OmniTrader Upgrade Forums
ATM
RTMs – A Significant Observation?
16 replies, 2728 viewings
|
OmniTrader Upgrade Forums
ATM
RTMs – A Significant Observation?
|
Last Activity 4/21/2021 9:48 AM 16 replies, 2728 viewings |
|
|
|
| Printer friendly version |
mholstius Veteran   Posts: 175 Joined: 1/13/2017  User Profile |
The multiple reversals, swings, trend line breaks, consolidations, etc. – all condensed into 1 year – make it particularly valuable for testing and development. I regularly use RTM strategies. They have statistical confirmation over the long run and appear to be utilized by Institutional algorithms - a significant consideration when I started my Follow The Money portfolio in Elite back in 2017. Because I’ll be using that FTM portfolio for the examples in this thread, I feel it would be helpful to do a quick review of how it works and allocates equity to trades… FTM uses 5 dynamic lists to select stocks daily based on their money flow and correlation to the Dow, SP500, Small Caps, Financials, and Energy - hopefully avoiding some survivorship bias. The chosen symbols are then paired with various RTMs, creating 200 strategies. Those 200 strategies allow it to vary the allocation to any symbol based on how many times the symbol is traded by the various combinations of correlation and RTMs. It’s a process similar to multiple timeframe confirmation. If a symbol is only selected once on a day, it allocates the default allocation to that trade. If selected by more than one strategy (different RTMs and/or correlated with more than 1 index), it can allocate as much as 5x more to the trade. Remember, this was developed before we had the benefits afforded by ATM… I’ve been trading FTM for the past 2 years using a small default allocation of 2% / trade. If a symbol was chosen by 1 strategy it would allocate 2% to the trade. If chosen by 5 or more strategies, it would allocate up to 10% to the trade (2% managed by each of the strategies). I felt that the subsequent diversification would be a benefit. The results of this study indicate that I was wrong. A problem with RTMs is that they can have a string of small profits followed by a significant loss that can negate the gains. This post will examine a method that has the potential to reduce that problem. The most significant losses with RTMs come when a market downturn is substantial enough to trigger multiple RTM trades without making the rebound that they need to be successful… The RTMs are trying to catch the proverbial falling knife. I’ve expended a lot of effort attempting to develop a filter to stop RTMs from entering those trades but haven’t been successful – yet. Everything I’ve tested has a negative effect on the good trades that eliminates any benefit derived by avoiding trades during the DDs. (If you have any suggestions for filters you may have developed to avoid the DDs, I certainly invite you to share them... or at least hint at what you’ve found) Here’s a diagram to illustrate what I’m addressing concerning the balance between normal and DD allocations; What I’ve found is that when the “falling knife” situation occurs, RTMs usually enter enough trades to drive the allocation to 100%. Even if you can filter out some of them, there are often enough unfiltered trades available to still be 100% allocated in the wrong direction. That’s the crucial point. It’s difficult to stop the RTMs from allocating ~100% in “catch the falling knife” situations. They’re doing what they’re designed to do and entering trades in stocks that will, unfortunately, continue declining. The fact that they usually result in 100% allocation is useful information, since those catch the knife DDs can be treated as a somewhat stable problem – a loss often in the neighborhood of -10% at ~100% allocation. The “light going on” for me was the realization that we should be able to improve results by increasing the allocation during the periods that the RTMs are trading correctly – making the profits large enough to offset the falling knife losses. If thought of as CALMAR; trying to increase the Return side of the equation while the DD side is relatively fixed. Historically, I’ve followed rather typical trading rules and tried to mitigate risk by taking many small trades at a low allocation. Unfortunately, those “catch the knife” losses usually have allocations at a much higher level ( ~100%) that can negate the sum of my lower allocation wins. So, if the “catch the knife” DDs can’t be changed, what happens if we significantly increase the allocations to the normal trades - and thereby improve the gain side of the gain / loss ratio? To test the concept, I significantly increased my default allocation in my FTM portfolio for 2018. Instead of entering trades using from 2-10% / trade, I increased the multiplier so that it used from 20-100% / trade. The snag below shows 3 charts from 1/2018 to the present: SPY FTM, 2X margin, Long Only using the original 2-10% allocation FTM, 2X margin, Long Only using the modified 20-100% allocation The 6 “catch the falling knife” DD periods in 2018 are numbered in all three charts; Both FTM charts are approximately 100% allocated during the 6 falling knife periods and suffer approximately equal DDs during those periods. The critical difference is that the Modified chart has substantially better returns during the highlighted periods between the catch the knife DDs due to its higher allocations (visible in the allocation pane). Significant improvements from the Original to the Modified higher allocations; HR from 60% to 69% TPM from 104 down to 18 CAR from -12% to +60% Equity from $86,858 to $165,599 Trading with high allocations is not what I’m used to doing. It was something I probably examined 2 years ago and dismissed as a curve fit. Unfortunately for those of us trading FTM, it appears that I was wrong. We would’ve been much better off using these higher allocations in 2018. (sorry, all… maybe some of you did it on your own?) I haven’t changed any of the systems used by the portfolio for these tests. Any changes to the results in 2018 are due to the change in allocation. Since these new figures are so much better, I was concerned that it was just a curve fit to the unique situation in 2018. I found that the “out of sample” results for the previous 5, 10, and 15 years also showed substantial improvement and similar stability, as shown below; Avg Ann DD increases by ~6% Avg % Invested doubles (~90% vs ~45%) ~80% fewer # trades / month (15-17/mo vs 80-90/mo) CAR approximately doubled with substantially higher ending equities As I said, trading with these higher allocations will be new for me. I may find it a bit uncomfortable at first, but all the data indicates that it could be a significant improvement. As a result, I’ve increased the allocation levels in my FTM accounts and plan to just let it run. There will undoubtedly be times when a “normal” trade (not in a falling knife DD) will be negative - but this historical data indicates that the results should be stable, even while allowing for a slightly higher Avg DD. Higher allocations appear to be a better choice when trading RTMs (at least with FTM). They appear to counterbalance the ~100% allocation during falling knife DDs, and the 5, 10, and 15-year charts show a much more stable continuation of the curve into 2018, even with its volatility. I’ve been out of town taking care of some medical issues and haven’t had time to test this further. Given the superior ability of ATM to rank and filter trades, I expect to obtain even better results when using it. Add to that the possibility of raising the HR with AI, and this higher RTM allocation concept definitely merits more attention and testing. Since I feel that the concept is a significant finding, I wanted to share the information in a timely manner. I look forward to any comments and observation. Maybe some of you had already discovered this, already been trading with higher allocations using RTMs, and could provide more information or data…??? Mark [Edited by mholstius on 1/29/2019 5:22 PM]  Attached file : 00 Allocations.png (25KB - 1904 downloads) Attached file : 01 3 charts.png (355KB - 1963 downloads) Attached file : 02 Compare 5 year.png (249KB - 1868 downloads) Attached file : 03 Compare 10 year.png (239KB - 1867 downloads) Attached file : 04 Compare 15 year.png (246KB - 1802 downloads) | ||
| Jim Dean Sage   Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
1. Higher allocations per trade significantly increases Risk of Ruin. 100%/trade is “asking for it” imho, unless a crystal ball is involved, or trades are protected by options, etc. 2. It’s wise as a trader, when things don’t go well, to periodically re-examine the basics of the trading strategy - as you’ve been doing. I suggest you re-examine the core components - RTM rules. Modify them to prevent the falling-knife issue. [Edited by Jim Dean on 1/29/2019 5:32 PM] | ||
| LSJ Legend  Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL User Profile |
I read of several different approaches of dealing with this. Mostly it entailed defining the market state to determine if RTM or trend trading was appropriate. Using a 200D moving average of the S&P500 and a 100D historical volatility of the S&P500 were a couple of the factors used to determine state for trend or RTM. I found it also stated that there is an art to assessing when to trend and when to revert. I'm with you on ATM being the tool that can help us find the right combination. I agree with Jim that the high allocation may be a risky solution. I have looked at Risk of Ruin calculations and find that high allocations don't suit me personally. A 50% allocation with a 70% hit rate has an 18% ROR. A little too much for an old guy like me who doesn't have that much runway left <:) The beauty of all of this though is that we should be able to suit our individual choices with ATM helping us find the right time and allocation for our strategies. | ||
| Buffalo Bill Legend Posts: 539 Joined: 10/3/2006 Location: Stafford, VA User Profile |
This is exactly the same experience I suffered thru last year. small gain after small gain then big loss. Some thoughts - use $VIX as a RTM cutoff? $VIX>25 lets say, or VTY price on the VIX? The day that the stock/index drops and triggers the RTM all-in scenario you would hope would have some tell-tale change like $VIX increase (for the market) or some change in the stock itself - maybe a drop bigger than the ave ATR over the last X days? Another factor - and this bit me - bad news causing a sudden drop. Had a pharma stock get entered by RTMs after a sudden drop from a long upswing. The drop was caused by the company's main drug failing to pass the FDA test. That isn't going to rebound and currently I know of no way to scan for things like that (bad news vs tech indicators) | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
No way to scan for "bad news" ... but you can scan more generically to eliminate symbols that "characteristically" have a major fraction of their moves up/down composed of either really big jumps or really big bars. This is typical of pharma stocks. It's almost impossible for tech analysis to predict those moves, good or ill ... sort of a crap shoot. So imho it's best to keep them out of mechanical trading focus lists. I'll very likely be coming out with a suite of Scans that deal with this and other "personality" features, later this year. | ||
| Mel Veteran Posts: 291 Joined: 12/30/2003 Location: Kensington User Profile |
Attached file : Dynamic Position Sizing.pdf (202KB - 2358 downloads) | ||
| Vinay Elite Posts: 640 Joined: 12/9/2011 Location: Planet Earth User Profile |
It also became apparent that it is very easy to curve fit RTM strategies, you just need to side step few trades by changing System or Filter parameters which results in most of the draw-downs. However in HRE trading these severe draw-downs will surely hit us and take away all the profits which we make on several trades. Now I have stopped trading RTM strategies and switched back to my old style of Trend trading. It also involves severe draw downs but I also get few home runs which takes care of these and still be profitable. I want to make it clear that I don't want to discourage anyone from using RTM strategies. Some of you might have found ways to avoid big drawdowns and profit from small but consistent income these strategies generate, but these are not for me. [Edited by Vinay on 1/30/2019 11:54 AM] | ||
| mholstius Veteran Posts: 175 Joined: 1/13/2017 User Profile |
As I said, I just wanted to share what I found: that the inevitable losses generated by RTMs “piling on” when there’s a significant drop in the market might be mitigated by increasing the normal allocation. To be honest, I was surprised by the magnitude of the improvement that came with increasing the allocation. I’m sure it’s only part of the puzzle. Decreasing the DDs remains a primary objective for all of us, and these results might be influenced by the weighting and ranking factors used in the FTM portfolio’s dynamic lists. The basic premise of increasing the allocations might influence the direction taken in further testing by folks using ATM, etc. I expect to see substantial improvement in the results when applying the benefits of ATM and the GA. To Jim and Larry… Most assuredly, the risk of ruin goes up with increased allocation – so I did some additional analysis. The FTM portfolio increases the trade allocation based on how many times a trade is triggered for a symbol in the various strategies that day: 1X = 20%, up to 5X = 100% The following capture breaks down the trades over the past 15 years by allocation with a couple of ROR calculations below the table; The table has a row for each of the allocation levels (20% to 100%) with conditional formatting to color code the performance in each column. The average allocation was 32.9% (last row, highlighted in orange). Row 1) 995 of the 1,663 trades over the past 15 years were done using 20% allocation; they accounted for 59.8% of the trades and 36.4% of the total equity invested. Row 5) Only 41 of the trades were made using the full 100% allocation; they accounted for 2.5% of the trades and 7.5% of the total equity invested. It’s interesting to note that in general, the performance (PPT, HR, etc.) increases as the allocation increases; the trades triggered by more than one strategy. This tells me that the multiple confirmation method in the FTM portfolio is working. Obviously, this would change if I limited the allocation to a max of 60% or 80% since different trades would then be taken. I’ve also included 2 popular Risk of Ruin calculations and their web sites. The 1st shows a 7% ROR but the calculation isn’t very specific. The 2nd (a Monte Carlo calculation that varies with each run on the web site) shows that over 15 years there’s about a 6% chance of losing 80% of equity. If you set the loss limit to a 100% loss, it calculates a 0% chance. Buffalo; I agree with your suggestion about using $VIX – and the new VAT method just released by Nirvana may help with research and development in that area. Mel; thanks for reminding us about your presentation about dynamic sizing and attaching it. I hope Nirvana can do more with that, as I’ve always found your concepts and ideas to be stimulating. Vinay; I hope that further testing and development will make the RTMs more stable and productive. This allocation idea may be a step in the right direction, but it needs more verification beyond this one example. Mark [Edited by mholstius on 1/30/2019 3:18 PM] Attached file : 00 allocations table.png (177KB - 1703 downloads) | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Glad to see you have included the RoR in your analysis. Suggestion - to probably widely varying degrees, “we” (collective) may believe that the core RTM logic that all this is built upon has been to some degree or another compromised by indirect, unintentional curve fitting. So - again just my suggestion - I’d suggest applying a factor of at least two to the RoR (ie double it), and similarly doubling guesses of actual vs historical DD’s, and halving good numbers like APR. jmho fwiw. | ||
| mholstius Veteran Posts: 175 Joined: 1/13/2017 User Profile |
Looking forward to your work on an improved RTM method. Mark | ||
| John J Veteran Posts: 272 Joined: 8/3/2010 Location: Leduc, AB User Profile |
I'm not sure how RTMs can be improved upon unless you're able to peek into the proverbial crystal ball for black swans...? You said you were able to improve your ROI for 2018 by increasing your allocation. Have you been able to run the same test in let's say 2001, which saw the huge 9/11 black swan drop? Whenever RTMs are being promoted it's almost always over a very long time span. As we all know, the markets have been in a predominant bull market for decades, with the occasional hiccups here and there. Any long RTM strategy would have been able to recover from those hiccups, but that's obviously not a great solace for the guy who went all in on 9/10... The notion that 2018 may have been an outlier and that market should resume it's upward slope doesn't cut it for me any longer. We can't get around the fact that the HRE is uncharted territory (no pun intended). But even if a new trending market is just around the corner, who really needs software to trade it? it's "difficult" markets like we've seen in 2018, 2008, 2001 that should be the only test beds (or litmus tests) for sophisticated trading platforms like OT/VT. If we can't eek out profits in years like that, then we need to start applying fundamental tools such as the piotroski stock screener or some discretionary "algorithms" to the mix (not sure what that means myself, but you get my drift...). | ||
| mholstius Veteran Posts: 175 Joined: 1/13/2017 User Profile |
Here are snags of 4 time periods to add to the 2018, 5, 10, and 15 year periods I posted above; 1/1/2001 thru 1/1/2004 1/1/2007 thru 1/1/2010 The whole range 1/1/2001 thru 1/1/2010 And for the unfortunate guy you mentioned who started on 9/1/2001; Hope that helps…? Mark [Edited by mholstius on 1/30/2019 5:25 PM] Attached file : 0 2001.png (123KB - 1577 downloads) Attached file : 1 2007.png (116KB - 1595 downloads) Attached file : 2 all.png (108KB - 1560 downloads) Attached file : 3 sept 2001.png (112KB - 1553 downloads) | ||
| LSJ Legend Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL User Profile |
[Edited by LSJ on 2/1/2019 2:44 AM] | ||
| Buffalo Bill Legend Posts: 539 Joined: 10/3/2006 Location: Stafford, VA User Profile |
| ||
| mholstius Veteran Posts: 175 Joined: 1/13/2017 User Profile |
If this was “just to get something started”, I can’t wait to see where you end up! Please keep us posted. Mark | ||
| DesertDude Veteran Posts: 108 Joined: 12/15/2003 Location: Texas User Profile |
Darrell | ||
| mholstius Veteran Posts: 175 Joined: 1/13/2017 User Profile |
To summarize, my theory is that during major DDs RTMs “pile on” - resulting in a higher allocation of equity during the DDs that can negate the gains made with lower average allocations during normal trading periods. We don’t avoid the DDs, but by raising the Avg % Invested during “normal / profitable” periods we build up enough equity to weather the storm and remain profitable over time. The statistics I shared in this thread (above) showed consistently better performance when using higher allocations that raise the Avg % Invested with the FTM RTM portfolio. I stated on 1/29 that I was going to adjust my allocations: “trading with these higher allocations will be new for me. I may find it a bit uncomfortable at first, but all the data indicates that it could be a significant improvement. As a result, I’ve increased the allocation levels in my FTM accounts and plan to just let it run. I’m happy to say that it’s working well so far… I’m trading Margin and IRA accounts using Nirvana’s servers with IB - an excellent combination that’s worked flawlessly for 2 years now. Those accounts are up the same percentages as these snags from OV; In slightly less than 3 months from 1/29/19 to today: Margin +28% IRA +14% (SPY +10%) Margin; IRA; I’ll be the first to admit that 3 months isn’t enough time to validate this, especially since we haven’t had any large DDs since January… However, it’s important to point out that this portfolio’s been trading live in Elite Traders since before 2018, and that these charts starting 15 months ago in January 2018 represent trades that were actually taken. They’re not just a theoretical BT or FT. The improved performance is just the result of modifying the Account Settings in OV, not any changes to the system or its choice of trades. Had we been using these allocations in 2018, we would’ve been up +110% using Margin and +50% in an IRA. To everyone trading it (myself included) … I’m sorry I didn’t recognize it sooner. It would’ve been nice to be writing about that. Some subscribers told me they were planning to use allocations higher than the low ones I originally posted and used in my own trading. I certainly hope you did. I’ll let it run unattended with these new allocations and update as things progress. In addition... I'm working on systems that take advantage of the new tools and concepts Ed and Jeff are currently sharing in their excellent Trade Secrets series. Their insights into Market Breadth and Volatility have been outstanding, and the knowledge they've shared is definitely improving my results. Those developments, coupled with the progress Nirvana is making in GA & AI, should make this an exceptional year. I’m looking forward to celebrating at the Bash… and happy that Ed is delaying it so they can concentrate on perfecting those enhancements and upgrades. Good luck in your trading… Mark [Edited by mholstius on 4/14/2019 9:04 AM] Attached file : 01 Margin results.png (136KB - 1544 downloads) Attached file : 02 IRA results.png (131KB - 1507 downloads) |
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
