OmniVest User Forums
General Discussion Topics
RTM Entry Allocations
6 replies, 1326 viewings
|
OmniVest User Forums
General Discussion Topics
RTM Entry Allocations
|
Last Activity 4/15/2019 7:43 PM 6 replies, 1326 viewings |
|
|
|
| Printer friendly version |
Mark Holstius Elite    Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL  User Profile |
For this thread, I’d like to get some feedback about modifying the entry protocol for RTMs. I’ll begin by acknowledging that the data is limited to the trade data I have, and my observations may not apply to the universe of RTM trades (a curve fit?) - but the concept I propose passes my “gut check” for reasonableness. The following chart compiles a lot of data from many spreadsheets. I hope it will address a question that’s been around since Nirvana introduced OV and the RTM Strategies. Would we be better off if we could enter / exit trades MOC on the day that triggers a trade? I think we can all acknowledge that’s not possible. Dynamic lists need to be run and price movement in the last 15 minutes of trading often affects the decision process - but if we could do it, would it help? Using the data from my 15 years of trading - yes & no. The chart below contains 3 plots of Closed Trade Equity curves using 4% allocations and my FTM portfolio trade data. It’s NOT the standard OV MTM equity graph; The dotted Green line is what the equity would’ve been if you were able to enter the trade MOC on the afternoon that triggered the trade. Ex: Enter on Monday’s Close (vs Tuesday MOO) and exit Friday MOO per normal OV. The solid Yellow line is the closed trade equity over the past 15 years for my FTM portfolio. It uses the current MOO OV protocol for both entries and exits. Ex: Price action on Monday results in a decision to enter a trade MOO on Tuesday morning, and the data on Thursday triggers a MOO exit on Friday morning. The dotted Red line is what the equity would’ve been if you entered the trade MOO on Tuesday and were able to exit the trade MOC on Thursday (vs Friday MOO). Both the Green and Red lines convey similar information: on average, the Opening price of the stocks traded by RTMs is higher than the previous day’s Close. If we were able to enter at the close of the previous day, on average we’d get a better entry price & see improved results. The exits are the same for the Green and Yellow lines (MOO), so the MOC entries must be lower. If we were able to exit at the Close of the previous day, on average we’d get a worse exit price & see worse results. The entries are all the same for the Red and Yellow lines (MOO), so the MOC exits must be lower. Green PPT: 1.62% Yellow PPT: 1.50% (Normal OV) Red PPT: 1.27% The chart below adds a Magenta line showing what would’ve happened if we were able to both enter Monday MOC and exit Thursday MOC (PPT 1.40%); Interesting observation - the OV protocol of MOO entries and exits has been better than the theoretical MOC entries and exits combination since 2009. Currently, we can only do MOO trades with OV, but I have a proposal that might be possible with some changes to the trade processor… What if we were able to modify the entry side of the protocol to benefit from this data that shows the Close of the day before is better on average than the Open the next day? I calculated the results keeping the MOO exits in the data but substituting trade entries in 3 different groups, cross-checking whether they had been possible given the historic price action on the entry days; A) Enter at the Open if the Open is < the Previous Close (this should get the lowest entry $) B) If the Open is >= the Previous Close, enter during the day using a Limit Order equal to the Previous Close. C) If neither A or B was possible, enter with a MOC order. Here is a table of the results; A) ~40% of the trades were entered at an Opening price less than the Previous Close (PPT 1.83% / HR 79%) B) ~40% would’ve been entered during the day with a Limit Order equal to the Previous Close (PPT 1.44% / HR 77%) C) ~20% were entered as a MOC order because neither A or B were possible during the day (PPT 0.74% / HR 66%) As expected, the best performance is with the lowest entry price (A). I appreciate that what follows is a curve fit, using the information I already have about the trades to determine how much to allocate - but it makes sense at the “gut level” even without that information. I propose to vary the trade allocation at the trade entry as follows; A) If the Open is < the previous Close, allocate 6% to a trade entered at the Open. B) If the Open is >= the previous Close, enter a Limit Order at the Previous Close and allocate 4% to it when / if it fills during the day. C) If the limit order isn’t filled during the day, allocate 2% to a MOC order. The following chart has a Teal colored equity line that shows what would’ve happened using those 3 entry allocation rules; The results since the recovery in 2009 are even better than Entries at the Previous Close, so I’ll bring up my question... Would it be more probable to obtain a lower entry price at the Open using option 1 or 2 below? 1) Wait for a few seconds after the Open to see if the Open is below the previous Close, then enter a Market Order at 6% allocation to get as good a fill as possible (hidden order). If the Open is >= the previous Close, then enter a Limit Order at the previous day’s Close setting a 4% allocation and wait for a fill during the day. 2) Start with a Limit Order at the previous day’s Close using 6% allocation (visible order). If the stock has downward pressure (Open is below the previous Close), hope that our order doesn’t have a large upward effect on the price and that we get a fill within a few seconds that’s below the Previous Close limit price. If not filled after a short time (Open >= Previous Close, etc.), change the allocation to 4% and let the Limit Order at the Previous Close run for the rest of the day. If there was no entry by the end of the day, enter a MOC order at 2% allocation and accept that price. If the Open is going to be lower, either a MOO order or a Limit order should have the effect of raising the Opening price (at least slightly) - I’d just like some opinions on which might be better. There are benefits and problems with either 1 or 2. With #1, the order would be hidden until the market opened (no effect on price), but would entail calculations to decide whether to enter a Market Order at 6% allocation or a Limit Order at the Previous Close using 4% allocation X seconds after the Open. With a visible Limit Order set to the Previous Close (#2), the question is how much we would “drag up” the Opening price if it were going to be lower. It would be simpler, and the only decision would be to change the allocation to 4% X seconds after the Open if it wasn’t filled. Thanks for your input, Mark [Edited by Mark Holstius on 7/30/2017 8:24 PM]  Attached file : 01 Equity Curves.png (74KB - 642 downloads) Attached file : 02 Equity Curves.png (86KB - 659 downloads) Attached file : 03 Results Table.png (13KB - 612 downloads) Attached file : 04 Equity Curves.png (88KB - 704 downloads) | ||
| gbarber Veteran Posts: 282 Joined: 12/30/2012 Location: Pearland, TX User Profile |
Perhaps it would be worthwhile for N to do that. It seems to me if these results can be generalized and the implemented in OV, it would be a great advertising feature that could draw more customers. I have one question. Do you have intraday data? It seems that you do since you made the statement that if a limit order is used you plotted the equity curve "when / if it fills during the day". Unless I interpret this wrong, you must have data to determine whether the limit would be filled. If you do have such data, why not also compute a curve that shows what would happen if a market order was placed when the trigger existed at 15 minutes before close. That could be achieved with current TP. I'll take a stab at an opinion on your questions at the end of your post regarding how hidden and visible orders might behave. Human nature says the visible order would push the price up. Probably to the limit price. But there may be factors that make that action vary depending on those factors. The size of your order versus the typical volume is likely going to make a difference. One could use that info to make a decision (but then one would also have to know how many people are following your trades and how big they make their trades). Market makers know the answer to that question. If anyone has access to one or more market makers, perhaps they would answer that question. Or the answer may be in one or more books written by market makers. It is also possible the answer could vary among different market makers. That's my guess. I could be way off the mark. | ||
| Jim Dean Elite  Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
I know limit orders have been requested and discussed a few years now, have I missed something? Does OV currently support them? If not, is there a firm date for that to be implemented? Further note - I'm not at all surprised that MOO is better than MOC for N RTM strat Exits, since as I recall, most or all of those strats have arbitrary (presumably tuned) N-bar exits (5 bars, etc). Supposedly, RTM trades are in the direction of the prevailing trend, which often might last many more than just five bars past the spring back reversal. So, it makes sense that shortening the duration of the trade (MOC) would on average reduce profits. Mark this is excellent work, with a statistical basis that's more robust that any other proof-testing I've heard of for N strats. Sure more could be done but imho this is meaningful as is. However, if Limit orders aren't supported yet after all the years of requesting them, isn't the proposed solution sort of on the wishful thinking side? Don't get me wrong, I'd love to see entry limit orders AND exit stop orders be implemented. [Edited by Jim Dean on 7/30/2017 11:12 PM] | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
Gary; “all” I have is the OHLC data I’ve downloaded for the ~9,400 trades over the past 15 years. As Jim said, for the limit order I just check to see if that price would’ve been hit on the day by checking the range for the day. Unfortunately, I don’t have any way to get that amount of intraday data… Jim: thanks for the kind words. No, there’s no current capability for using Limit orders in OV. I hope this extensive “proof of concept” study might have a positive effect on that…??? If limits, stops, and intraday exits could be implemented in OV simulations we could verify whether the idea is robust or just a fluke of history with my trades. As you know, the problem with OHLC data is that you can’t know the sequence of the prices, so you can’t tell whether a stop or profit target was hit first during a day with a range that covered both. Intraday data could solve that, but that’s another level… I did some experimenting with profit targets and stop limits for each of the 9,400 trades. In my preliminary runs they both showed a decrease in overall profit & HR. I hope to refine the study over the next week or so and post some more about it. Mark | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
Sequence of OHLC points re entry limits and exit stops (but not entry stop-limits) isn't important unless your trade plan permits entry and exit on same day, so testing is legit for OV processing. I can offer a lot of suggestions re alternative stop logic that would be worth testing, given that you have the trade and price history. Call to discuss if you'd like. [Edited by Jim Dean on 7/31/2017 6:10 AM] | ||
| gbarber Veteran Posts: 282 Joined: 12/30/2012 Location: Pearland, TX User Profile |
a) I recalled that Jeff Drake already did quite a lot of testing on MOC orders. He found that there was improvement with various RTM strategies including RTM-7, VBX-3, swing 13, and XLS-19 V2. I think he was using the S&P100 as the list. b) I found an article about market makers actions. https://www.fidessa.com/document/4028?ajax=true This article indicates there was a significant change in 1997 to the rules NASDAQ market makers have to follow called Order Handling Rules. "The first rule was the requirement for market makers to display the size and price of client limit orders to the marketplace, known as the Limit Order Display Rule. According to the SEC, this rule gave “investors the ability to directly advertise their trading interest…. enabling them to compete with market maker quotations and narrow the size of the bid-ask spread. The second rule was the requirement for market makers to display and honor quotes from dealers on the ECNs, a mandate known as the revised Quote Rule. In other words, previously undisclosed (and more competitively priced) orders on ECNs would now be published and, in the case that the quotes were more competitive than their own bid-ask spread, the market makers would have to honor those dealer’s prices." Under these rules the market makers ability to control spread is not as drastic as I had speculated. So maybe your option 2 would work better than 1 because you still have a good chance of getting a lower price if it exists at market open and you have a limit order so you don't have risk of taking the trade at a price higher than you wanted. | ||
| Steve2 Elite Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD User Profile |
Limit orders have been on the feature request list for years. I did some limited experiments with a similar strategy several years ago (don't know whether I still have the spreadsheets). The strategy I modeled was to submit a day-limit opening order at the previous trading day's closing price. If the order didn't fill (at open or during the day) then the trade was abandoned. Closing orders were standard next day MOO orders. This resulted in improved results but I used a much small set of trading data than you have. Subjectively, I think we would see much improved results if OV provided better capabilities for managing entries and exits. I would be in favor of making this high priority as long as whatever capabilities are provided are also supported for backtesting. Steve |
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend(un)/Freeze thread | |
Toggle e-mail notification |
