Posts: 175
Joined: 1/13/2017
User Profile I’ve just returned from a family vacation and watched the recordings of the first 2 weeks of Jeff Drake’s “ATM Master Course”… and the course is incredible.
As I’ve said so many times before… if you don’t have ATM, you really should buy it.
If you DO have ATM, Jeff’s course is unquestionably the best way to get the most from your investment.
I thought I was somewhat familiar with ATM, but just the first 2 weeks of Jeff’s course have given me countless new ideas to improve my trading with ATM.
ATM is a revolutionary piece of software. With the addition of this outstanding course, and the combined efforts of the talented traders in this group, I’m definitely eager to see what we’ll accomplish over the next few months…
OK, on to the purpose of this thread - how to use ATM to improve our results with ATM. Trust me, it’s not as convoluted as it sounds.
Jeff B started a thread on the “No Trade Market State” introduced in Jeff’s ATM course. I asked him to follow some specific rules for setting up his ATM system so that he could generate the maximum number of trades. He then posted his export of the 38,135 trades generated and I imported then into the attached spreadsheet ("Analyzing Macro And Micro ATM Trades"). Thanks, Jeff...
His thread is here:
https://www.omnitrader.com/currentclients/otforum/thread-view.asp?threadid=15822&posts=16
These are the steps I shared with Jeff in his thread to obtain all the raw trades the system could generate;
1) Check “Use These Strategies” in each state and then go down and check all the Strategies in the list in every market state. That way, the strategy used will show up along with the market state at the time of the trade in the export of the trades.
2) Set the allocation in each market state to 2% / trade and 100 Longs and 100 Shorts.
3) Set the account starting balance to $1M. That way a 2% allocation will allow a trade in a symbol that trades at $20,000 / share and it won’t be missing out on high priced stocks.
4) Set the Leverage at 6X and check “Use leverage to Increase The Number Of Trades”. With this, you stand less of a chance of missing trades due to hitting the equity ceiling.
I do all this because when analyzing a system, I want to look at all the trades the system can potentially generate. I don’t care about ridiculously high ending equities, or big DDs. I want a large number of diverse trades to analyze before applying the excellent ranking and filtering capabilities of ATM.
The nice part is that ATM itself can be used to generate the trades and label them with both the Strategies and the Market State at the time of the trade - something we’ve never been able to do before.
I’ve plugged Jeff B’s trade data (38,135 trades) into the attached spreadsheet.
I’ll use it to show how the analysis can help us find potentially good pairings of Strategies with Market States.
Overview:
There are 3 tabs in the spreadsheet;
The core of the data processing is in the “Analysis” tab.
The Exported raw trade data is plugged into the “Trades” tab.
The Market States for each day are in the “States” tab (Macro & Micro daily states are included)
#1) I’ve repeated the group of Jeff’s 21 Strategies 9 times in rows 2 thru 192 and paired each group of 21 Strategies with each of the 9 Macro & Micro Market States.
#2) I’ve found that “Normalizing” the returns according to how many days were available to trade in a particular state is a useful way to compare the efficiency of the strategies (comparing Apples to Apples).
#3) You can select any range of dates to be evaluated.
#4) You can substitute your own groups of Strategies and Market States if you want to modify the spreadsheet to use it yourself - you just need to ensure that the formulas in the yellow shaded columns match the number of rows necessary for your own analysis.
The data can be sorted in numerous ways to help isolate possibly good pairings of Strategies and States.
This is a sort by Strategy;
This is a similar sort by State;
Can this method be used as a “double optimization” to find perfect combinations and ridiculous equity curves that probably won’t trade well at the HRE? Sure - but if done with some prudence, it can be a useful tool to improve results. The key is to input as many trades as possible using as many symbols as possible to get representative results.
In particular, if you just input the trades already selected using ATM’s filtering and ranking, then you will definitely be curve fitting. However, if you input a large, representative sample of a variety of diverse trades over time then this can definitely help find robust combinations that ATM will then refine.
I’ll close with 2 more plugs…
I heartily recommend Jeff’s ATM course being offered by Nirvana - the first 2 weeks have been extremely informative and helpful.
And as always, please consider the kids at Saint Jude Children’s Hospital;
https://www.stjude.org/
Thanks - and good luck in your trading,
Mark
I attached "Analyzing Macro And Micro ATM Trades Ver 09" and it now appears to work correctly without any errors on opening after downloading.
Not sure what was going on, but please let me know if you have any problems with it.
Attached file : 00 Overview.png (495KB - 1782 downloads) Attached file : 01 Sort Example 01.png (460KB - 1838 downloads) Attached file : 02 Sort Example 02.png (470KB - 1767 downloads) Attached file : Analyzing Macro And Micro ATM Trades Ver 09.xlsx (3782KB - 510 downloads)Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 262
Joined: 8/19/2005
Location: Boulder, CO
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 262
Joined: 8/19/2005
Location: Boulder, CO
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 175
Joined: 1/13/2017
User Profile I was distracted by some other issues yesterday.
First, I'll try to attach a newly saved version of the spreadsheet to this reply and see if it downloads better.
Mark
OK, I attached "Analyzing Macro And Micro ATM Trades Ver 09" to this reply and the original post and it now appears to work correctly without any errors on opening after downloading.
Not sure what was going on, but please let me know if you have any problems with it.
Attached file : Analyzing Macro And Micro ATM Trades Ver 09.xlsx (3782KB - 517 downloads)Posts: 175
Joined: 1/13/2017
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 262
Joined: 8/19/2005
Location: Boulder, CO
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 3022
Joined: 9/21/2006
Location: L'ville, GA
User Profile Posts: 1543
Joined: 12/12/2003
Location: BALDWINSVILLE, NEW YORK
User Profile Posts: 3022
Joined: 9/21/2006
Location: L'ville, GA
User Profile Posts: 175
Joined: 1/13/2017
User Profile Sorry, but I need to add some clarity to this, as it doesn’t apply in Omnivest. In my Elite FTM portfolio, we regularly have multiple trades open in 1 symbol at IB that are controlled by multiple trade plans, with entries and / or exits on different days.
Here’s a quick snag I made of 5 different positions in the live account. In each case, the 5 concurrent trades in each of the 5 symbols were controlled by 5 different strategies / trade plans;
No criticism intended, just want to clarify that it happens regularly in OV coupled with IB. I'm adapting to OT rules...
Mark
Attached file : Multiple Trade Plans.png (404KB - 1506 downloads)Posts: 3022
Joined: 9/21/2006
Location: L'ville, GA
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 3022
Joined: 9/21/2006
Location: L'ville, GA
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile The prior post has a link to my google drive if you'd like the full spreadsheet.
7-zip files can be opened with most archive tools -- Peazip is one that I use on Windows. I tried straight zip but it was close to the 7MB limit.
I'll put the zip file version on my google drive also for those that don't want to deal with 7zip.
Attached file : Analyzing Macro And Micro ATM Trades Ver 09-Revised.csv.7z (3429KB - 495 downloads)Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile A word on the strategy names in the spreadsheet:
1) Strategies beginning with ATM- mean I use it with an ATM method. I do this so I don't accidentally change it. It also groups them together in the strategy list.
2) MOC in the name indicates that the orders are Market on Close.
3) In in the name indicates that it has an inactivity stop.
4) The numbers following In are the values of the parameters for the Inactivity stop. For testing, I wanted to see if different values improved or hindered the effectiveness of the strategy.
With the above naming convention, I'm able to perform an additive analysis of the base strategy against the modifications. One note. You will not see a strategy consisting of the Base + Inactivity. This is because I only added the Inactivity stop on top of the MOC. Why? Because that's what I did -- just that simple.
The Power of the Analysis Spreadsheet
Maybe this is obvious. Let's compare strategies in different market states:
Bear and Bear Market State
Bear and Bull Market State
Here we have a Bear and Bear MS vs a Bear and Bull MS. Looking at the CRT-3 strategy and its derivatives, we can see that in the Bear and Bear MS, the derivatives outperformed the base strategy. However, in the Bear and Bull MS, two of the derivatives marginally outperformed the base.
We can also observe that in both MS, 2 of the derivative strategies produce the same result. Do we need both derivatives? Why did I create two? I actually did it because I was optimizing the MOC version using Strategy Wizard 4 and in the Bull Default MS, the performance is different.
But, if we observe the results, we can actually see that the base strategy (CRT-3) outperforms all derivatives. So, without this analysis spreadsheet, I would have likely ignored the base strategy as a candidate because I was looking to establish a MOC base set. Am I over analyzing? Am I over optimizing? Is over optimizing bad? Personally I get into a pattern of seeing something and trying to make it better. Once I find something better, then I assume it will always be better -- that's my flaw and the spreadsheet highlights it.
This raises a question... If the base CRT-3 strategy outperforms the MOC version, how would I trade it using AutoTrade and ATM -- using a mix of Market and MOC strategies? I don't know the answer to this question yet. The challenge is that AutoTrade runs at a specific time. But more importantly, when OT does its analysis and simulation to produce the trades, what price did it use? Was it the market open price on the day following the signal? I don't know. I do know that for MOC, OT uses the close price for the bar. Since I have a difference in time of execution and AutoTrade only processes once a day, how can I solve this? I suspect I could run 2 instances of OT, splitting my ATM method apart into 2 -- one containing the MOC strategies and one containing the Market strategies. In this manner, I could run the analysis at the appropriate time. Perhaps that will work. Anyone have an answer?
Okay, the above is a perfect reason for me to use this analysis technique -- tuning the strategy to the MS and checking that I didn't over analyze and over tune.
One last exciting feature of this spreadsheet... the date range. Did you see it?
Change the dates and all the stats are updated! If I do that, do the same strategies that I saw as strong over the long run hold up in the date ranges I may be concerned about? Or how about the last 3, 5, or 10 years? Which strategies continue to prove strong? That's powerful!
As you can tell, I'm a fan of the spreadsheet. Mark, thanks for sharing it with the community!
Attached file : Results Only Macro And Micro ATM Trades Ver Brown09.xlsx (42KB - 352 downloads) Attached file : BearBearCRT-3.png (28KB - 1049 downloads) Attached file : BearBullCRT-3.png (28KB - 1070 downloads) Attached file : BullDefaultCRT-3.png (29KB - 1065 downloads) Attached file : DateRange.png (6KB - 1023 downloads)Posts: 175
Joined: 1/13/2017
User Profile Posts: 175
Joined: 1/13/2017
User Profile This spreadsheet was created to simplify the process, and make it valid by not just curve fitting to the best outcomes already calculated by ATM.
However, there’s a problem in obtaining all the trades possible for every State & Strategy pairing in order to make a fair, valid, and useful comparison of the relative performance of each pair.
Let’s say there happened to be an excellent trade in Apple. OT only allows one trade in a symbol / day. Even though the trade would be taken by all the other Market States or Strategies it would only be credited to the first in the list - and all the rest of the States or Strategies wouldn’t get credit for the potential of that trade in the exported trades.
To resolve the problem, we have to individually pair each Market State with each Strategy, run a basic ATM Analysis with few restrictions, and export each set of trades. Thanks to ATM, each trade will be labelled with the Market State and Strategy used for the trade. We then combine all of the trade exports and import that large data set into the trades tab in this spreadsheet. It’s admittedly a tedious process at the moment, but essential to obtain the unbiased comparison we need.
To demonstrate the difference, Jeff’s original export of his ATM Method containing the 9 Macro & Micro Market States and associated Strategies had an export of 38,136 trades.
When he took the time to individually pair each State with an individual Strategy, the combination of all his trade exports contained 288,371 trades - almost 8X the number of trades.
This extensive data set now allows a valid comparison of the raw data, pre-ATM performance of every State and Strategy pairing. It allows us to find the best pairings to feed into ATM while avoiding the curve fit solution of “picking the best” after ATM runs.
Thanks again for your work obtaining the data, Jeff…
The latest version of my Analysis spreadsheet is too big to post (~40MB), but can be downloaded from the following Dropbox link;
https://www.dropbox.com/s/eez2ohp4rn9i1ln/Analyzing%20Macro%20And%20Micro%20ATM%20Trades%20Ver%2040.xlsm?dl=0
After clicking on the link, you should be presented with an option to download it by clicking on the “ … “ at the top right of the screen;
I added a number of improvements, using Jeff’s 288,371 trades for development.
It now has 3 major sections. Each can be sorted to find the best opportunity for any selected date range.
It analyzes every combination of State and Strategy and calculates the following statistics for your chosen date range;
% Sum = total return for that pairing of State and Strategy during the period
# Days In State = # of days the state (ex: Bear and Bear) was active during the period
% / Days In State = a normalization based on the % Sum / # Days the state was active during the period
# Days In Trade = market Days the State & Strategy pair had open trades during the period
% / Day In Trade = normalization based on efficiency (esp if comparing short and long term strategies)
# Trades = all trades for that State & Strategy pairing during the period
Trades / Day = # Trades / Total # days during the period (would be 4,855 days from 1/1/00 to present)
PPT = during the period for that pairing
HR = during the period for that pairing
Every Market State is analyzed in a similar manner for Long, Short, and Both performance;
Every Strategy is analyzed for Long and Short performance;
This is an overview of the Analysis tab;
Long trade data is automatically highlighted in green, short data in red, and every section is individually sortable to help identify the best opportunities.
I’ve added a “Setup Help” tab that uses a Macro to find and identify all the Market States and Strategies available (Long & Short) in the raw trades entered on the Trades tab. This eliminates the tedious process of identifying them using sorts, etc., for a large number of trades. The “Setup Help” tab also gives instructions on how to set up ATM to obtain the trades, and where to modify the spreadsheet for your data, making the entry of data much easier than in the original spreadsheet.
IMPORTANT: the spreadsheet was set up for Manual Calculations to make it easier to modify during development, since it takes a while for all the calculations to run.
You need to go to "Formulas / Calculate Now" to make it calculate - or if you want it to calculate automatically, you need to change to Automatic Calculations in your excel preferences.
Please pass on any suggestions for improvements you might have, or any bugs you might find.
ATM opens up so many possibilities for adapting to changes in the market.
I’m looking forward to the third week of Jeff’s course, “Matching Strategies To Markets”, this afternoon.
Mark
Attached file : 00 Dropbox Link.png (99KB - 1013 downloads) Attached file : 01 Basic analysis of pairs.png (189KB - 1010 downloads) Attached file : 02 State analysis.png (333KB - 1008 downloads) Attached file : 03 Strategy analysis.png (459KB - 1039 downloads) Attached file : 04 Analysis Tab.png (721KB - 998 downloads) Attached file : 05 Setup Help Tab.png (294KB - 1021 downloads)Posts: 515
Joined: 8/17/2006
Location: Citrus Springs, FL
User Profile Posts: 262
Joined: 8/19/2005
Location: Boulder, CO
User Profile I'd like to share my approach to getting the data to use in Mark's spreadsheet. His spreadsheet uses the %G/L of each trade, and the sign (+/-) from the # of shares. So, I'm using the following process to generate and gather trades by Market State, where each trade is just 1 share.
I created a set of ATM methods - one for each market state, and selected all of them for processing:
In each ATM method, I configured the Allocation tab with "Fixed Trade Size" and 1 share per trade, allowing 100% long, 100% short, 100 long trades and 100 short trades (this should be increased if you have 1000's of concurrent trades). In this example, I am configuring the "#ATM M&M Bear and Bear" method. So, ONLY the "Bear and Bear" market state TradeSize is set to 1. For all the other market states in this method, the TradeSize is set to 0, so that trades are gathered only for the "Bear and Bear" state. Repeat this process for each of the market state methods.
In all the methods for all the market states, "Use Portfolio Simulation Settings for Strategies" is selected for all the market states. NOTE: The strategy to be processed is controlled by the PortSim "Trade Source" setting.
In the PortSim Settings "Trade Source" tab, select the Focus List and Strategy to be analyzed:
In the PortSim Settings "Account Settings" tab, make sure you have sufficient funds allocated available to purchase 1 share of all possible in-trade symbols. If you have thousands of potential concurrent trades (assuming you have a very large number of symbols in your focus list), then make sure you set this to a high value, e.g. $1M. And, configure "Increase Number of Trades" so that all potential trades are included in the analysis.
In the PortSim Settings "Trading Parameters", I configured Min Shares and Max Shares to 1, and "Per Trade Allocation." (Frankly, I'm not sure whether these settings matter in this scenario, but it doesn't hurt to set them.)
Since this process is designed to only trade 1 share, I do not want commissions to be a factor in the analysis. So, I set commissions to $0.
Run the simulation, and at completion you will have a set of "Strategy-by-MarketState" trades for inputting into Mark's spreadsheet.
Attached file : 2018-08-17_17-14-06 ATM Market State Methods.png (7KB - 932 downloads) Attached file : 2018-08-17_17-21-15 PortSim Settings Trade Source.png (13KB - 919 downloads) Attached file : 2018-08-17_17-23-18 PortSim Settings Account Settings.png (12KB - 927 downloads) Attached file : 2018-08-17_17-24-04 PortSim Settings Trading Parms.png (19KB - 926 downloads) Attached file : 2018-08-17_17-25-05 PortSim Settings Commissions.png (18KB - 930 downloads) Attached file : 2018-08-17_17-19-36 Method Strategies.png (41KB - 904 downloads) Attached file : 2018-08-17_22-56-51 Method Allocation.png (102KB - 919 downloads)Posts: 262
Joined: 8/19/2005
Location: Boulder, CO
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 175
Joined: 1/13/2017
User Profile Thanks for your suggestion of an easier way to cycle thru the various strategies to get the trades data.
I’d forgotten about using the Simulation Settings to pick the strategy being used - and that can be a big time saver (I need to do a couple hundred runs for my project).
On the serendipity side, I started using your method this morning - and it helped me find a good way to illustrate and document a small discrepancy in the trades exported when using concurrent states that I’ll be sending to Barry.
Since Nirvana can’t address anything until Monday, please make the following adjustment to your method to be sure you’re getting accurate trade data;
Rather than just setting the Default Fixed Trade Size to zero, also set the Symbols tab so that the Market State you don’t want to trade has no symbols in it…
In my testing this morning, I found that this will avoid the discrepancy I’ll be documenting for Nirvana.
Thanks for your input - we’ll move the ball forward faster working together.
Mark
Attached file : symbols to zero.png (74KB - 831 downloads)Posts: 3022
Joined: 9/21/2006
Location: L'ville, GA
User Profile Posts: 175
Joined: 1/13/2017
User Profile Posts: 6
Joined: 4/30/2006
Location: Longmont, CO
User Profile Posts: 168
Joined: 5/11/2005
Location: Brown Deer, WI
User Profile Posts: 175
Joined: 1/13/2017
User Profile There are so many ways to use ATM, and we’ll undoubtedly improve the results even more in the future.
For Jim (and anyone interested)..
I've attached a file to this reply with all the Macro & Micro market states.
It's also color coded a bit to help see the formulas as they're combined.
Good luck in your testing,
Mark
Attached file : Macro and Micro market state formulas.rtf (4KB - 323 downloads)