Please keep me up to date on your progress and thoughts with testing the new models and I will do the same.

    Thank you,

    John Locke
     
  2. DukeB

    DukeB Guest

    Hello, John...
    I am probably the least qualified to be commenting on this subject. However, Ken asked me to give my impressions of what I have seen with OV.
    His main concern was that delta was not stable. That it changed radically for no apparent reason. He also mentioned the criteria about how delta was calculated.
    First of all, the deltas, as far as I could tell, were fairly stable most of the time. However, they were inconsistent in back trader if one would leave a given date and return to that same date.
    This seemed strange to me since the software should be using historical data. Historical data should not change. This got my attention since I have developed a lot of software in many languages over the years.
    Ken said they were ready to release the the beta as final. Below is an email from Ken and my comments in red.

    Hi Duke,
    There are several settings that will affect the modeling and this in turn will affect the greeks, especially delta.
    The Define settings for the strike intervals and extents will change the Analyze graph modeling and greeks. Check your Define settings and verify they
    are the same for each instance of your testing. (see the below screenshot for some suggested settings).
    The individual option IV’s are the raw data for the Variable modeling. You can see this in the Model \ Volatility \ Vertical Skew \ Skew curve
    window as shown below. To see the Skew graph, highlight one of the option expirations in the Vertical Skew Curve grid. Then click Skew Graph.
    The Use combined call and put skews box will also affect the modeling and greeks.

    From and end user’s point of view this is a flawed approach. When a trader changes settings in the OV the last thing that should be on his mind is the magnitude effect on delta. Changes are made for reasons of usability (to see more strikes etc). The software should use the data necessary to compute delta regardless of the users choice. If the best calculation is ‘x’ number of data points then the software should use ‘x’ number of data points without regard to the trader’s choice of extents etc.

    =========================================================================================================================
    When in BackTrader, how you arrive at the date your testing is relevant because the Vertical Skew Curves store 3 Day Running Averages of ATM IV’s .
    If you were in a forward date and then went back to your test date, the Vertical Skew Curve data stored will be different than if you were advancing forward from a prior date.
    BackTrader works best if you advance forward one day at a time for this reason. That will best simulate live market conditions and give more consistent modeling and
    Greek calculations.

    The choice of a user to go forward and backward is a foreseen event. That is, it should be expected. Good software design will accommodate foreseen user behavior.
    Each time a new date is selected, it appears a new 3 day calculation should take place using the relevant dates. This goes for any date sensitive calculation. Further that the results should match when calculating on the same date and time
    .

    ==================================================================================================================================== ​

    Here is some additional info on how the modeling curves are generated which helps explain why there can be delta differences in BackTrader for the same date/time/position, different test.

    The program does independent modeling calculations each time you open a matrix. There is an algorithm designed to find the best modeling curve for the matrix IV data,

    which can run up to 500 iterations to locate the best fit. This occurs each time you open a matrix, whether it’s in BackTrader or with live data, so that the modeling is as
    up to date and accurate as possible. This algorithm has several parameters which trigger when the iterations are complete. Therefore the modeling curve can vary slightly
    from instance to instance using the same BackTrader data.


    I believe that your statement ‘so that the modeling is as up to date and accurate as possible’ is difficult to defend when calculations on historical data (data that does not change) produce different results at the same given date and time. The number of iterations is irrelevant given the data being used is the same each time.

    =====================================================================================================================================
    However, if all other Define and Model settings are matched and the BackTrader date sequence is exactly the same for each test, the differences from the modeling curve fit algorithm
    should be minimal.
    I have seen some fairly large delta issues. On 12/4 , which was after I received the latest version, my email to you showed some very large discrepancies from TOS to OV. If you remember, TOS and OV deltas were moving a bit after hours. The issue was that the delta for OV was at times 150% over that of TOS. Also, the delta for OV would change incrementally from 90 to appx 50 and go back and forth. TOS was at 37. The changes in TOS were very minute while OV had very large variances. Since this is the same version of beta that I am now using, I would think that this behavior could occur in the future.
    Due to the way you explain the software design with varying results due to user interaction and the way software calculates delta it would seem that it is just a matter of time before the these variable inputs conspire to produce results that could be (at the risk of being an alarmist) financially damaging to someone using OV to manage their portfolio.
    I have read concerns among some traders stating exactly this.
    I realize I am being very critical. But this is a beta and being critical is, after all, what it is all about.
    Now, after having said that, there are many, many users that are very happy with OV. I have heard experienced users say that materially the calculations work fine.
    Again, Ken, if it were me I would get a very detailed handle on the way calculations produce differing results.
    John, my comments while I believe are true, may not matter at all as far as the efficacy of the product. I offer it up here in the interest of disclosure about a subject that appears to be very important to some individuals.

    Regards,
    Duke
     
    TheSpeculator152 and Andrei like this.
  3. Andrei

    Andrei Well-Known Member

    Duke, give yourself a bit more credit. I think your comments are spot on.

    The problems with OV are a major pain to deal with. It is completely unacceptable for a vendor to introduce changes of such magnitude with so little testing, thought and consideration.

    However, it had one positive side effect on my trading: it forced me to pay more attention to other models and thus become less mechanical and more aware of what the market is telling me. Hope it will be reflected in my P/L as well;)
     
    Last edited: Dec 13, 2015
    GreenZone, Paul and TheSpeculator152 like this.
  4. TheSpeculator152

    TheSpeculator152 Well-Known Member

    @DukeB, spot on.

    I know quite a bit about software designed as well and completely agree with your viewpoints. Changing calculations based on whether one is going backwards or forwards does not make any sense at all.

    When reading your post I'm getting quite concerned as to whether I should use OV as the basis for my options trading. The more I learn and the more I study, the more concerned I become OV have seen its glory days already.


    Sent from my iPhone using Tapatalk
     
    GreenZone likes this.
  5. Kevin Lee

    Kevin Lee Well-Known Member

    This is a good video to watch. JL explaining the model confusion.



    This video was good done before the v7.80 launched though.
     
    Last edited: Dec 14, 2015
  6. John Locke 1

    John Locke 1 Guest

    Duke,

    I agree with you completely.

    And Andei I agree with you as well.

    The good thing about this is that more people are aware of how "none" of the models is a great predictor (they are loose interpretations at best) and now realize how unscientific this whole trading thing is.

    Hopefully this will focus those stuck on "following directions" into discovering the art. They need to think conceptually, worry less about exact science and learn the flow :).

    And hopefully we will end up with a better model as well, but if we don't a good trader can easily work with what we have.
     
    Andrei likes this.
  7. DukeB

    DukeB Guest

    John,
    Thank you for your response.
    Kevin sent me a video where you are explaining that the available analysis platforms are basically estimations. Thank you Kevin. I agree with you completely that the art of trading rather than the science of trading is where focus should be placed. I tend to reside mostly with technical analysis of charts for short term trades as in day trading the e-mini. However, I am trying to learn market feel. Since I am probably the least competent trader in this forum I sometimes have to really study what is being said by the members and find myself a bit envious of their mastery of the subject and immediate comprehension of the comments by other members. Their responses prompt even more study. So it is back to school on a daily basis.
    Again, thank you for the time you took to enlighten me on the details of the various models.
     
  8. DukeB

    DukeB Guest

    I believe the deltas are more stable, not necessarily more accurate. The best comparison is by John Locke. He has much more experience and knowledge about the various models.
     
  9. ACS

    ACS Well-Known Member

    In the Monday morning update, John said he noted some issues with the new model that might require changes in Delta guide lines in some of the trades to the down side. Stay tuned.
     
  10. TheSpeculator152

    TheSpeculator152 Well-Known Member

    Thanks. I'm wondering when ONE will release it's updated version with a better volatility model. I saw demo of it yesterday and it looks absolutely awesome compared to OV!
     
    GreenZone likes this.
  11. Boomer34

    Boomer34 Well-Known Member

    This sounds like GREAT news...any idea on the timeframe?
     
  12. TheSpeculator152

    TheSpeculator152 Well-Known Member

    No. I hear January but the beta testers in this forum might have more details.
     
    Boomer34 likes this.

Share This Page