OmniVest User Forums
General Discussion Topics
ATM: First Impressions
40 replies, 5306 viewings
|
OmniVest User Forums
General Discussion Topics
ATM: First Impressions
|
Last Activity 4/15/2019 7:43 PM 40 replies, 5306 viewings |
|
|
| Printer friendly version |
Mark Holstius Elite    Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL  User Profile |
The ability to alter our choice of Strategies, Allocations, Long / Short ratio, number of trades, and ranking method based on Market State are all obvious advantages. We’ve only had it for a few days now, so I feel like my 8 month old granddaughter - just learning to crawl. With that in mind, I want to share some of the good things I’ve found even at this basic beginning level. I wanted to demonstrate to myself that the results we’ve seen from Nirvana weren’t just a curve fit to a specific combination of stocks, strategies, and settings... To do this, I obtained various lists for testing by using Omniscan in OT to obtain the top 300 liquid stocks; I then split that large list into 3 random lists, each with a unique set of 99 stocks + SPY. To vary the Strategies used by ATM, I kept 3 of those used by Nirvana and added 5 new ones; I kept the same Market State definitions used by Nirvana, and then used Jim Dean’s rigorously tested “step back and squint” method to choose which strategies to use in each market state. Results of my first test; This first run produced a much smoother curve using ATM, but not exceedingly better than the % of equity results - until you look closely at the sidebar statistics. ATM vs % Equity (both using 10% of equity) 1) MDD 20.7% vs 75.7% 2) Return-Drawdown Ratio 10X higher 3) Avg Ann MDD 8.0% vs 29.6% 4) Avg % Inv 57% vs 166% I feel the most important may be #4... The ATM ending equity is 3X greater while only utilizing 1/3 of the Avg % Invested. Since there’s so much equity available, I retested with ATM % Invested at 20%; Now the ATM excess return is obvious, and the sidebar statistics are again very positive; ATM vs % Equity (ATM at 20%) 1) MDD 25.2% vs 75.7% 2) Return-Drawdown Ratio is 30X higher 3) Avg Ann MDD is 12.2% vs 29.6% 4) Avg % Inv 81% vs 166% Now to my second big question… Are these settings only applicable to this specific set of symbols - a curve fit? In an attempt to answer this, I used the identical ATM settings with the other 2 lists I created. Each list has a unique set of 99 stocks + SPY (necessary for Market State determination). Using list 2; Using List 3; Again, ATM results are much better - with much lower Avg % Invested. Using the combined list of all 300 stocks; The stability of the improvement using ATM with unique stock lists is impressive. This particular test using different lists may not be perfect, but it helps to demonstrate to me that the results are not dependent on a specific set of symbols paired with a specific ATM method (a curve fit). There are many more ATM capabilities that I didn’t modify at all in this set of tests; Market State Definitions, Filters, Ranking, Allocation, Long / Short ratios… I simply wanted to pass on what I’ve learned this first week at a very basic level. I look forward to learning how to fully utilize the ATM capabilities - developing better methods to use both in OT and Omnivest. Mark [Edited by Mark Holstius on 1/21/2018 1:14 PM]  Attached file : 00 Stocks List.jpg (61KB - 2060 downloads) Attached file : 01 Strategies List.jpg (112KB - 2070 downloads) Attached file : 02 First Run.jpg (380KB - 2103 downloads) Attached file : 03-Comparisons.jpg (461KB - 2041 downloads) Attached file : 04 Results at 20.jpg (367KB - 2069 downloads) Attached file : 05-Comparisons at 20.jpg (456KB - 2045 downloads) Attached file : 06-List 2.jpg (433KB - 2039 downloads) Attached file : 07-List 3.jpg (394KB - 2055 downloads) Attached file : 08-All 300 stocks.jpg (413KB - 2085 downloads) | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
Impressive - as always! Thanks! I'm so rusty on OT it will probably take a while for me to get familiar with ATM. I would be really interested in seeing how the results varied when changing the start date of the simulation. I know this might sound silly, but I performed (and posted results for) lots of test on OV that showed wildly varying results over significant backtest periods (years) simply based on what day in January the test was initiated. I'm just an old legally blind crufty dude who can hardly use the computer these days. Your help validating this latest Nirvana tool is appreciated! Keith | ||
| KC Kid Member Posts: 23 Joined: 6/20/2013 Location: Olathe, KS User Profile |
In your evaluation, the one thing you did not vary was choice of strategies in the mix. (Edit: Sorry, I misstated. You did change up the strategies, but my point below of having trending strats in the mix must be the reason for the big drawdown periods in the basic % of equity run.) I jumped on the ATM bandwagon, but have not yet fired it up to do testing. But something has been eating at me that I want to test. Why is ATM avoiding the big drawdowns, I was wondering? Afterall, this new raging bull market state only comes into play primarily in 2017. The rest of the time, only RTM strategies are in use during long and short periods. Next, I was thinking "When have I seen an RTM equity curve published by Nirvana that suffered such drawdowns as show here approx 2002, 2008 & 2011?" Answer: Never. None of the RTM strats suffer such significant drawdowns. But this curve does. And I think the answer is that the base curve contains several trend following strategies that are enabled and trading during periods where they are out of favor. The simple test would be to enable only RTM strats for a normal, an ATS and an ATM setup and compare the equity curves. I suspect the unenhanced % of equity curve will not have nearly the severe drawdowns as the current 'brochure" curve and the ATM curve will be much closer to the ATS curve, ATM possibly still benefitting from the different market balance setting in bull vs bear market states. I think ATM may be a good platform to develop and experiment with market state switching. I would like to find out if market state could be detected with more granularity to switch in trending strats into the mix more often. Surely there were periods outside of 2017 where trending was the way to go. Anyway, I'm not sure ATM is really going to deliver the shock and awe offered at its unveiling. Love to see if anyone can run my test and post results before I have a chance. [Edited by KC Kid on 1/21/2018 8:24 PM] | ||
| Jim Dean Elite  Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
Could it be used to “curve fit”? Sure - but it would be counterproductive to use it that way imho. The way to avoid it is to apply rules for filtering states and ranking trades and applying allocations that “inherently make logical sense” - rather than blindly or mechanically iterating though hundreds of options to find the best outputs. Summary: keep your brain engaged; don’t over-tweak results. Aside: I suspect the most revealing tests for “comfort level” purposes would be to use less sophisticated strategies (such as ones provided with base OT package), which are less likely to have been inadvertently over-tuned during development. Before/after comparisons of those kernel strategies could provide a challenging “acid test”. | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
Still a nice problem to have. My guess is small accounts and those with higher ticket charges will need to be careful. Keith | ||
| hafnium Member Posts: 43 Joined: 10/11/2012 Location: Vienna, VA User Profile |
The answer to your question about why/how the ATM curves avoid the big drawdowns is fairly obvious. The same 3 RTM strategies are being used in the Bear and the Default Market States. Same Equity Allocation (10%). What is different between Bear/Default modes is the Long/Short Balancing. In the Default mode the balance is 10 Longs / 0 Shorts. In the Bear Mode its 2 Longs / 5 Shorts. By the way, the trending strategies only fire in the Raging Bull mkt. [Edited by hafnium on 1/22/2018 7:19 AM] | ||
| Joe New User Posts: 3 Joined: 6/20/2017 Location: Nj User Profile |
Thanks Joe [Edited by Joe on 1/22/2018 8:29 AM] | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
I agree that the % of Equity curve suffers in the comparison because of its use of trending strategies during bear markets, but that’s simply a consequence of the structure of ATM. Jeff’s research with ATM showed us that one of the underlying causes of the poor performance of RTMs in 2017 was the lack of volatility. That lead to the realization that we should be using more trending strategies during those periods. For ATM to have those trending strategies (or any particular strategy) available to use at the appropriate time, they have to be included in the “base” list of strategies - and all of the strategies in that base list have to be used at all times by the % Of Equity calculation. Presently, there’s no way around that. In this set of tests, I was more interested in applying the same ATM method to 3 different lists - each containing a unique set of stocks. The fact that all 3 runs showed consistent results satisfied me that a particular ATM Method created isn’t just a curve fit only applicable to a particular set of strategies and stocks, but rather that it can be quite robust. With that fundamental question answered, I’ll probably stop plotting the % Of Equity curve and use the ATM Method created as the base of comparison. Then I’ll save individual variations of the Method, adjusting other significant variables one at a time in each, and analyze their effect on the results. ATM is an incredible tool. It allows us to refine things and see relationships like never before. I’m sure the rewards will be well worth any effort involved, and I’m looking forward to Ed’s webinar this afternoon in hopes of learning more. Mark | ||
| gbarber Veteran Posts: 282 Joined: 12/30/2012 Location: Pearland, TX User Profile |
| ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
| ||
| Angela Duran Veteran Posts: 168 Joined: 10/11/2012 Location: Nirvana Systems User Profile |
Here's the registration link to today's webinar: https://attendee.gotowebinar.com/register/6177542697186907650 | ||
| gbarber Veteran Posts: 282 Joined: 12/30/2012 Location: Pearland, TX User Profile |
| ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
| ||
| Joe New User Posts: 3 Joined: 6/20/2017 Location: Nj User Profile |
I like your thought process in testing other symbols and some other strategies in ATM. Have you tried adding a bunch of random strategies and seeing what happens? Also have you performed any walk forward tests? Joe [Edited by Joe on 1/22/2018 1:08 PM] | ||
| hafnium Member Posts: 43 Joined: 10/11/2012 Location: Vienna, VA User Profile |
I believe that gxtrader comissions are .005 /share (1$ min per order). | ||
| Tim Lambie Member Posts: 6 Joined: 10/11/2012 User Profile |
I just ran Ed’s marketing model with these commission rates and got $68.5M, instead of 87m with no commissions. Still an awesome outcome, assuming robustness testing by Mark and others works out. Tim | ||
| Tim Lambie Member Posts: 6 Joined: 10/11/2012 User Profile |
We were led to believe at one point we are on the cheaper tiered model, albeit with only one tier but, with the pricing outlined above. I will start a new thread to follow-up since this orthogonal to the ATM discussion. Tim | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
But if i was throwing orders as large as Ed’s simulation grows to, I’d bet slippage would b the big issue. Try it with a 5k account with a more r asonable 1.50 minimum Ticket... | ||
| Steve2 Elite Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD User Profile |
| ||
| MarkG Member Posts: 6 Joined: 7/19/2015 Location: New York User Profile |
| ||
| SteveL Veteran Posts: 189 Joined: 10/11/2012 Location: Boulder, CO User Profile |
First, ATM or ATMS (Adaptive Trade Management System) is not a strategy. I would describe it as enhanced Portfolio Simulation and Trading. With ATMS you have the ability to configure which set of strategies are being used during different Market States (which can be configured to your heart's content), AND AutoTrade will be able to trade that same ATMS configuration (I believe the AutoTrade upgrade is targeted for release this coming week). So, don't think of ATMS as a strategy, but rather a way to simulate and trade with different strategies during different market states. To answer your question as I interpret it, here is a snapshot of the shipped default ATMS configuration with the # of shorts configured to 0 (i.e. LONGS ONLY), AND Leverage set to 1 (instead of 2). The strategies and configuration variables are shown in the lower right-hand corner of the snapshot. NOTE: Commissions are NOT included. [Edited by SteveL on 2/1/2018 7:38 AM] Attached file : 2018-01-28_15-20-08 ATM LongsOnly.png (170KB - 1585 downloads) | ||
| SteveL Veteran Posts: 189 Joined: 10/11/2012 Location: Boulder, CO User Profile |
Here's the comparison with the original ATMS "Universal method", which includes shorts. NOTE: This is still 1x leverage. I'll post a 2x leverage in the next post. NOTE: Commissions are NOT included. [Edited by SteveL on 2/1/2018 7:39 AM] Attached file : 2018-01-28_15-32-09. ATM Universal 1x leveragepng.png (174KB - 1534 downloads) | ||
| SteveL Veteran Posts: 189 Joined: 10/11/2012 Location: Boulder, CO User Profile |
Here's the original Universal method with 2x leverage. That's not applicable to IRA trading. But, I'm including it because that is the way it comes initially configured by Nirvana. For comparison, you also see the Longs Only with 2x leverage. NOTE: Commissions are NOT included. [Edited by SteveL on 2/1/2018 7:39 AM] Attached file : 2018-01-28_15-36-45 Universal with 2x leverage.png (168KB - 1551 downloads) | ||
| MarkG Member Posts: 6 Joined: 7/19/2015 Location: New York User Profile |
| ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
Personally, I think of it as Next-Gen Strategy Builder. |
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend(un)/Freeze thread | |
Toggle e-mail notification |
