Current location | Thread information | |
![]() ![]() ![]() ![]() ![]() ![]() |
Last Activity 4/21/2021 9:48 AM 46 replies, 10804 viewings |
|
Printer friendly version |
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
I have applied ATM Micro Macro to the bundle of Xsuite strategies. The first run was with all the default settings and all the strategies selected for each market state. The portfolio size is $40K no leverage long only. I used 5 years of data, exported the Port Sim results, and then ran an Excel pivot table to see what I might learn. The list was the first 300 stocks from the Omniscan list: My goal is to whittle down the drawdown to 10% or less. It is not lost on me that once drawdown exceeds 7% the percentage needed to recover starts getting exponential. The other factors that I personally desire are the gut check numbers. It is exciting to see the huge gains but if the drawdown is over 20% I just know I won’t be able to sit still. (I’m too old for fireworks!) I want to see the Calmar ratio greater than one and the max consecutive losses is another psychological number. I don’t think I would have the confidence to sit through more than about 8 consecutive losses either. The first result from Excel is: Excel detail including the strategies: Armed with this information the strategy selections are adjusted to eliminate negative returns from the strategy selections for each state. The ROI went from 130% to 192% while the drawdown went from 22% down to 13.5% and the Calmar ratio went from 0.63 to 1.31. This is a good start and is just a simple analysis based on the Excel result. I’m sure there are a lot more possibilities. [Edited by LSJ on 3/15/2018 12:39 PM] ![]() ![]() ![]() ![]() ![]() | ||
^ Top | |||
mholstius![]() Veteran ![]() ![]() ![]() ![]() Posts: 175 Joined: 1/13/2017 ![]() |
Thanks for sharing your results, Larry FWIW: To get the images into a post; At the point you want the image, insert; xxxx = paste the "copy Link Address" from right clicking on the image below the post. Mark [Edited by mholstius on 3/15/2018 9:54 AM] ![]() ![]() | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Thanks for the help. I used html tags but that didn't work. Trying to make it work. Did this once a long time ago and forgot ow i did it! | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Larry: So you have aroused my curiosity. How exactly did you obtain these quoted results?: Did you construct Methods for each of the individual Market States and then run a Portfolio Simulation for that individual isolated Method? If so it is a great paradigm for finding strategies in a given Market State that just are not working. Thanks, Tom Helget [Edited by EYEGUY on 3/15/2018 10:16 AM] ![]() | ||
^ Top | |||
mholstius![]() Veteran ![]() ![]() ![]() ![]() Posts: 175 Joined: 1/13/2017 ![]() |
Looks great - and good info... we all want smaller DDs! FWIW: If you're using snagit, it also helps some folks with smaller screens if you can resize the image to 750-800 pixels wide. Thanks for sharing your data! Mark | ||
^ Top | |||
mholstius![]() Veteran ![]() ![]() ![]() ![]() Posts: 175 Joined: 1/13/2017 ![]() |
Another hint on posting images... If someone has posted an image, in any thread and in any forum, you can also right click on that image and then use that copied link to insert that image into your current post... Might be handy if you have a question about someone else's chart, etc...??? Mark [Edited by mholstius on 3/15/2018 10:27 AM] | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
I found this to be really simple for my old brain. In Port Sim you view trades. Then choose export report and select .xlsx from the dropdown. In the new Excel file you find Pivot table in the menu bar. It then gives you choices to select what you want to see. You can then use the Sort function to group the results the way you want. This may just be a Rubic's Cube for trading results and I'd be interested if you found some useful variations. (Apologies to Mark for "Micro Macro" instead of the other way around.) | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Larry: For those of us who don't know about the Insert...Recommended Pivot Tables function of Excel here is what I did. First I exported the Trades in my Portfolio Simulation in .xlsx form - this is Data Only. Note that this data includes both the Market State and the Strategy: Then I opened the Excel file thus generated and highlighted all the data (this might be unnecessary but it worked). I went to Insert....Recommended Pivot Tables and chose one of the recommended formats and voila: What a truly great Excel function! Thanks for pointing it out to me. Tom Helget [Edited by EYEGUY on 3/15/2018 11:31 AM] ![]() ![]() | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Tom, Thanks for making that clearer for everyone. I am overflowing with ideas I want to try with all the great resources and the contributions by everyone, especially Mark's. | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Larry: Creating the second table you show with the MarketState and the Individual Strategies was a bit more difficult. First I copied the G/L($) Column so that it was contiguous with the MarketState and Strategy Column. I then highlighted the three columns (not shown). I chose Insert....Pivot Column and selected to display the resultant Pivot Table at $M$3 on the current page. I was then presented with a check box menu for the Pivot Table Fields as shown: I hope this helps anyone else who, like me, might not be familiar with Pivot Tables in Excel. Tom Helget [Edited by EYEGUY on 3/15/2018 12:08 PM] ![]() | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Tom, There is an easier way. Once the exported data is selected in the spreadsheet I entered "Pivot Table" in "tell me what you want to do" in Excel toolbar. I then just checked the items I wanted and it created the in a new sheet. [Edited by LSJ on 3/15/2018 12:34 PM] ![]() ![]() | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Larry: Yes, that is way easier! Trust Microsoft for thinking of everything. Thanks for the Excel "pointers". Tom Helget | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
I ran two different tests to try to glean the best strategies for the Method using Xsuite. The first test was using Mark's suggestion of testing a single strategy in one market state and testing each strategy in the profile to see and compare performance. The second test was as above where the trades were exported to Excel and a pivot table used to see the performance of each strategy. The results of the tests don't lead to the same conclusion. In the pivot table some strategies produce net losing trades while the same strategy produces a gain in the portfolio simulation. Please help interpret this result! ![]() | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Larry: Yes, that was the source of my confusion as well. I think the problem here is that the individual strategy is influenced by all that has gone on before it and after it. In isolation the strategy might perform just fine. But as a result of the Market State that preceded it more cash might be made available to it resulting in maybe more trades being taken than could not be "afforded" in the isolated state and some of those additional trades might lead to a catastrophic loss (that would have never occurred due to the lack of available capital to the isolated strategy). Remember that more available cash buys more trades and those additional trades have a lower Ranking Value and so might be considered less attractive. I hope maybe that makes some sense. Tom Helget | ||
^ Top | |||
mholstius![]() Veteran ![]() ![]() ![]() ![]() Posts: 175 Joined: 1/13/2017 ![]() |
I suspect the Excel pivot table results were with all market states trading at the same time...??? When using my one strategy and one state at a time method, you'll have to do each state and each strategy at a time and record the results. (9 states * X strategies). It gets tedious, but it also lets you see the equity curve for each state & strategy pair - and I like to see stability in that curve. Ending statistics alone can be misleading. I'd also suggest allowing the trades to stay open when the state ends. Hope that helps - I'll be gone for about a week. Hope someone can find the holy grail while I'm gone. ;-) Mark [Edited by mholstius on 3/17/2018 7:29 PM] | ||
^ Top | |||
Buffalo Bill![]() Legend ![]() ![]() Posts: 539 Joined: 10/3/2006 Location: Stafford, VA ![]() |
Another idea - use these OL systems into 2 confirm blocks in each strat. First one turn on the L&S Macro state you want to test, second the L&S Micro state. It will only fire when the Macro+Micro is in the correct state. You can test a bunch of strats at once for each of the 9 states then see which strats you like **Make sure BOTH confirm blocks are set to ref SPY as the symbol. For extra credit you can then run a couple more tests. Add a third confirm block, take out the SPY ref and set it to match the 1st block - so the SPY and current symbol are in the same Macro state. See if that helps. Add fourth block, take out SPY ref, to then have SPY Macro & Micro = Symbol Macro and Micro and see if that is better. Or maybe just symbol Micro = SPY Micro is best. [Edited by Buffalo Bill on 3/17/2018 7:49 PM] ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() ![]() | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Mark, I don't believe that is the case. The allocation was set to zero for all states but one - Default and Bull and that is the only state that was used for testing so far. The ATM methods were duplicated and one was used to test the strategies one at a time. The other was used testing all 5 strategies and the trades were exported to Excel for the pivot table. All trades were left open as well and the goal was Max avg Ann ROI/Avg Ann MDD. I did only test one market state because I was curious how they would compare. I want to understand why this difference before I do a bunch of tests the results of which I am not sure how to interpret. I will continue testing and update if I find an explanation. | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Bill - thanks for that. More great ideas for working with ATM and it is only getting started. | ||
^ Top | |||
mgerber![]() Member ![]() Posts: 40 Joined: 12/18/2003 Location: Issaquah, WA ![]() |
LSJ, thanks for your pivot table posts; potentially very useful! Looking at the simulation statistics in your original post, I am puzzled by one stat. The port sim run made money, but the Average Profit % shows a negative value. ANYBODY: How is that possible? Mark G. [Edited by mgerber on 3/18/2018 10:27 AM] ![]() | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Mark: Yeah there has to be something wrong going on here that Nirvana's programmers need to address. Tom Helget | ||
^ Top | |||
mgerber![]() Member ![]() Posts: 40 Joined: 12/18/2003 Location: Issaquah, WA ![]() |
Barry, Would you please directly address this? I have seen this happen several times before, just never got around to asking about the anomaly. --Mark G. | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Tom, That seems like a very plausible explanation. It was a small $40K 1x leveraged acct and maybe didn't have enough money. Back to the drawing board - thanks. Originally written by THELGET on 3/17/2018 8:19 PM Larry: Yes, that was the source of my confusion as well. I think the problem here is that the individual strategy is influenced by all that has gone on before it and after it. In isolation the strategy might perform just fine. But as a result of the Market State that preceded it more cash might be made available to it resulting in maybe more trades being taken than could not be "afforded" in the isolated state and some of those additional trades might lead to a catastrophic loss (that would have never occurred due to the lack of available capital to the isolated strategy). Remember that more available cash buys more trades and those additional trades have a lower Ranking Value and so might be considered less attractive. I hope maybe that makes some sense. Tom Helget | ||
^ Top | |||
EYEGUY![]() Icon ![]() ![]() ![]() Posts: 1543 Joined: 12/12/2003 Location: BALDWINSVILLE, NEW YORK ![]() |
Mark: OK, I performed your suggestion on one of my own Profiles with the Micro Macro technique. I then choose the better performing Strategies in each Market State: This gave the following Portfolio Simulation with way more Drawdown than I would like to see: Excel Pivot Table Analysis showed Strategies that one might have expected to perform well from the original Strategy Analysis going into the red: And, yes, these numbers are correct. Remember the Market State Totals are given at the top of their component Strategy List. Note that some unrecorded trades are still in play and so the Portfolio Simulation Final Equity disagrees slightly with the Excel Pivot Table Total. Tom Helget [Edited by EYEGUY on 3/18/2018 6:54 PM] ![]() ![]() ![]() ![]() ![]() | ||
^ Top | |||
LSJ![]() Legend ![]() Posts: 515 Joined: 8/17/2006 Location: Citrus Springs, FL ![]() |
Tom, I'm trying to understand all of this to see what it tells me. I added up the net gains from all 12 Bear Bear states and came up with 1,902,862. Looking at the Pivot Table output for Bear Bear the total is 9,096,448.114. I'm not sure why these are so far apart. My first assumption is that if I used the same data for the Port Sim and the pivot table that there should be close to the same result. Second, if through an analysis of the individual strategies I eliminated the negative performers and re-ran the Port Sim I should get a better result. By the same token I would expect a pivot table to verify that result with just minor differences. So, whatever is going on here I haven't completely grasped it. Seems like we're building railroad tracks from opposite ends and not meeting in he middle <:) Any light you can shed would be appreciated. | ||
^ Top | |||
Buffalo Bill![]() Legend ![]() ![]() Posts: 539 Joined: 10/3/2006 Location: Stafford, VA ![]() |
LSJ/Tom/Mark One thing about using ATM effected results to rank strategies for use in ATM is that the signals from the strategies are not raw they get ranked by ATM and only X amt of L or S get thru. ATM is directly effecting the results of the strategy - so was the strategy actually good (or bad) or was it ATM ranking and/or trade cutoffs that drove the results one way or another? IMHO using Port Sim is a better way to look at the over-all quality of the strats themselves in the Macro/Micro market (using the confirm blocks I provided) since it does NOT add the artificial, non-strategy-derived rankings or cutting off L&S signals arbitrarily at 2/day or something. If the sum total of a strategy's signals do well in PS - >70% HR, >1% PPT for ex - then things should only get better when ATM uses it's magic to rank those signals against other signals from other strats, etc. In addition, using ATM results to pick ATM strats could also be a form of curve fitting Food for thought I'm in the middle of running port sims on a long list of Nirvana provided strategies in each of the 9 Macro/Micro states (M&M states). This is Port sim data (FT = 1/1/2007+), no ATM, using the confirm blocks to set each MM state. I will clean up the results and post when I'm done FWIW. No pivot table but it will be an excel spreadsheet! Ha [Edited by Buffalo Bill on 3/19/2018 10:42 AM] |
|
Legend | Action | Notification | |||
Administrator
Forum Moderator |
Registered User
Unregistered User |
![]() |
Toggle e-mail notification |