mholstius Veteran   Posts: 175 Joined: 1/13/2017  User Profile | I totally agree with both you & Ken in that MOO orders aren’t ideal, and it took me some time to put together the following test that might help with the discussion. I used a data set of 48,014 unique long trades from 1/1/2000 thru 8/14/18 generated by a variety of RTMs (CRT3, NSP-41, RTM7, T3-S3, VBX-3, XLS-19, XMFI, XULT) on 371 symbols. I used the MarketXLS extension for Excel to download the closing prices on the signal days of each trade (one day prior to the entry days and one day prior to the exit days). I calculated the return for each of the 4 possible combinations of MOO and MOC entries and exits and tabulated the results (thereby avoiding the effects of compounding returns); Column X calculates the difference in % sum of each combination vs the normal MOO entry and exit (#1). Columns Y-AB show how often a combination was either the best or worst outcome for a particular trade. The combination with the best return (#3: MOC Entry & MOO Exit) was the best choice for 32% of the trades and was the worst choice for 19% of them. The 2 pairs with MOC entries (#3 & 4) produced better results, but it’s interesting to note that MOO exits produced better results - when paired with either MOO or MOC entries. We can all remember times when we wished we’d gotten out of a trade the day before, but a MOO exit the next day was more beneficial in this set of 48,000 trades. Given all that, there are some serious practical problems with MOC orders in an autonomous system. With IB, a MOC order must be sent at least 15 minutes before the close, so you’d have to start the analysis at some point earlier based on the complexity of the system. Even a relatively simple system would probably need to be started 45 minutes before the close, and I really don’t know how much time would be required for something like my FTM portfolio containing 9 dynamic lists. Maybe someone who’s using MOC orders can comment on their experiences? We all know that a lot happens in the 30 minutes before close. I think it’s reasonable to expect that making a decision to enter a trade 30-45 minutes before the close would generate some false positives: trades that wouldn’t be taken by the system after analyzing all the data in the evening. Those trades (entered erroneously MOC) would then be “orphaned”, have to be managed manually, probably lead to difficulties in calculating equity by the system, and not be beneficial in an automatic system. Back in August, 2017 I investigated a compromise: let the analysis run in the evening and then take a smaller 1% allocation MOO entry for all trades along with a larger 10% Day Limit order set at the previous Close minus 2%; This would avoid the problem of orphaned trades and appears to hold promise, but would need to be investigated in more depth. Here’s a link to that post in the OV forum with much more detail on the subject; https://www.omnitrader.com/currentclients/omnivestforum/thread-view.asp?threadid=7845 Further down in that thread, I analyzed using no MOO orders, but just 2 Limit orders; Fewer trades and better results, but like I said - this would need to be tested some more, and the implementation might be difficult. I’m also not trying to detract from what you’ve documented so well, Steve, but can I suggest some further investigation into your results to see if we can extract more information? Your Market Orders info shows a MDD of 75.4% and 4,289 trades. Your MOO Orders info shows a MDD of 88.3% and 12,622 trades (considerably more trades). Maybe there’s something going on that's not obvious from the charts…??? Anyway, I hope this discussion will yield some more information as people (hopefully) respond with what they’ve experienced trading with different order types. Mark [Edited by mholstius on 11/27/2018 11:22 AM]  Attached file : MOC v MOO results.png (76KB - 1589 downloads) |
