Chat Session Logs

4:29 pm:<LS_Bill> hello everyone
4:30 pm:<Brian> Hi Bill
4:30 pm:<LS_Bill> have a change to play with beta 2 yet Brian
4:30 pm:<LS_Bill> chance
4:32 pm:<Brian> looking at optimizer at the moment
4:32 pm:<rg> hi Bill, Chad, all
4:32 pm:<Brian> too many goodies ;-)
4:32 pm:<Brian> goodies haha
4:32 pm:<LS_Bill> optimization is a real goodie
4:33 pm:<rg> yes, very nice
4:33 pm:<rg> care for a suggestion?
4:33 pm:<LS_Bill> always ;-)
4:33 pm:<rg> ;-)
4:33 pm:<rg> I'm sure you've already got this on the list
4:33 pm:<rg> future version...
4:33 pm:<rg> ability to "click" on one of the iterations (in the list of optimization results) to drill down into its detail
4:33 pm:<rg> ;-))
4:34 pm:<LS_Chad> Interesting idea...however...
4:34 pm:<rg> heh heh... no mean feat, i realize...
4:34 pm:<rg> i really like what you've done so far
4:34 pm:<LS_Chad> as it currently works, you can just run the trading system right after the optimization is through...
4:34 pm:<rg> yes, of course
4:34 pm:<LS_Chad> and the optimized values will be left filled into the V# variables
4:35 pm:<LS_Chad> so you should then see a full report of optimized results
4:35 pm:<rg> oh, are you saying the top-most optimization in the list remains for a "manual" backtest?
4:35 pm:<LS_Chad> right...
4:35 pm:<rg> interesting... i will try that, thank you, sir :)
4:35 pm:<LS_Chad> BUT...this only applies to the V# variables...not the periodicity...currently...
4:35 pm:<rg> ok, the latter is easy to set
4:36 pm:<LS_Chad> you would have to adjust the periodicity manually...if that was an optimizing parameter
4:36 pm:<rg> got it
4:36 pm:<rg> thanks, Chad, awesome work on this :)
4:36 pm:<LS_Chad> when I say "optimized" values...it just means the ones that were shown at the very top of the resulting list/table
4:36 pm:<LS_Chad> based on sort criteria...
4:36 pm:<rg> yes, that's good... that's the one i'd want to look at first anyway ha ha
4:36 pm:<LS_Chad> which in general, would be the most likely candidate that you would want more info on
4:37 pm:<F_trader> hi guys :)
4:37 pm:<rg> hi
4:37 pm:<LS_Chad> I could go ahead and adjust the periodicity...but not sure all users would want the optimization to actually manipulate a trading system variable like this
4:37 pm:<LS_Chad> would be easy enough to do though
4:38 pm:<F_trader> Bill, for info, the notes.d01 file removal procedure (and replacement by an empty one) did not work
4:38 pm:<F_trader> * set SL_SYM_T_NOTE(28) error:
4:38 pm:<F_trader> member has null previous-pointer but is not first member
4:38 pm:<F_trader> was the error message
4:39 pm:<rg> chad, I've no strong thoughts about it either way... ultimately a click on a line in the report to set all the variables, but it's really wonderful right now so i'm just going to be happy.
4:39 pm:<LS_Chad> yea....that would be nice....I'll think on that one...not real easy for me to pick up on what iteration a user is double-clicking on in a text document
4:39 pm:<LS_Bill> eddy, this will require a utility function in the database utilities menu to "Initialize Notes Database"
4:40 pm:<Brian> Chad, yes I think I'd rather make the final choice after looking at results -- maybe not top of list -- less drawdown etc
4:40 pm:<rg> yes, it might eventually have to be a list of some sort, not simply a text document
4:40 pm:<rg> so that it's selectable.
4:40 pm:<LS_Chad> One question regarding the report display...if a user has given a user variable a title...would you prefer to see that title as the column heading in the report...instead of seeing just V#1?
4:41 pm:<rg> But the basics are very solid and my suggestion was not meant as a criticism.
4:41 pm:<F_trader> ok Bill
4:41 pm:<LS_Chad> right.....however, there can be unlimited iterations, millions, as you know...so..could be rather complex
4:41 pm:<LS_Chad> I understand...sure
4:42 pm:<rg> That probably makes the most sense, though since i write these things for myself, i generally know what they are... of course, when my systems become as complex as Eddy's labeling the variables will probably be required!
4:42 pm:<LS_Chad> I've added an optimization button the the trading system window that will allow you to quickly create an optimization out of an existing system....defaulting the name to the same name as the system, but immediately giving you a chance to change it
4:42 pm:<Brian> yes to title name
4:42 pm:<Brian> nice
4:42 pm:<rg> that's good... first time i had to go hunting on the main menu to find it... wondered about that with all the open space on the backtest page
4:43 pm:<LS_Chad> OK..I'll start showing the title....(cutting it off after 8 characters however for spatial considerations)
4:43 pm:<LS_Chad> Also added a button to allow you to quickly access the User Variables window from the Optimization window
4:43 pm:<F_trader> Chad : could we use a V variables (not one used for iteration) as sorting criteria ?
4:44 pm:<rg> um... if 8 characters are in lieu of V#xxx then I probably need to stay with the numeric portion... most of my variables are assigned names that are much longer than 8 characters, with a hierarchy such as aaaaa_bbbbbb_cccccc
4:44 pm:<LS_Bill> nice idea eddy
4:44 pm:<rg> run the risk of all of them reading aaaaaaaa..
4:44 pm:<rg> and i won't be able to tell any of them apart!
4:44 pm:<LS_Chad> well..if you had a bunch of 32 character titles....it would eat up the report header....
4:45 pm:<rg> i agree with you
4:45 pm:<LS_Chad> so I'd have to cut them off if I do it that way...
4:45 pm:<F_trader> the idea behind giving full flexibility in defining our own sorting criteria...
4:45 pm:<LS_Chad> personally...I think just putting V#1 in there is a little cleaner
4:45 pm:<rg> probably needs to be an option, chad -- those who define all their V# within 8 chars will be happy, i'll have to use the V# designation due to my naming convention.
4:45 pm:<LS_Chad> That could probably be done Eddy...I'll put on list...
4:46 pm:<Brian> would it make any sense to have a separate group of variables like O#23 for optimization so as to keep them separate and not have to think about the Context you discussed?
4:46 pm:<rg> which is okay by me since i only write for myself.
4:46 pm:<rg> there's a good point, Brian
4:46 pm:<F_trader> that opens quite nice flexibility.....
4:46 pm:<F_trader> PS : i would like to keep as well V#165 in report
4:47 pm:<LS_Bill> you could title the V# "165whatever"
4:47 pm:<LS_Bill> so you know the number and some text
4:47 pm:<LS_Bill> so you see
4:48 pm:<LS_Chad> I'm leaning towards V#165 myself...but may make it an option...
4:48 pm:<F_trader> that is what i do already on my QP, so that even with min colum size, i may know who is who :-)
4:50 pm:<rg> i might have to just expand my variables and just set aside a range for optimization and keep the names small; that will work.
4:51 pm:<F_trader> Derek, was your issue with signal marker on tick charts solved ?
4:51 pm:<LS_Chad> Eddy...thinking more about you suggestion to sort on a non-optimized V# variable....doesn't make much sense...
4:51 pm:<LS_Chad> because that would remain constant throughout
4:52 pm:<LS_Chad> or did I misunderstand?
4:52 pm:<DerekHome> Going to work on it later this evening. Thanks for all your help Eddy.
4:53 pm:<F_trader> Chad, i got all global report variables like MAE, MFE, some home made Drawdown values available as V variables in my custom reports
4:53 pm:<F_trader> i would like to sort my optimization run against this kind of variables, taking their value at the end of each runs
4:54 pm:<F_trader> is that more clear ?
4:54 pm:<LS_Bill> next update will have File: Open: Optimization so you can get a list and open one or more by name
4:55 pm:<Duker> Chad, are you taking suggestions for basic benchmark report items like buy &H hld return totals & percents and volatility also what about portfolio level items like % of capital etc for entry
4:55 pm:<LS_Bill> File: New: Optimization will prompt for a new and create a new opti setup window
4:55 pm:<LS_Chad> those would just need to become hard coded results for trading systems
4:55 pm:<LS_Bill> the Open Menu will list all named optimization like other objects...
4:56 pm:<LS_Chad> Duke...email me the results you'd like to see...along with details explaining them.....
4:56 pm:<Brian> just ran my first opt -- very cool ;-) for some reason all my Avg Win/Avg Loss reports are 0 or NA -- know what's happening with that?
4:56 pm:<LS_Chad> I agree Buy and Hold should certainly be there...we'll get that in there
4:56 pm:<LS_Bill> File: Save and File: Save as will save / save as the front most optimization
4:56 pm:<F_trader> but there are many ways of calculating MAE / MFE etc
4:56 pm:<F_trader> the idea is to be able to optimize against any variables that you use to evaluate a backtest runs
4:57 pm:<Duker> ok I will
4:57 pm:<LS_Chad> I assume Buy and Hold should just buy at the open of the first bar, and sell at the close of the last bar
4:57 pm:<Duker> correct period tested return
4:57 pm:<LS_Chad> Eddy, so you're saying you're able to leave certain V# variables with these MAE values after the trading system run....
4:58 pm:<Duker> also volatility of period tested B&H return so measure portfolio level risk against
4:58 pm:<LS_Chad> ok...make me a list....sort by importance...and give details....
4:59 pm:<F_trader> yes Chad : i generate all the data of the summary report as V variables + about 15 custom one
5 pm:<LS_Chad> ok..maybe I could just add all the V#'s at the end of the current sorted list...and allow you to sort by an optimized one, or any one you wish....(in addition to periodicity I guess)
5 pm:<F_trader> for example, i will have a variables telling me that my system hit 65% of it partial target profit or that 52% of the stop loss hit are trailing stop (the other one being initial stop) etc etc
5:01 pm:<F_trader> yep, using a V# variables as sorting criteria will make the trick , as long as the value used is the one stored at the end of each run...
5:02 pm:<Brian> would like to be able to sort result list by clicking on column heading rather than having to re-run to see a different sort -- any chance of that for the future?
5:03 pm:<LS_Chad> sure...open it in Excel and have at it ;)
5:04 pm:<Brian> hmmm, but nice if integrated into IRT
5:04 pm:<LS_Chad> yea....be tough to do on our end...I'll consider it
5:05 pm:<LS_Bill> it would be nice if IRT could send an 'open" command to Excel to open the file
5:05 pm:<F_trader> yep, that would be a nice touch...
5:05 pm:<LS_Bill> this can be done on the mac platform using apple events, to tell another application to open a file
5:06 pm:<LS_Chad> So if you have an indicator periods (or anything else) you'd like to use as an optimization parameter....email it to me....and if you have any additional backtesting reports you'd like to see....let me know those as well
5:06 pm:<LS_Bill> that would certainly be easier that us reinventing the wheel
5:06 pm:<Brian> auto export results to excel? that would be a lot easier
5:06 pm:<LS_Bill> we are adding Fractals to the RTL token list in next updated
5:07 pm:<LS_Bill> fractals would be a candidate to optimize on the "bars" parameter
5:08 pm:<F_trader> Bill : i would love to be able to open directly under excel a backtest report made of only NOTE token.. presently need to remove all std report lines, takes some times..
5:08 pm:<LS_Bill> so you would like a notes only report produced?
5:09 pm:<LS_Bill> with time stamps etc.
5:09 pm:<F_trader> exact
5:09 pm:<F_trader> just removing all other lines, ie just taking NOTE
5:11 pm:<F_trader> got all trade individual report under a single NOTE token.. and some NOTE token at the end of the backtest for producing the full report
5:11 pm:<Brian> Chad, I'm gettin 0 or NA under Avg Win/Avg Loss -- there are some iterations with winners and losers.....ideas?
5:12 pm:<LS_Chad> NA means that average loss was 0
5:12 pm:<LS_Chad> 0 means that Avg Win was 0
5:12 pm:<LS_Chad> if Avg Loss is 0, the result is undefined
5:13 pm:<Brian> but don't think that would be the case with all -- I see total gains, losses etc
5:14 pm:<Brian> about 100 iterations
5:14 pm:<F_trader> By the way, Bill, as we discussed a long time ago, would it be possible in a NOTE to have %STIME return an hh:mm:ss time stamp print / %TIME only return hh:mm
5:15 pm:<LS_Chad> so you either have NA or 0 for every Avg Win / Avg Loss
5:15 pm:<Brian> yes
5:15 pm:<Brian> at least so far
5:15 pm:<Brian> just ran it for last month, doing last year now
5:16 pm:<LS_Chad> yea..that would point to the fact that there are either only winning trades or losing trades (or neither) in each iteration
5:17 pm:<Brian> but I saw many that had %Profitable not equal to 0 or 100
5:18 pm:<LS_Chad> OK....I'll do some more testing on my end and see what I can find....
5:18 pm:<LS_Chad> if you run that trading system manually....do you get a valid AWAL result?
5:19 pm:<Brian> ok thanks, will see what this longer test does -- 23% done
5:19 pm:<Brian> yes
5:19 pm:<LS_Bill> Eddy: %TIME will display hh:mm:ss if the bar time has non-zero seconds, e.g. in a 20 sec bar case
5:20 pm:<LS_Bill> when the display would be hh:mm:00, IRT trims off the :00 seconds
5:20 pm:<LS_Bill> so whenever TIME is not an even multiple of 60 seconds, you will see hh:mm:ss
5:21 pm:<LS_Bill> in the NOTE output
5:21 pm:<F_trader> yes, it does it now / just tested it :)
5:22 pm:<F_trader> strange, got the impression it was not working as such last time i tested it...
5:22 pm:<F_trader> anyhow, problem solved...
5:23 pm:<LS_Bill> could have just been a minute periodicity...
5:23 pm:<F_trader> may be
5:24 pm:<Brian> love the "progress report" on the optimizer title bar with the average number as it goes along
5:25 pm:<F_trader> Bill, about automated trading, we need to discuss one day how to update a V variable by picking a value inside a generated text file....
5:26 pm:<F_trader> don't need any complex feature, but just a convenient way to achieve such a V variables update from an outside source which is a text file with few values in it...
5:27 pm:<Duker> Bill & Chad I sent you some pics of WL backtesting and optimization reports let me know what you may need me to expand on
5:28 pm:<LS_Chad> OK...will take a look.
5:28 pm:<LS_Chad> Please give me the specific values you'd like to see as well...with an explanation of what they are.
5:29 pm:<F_trader> even without reading inside the file, updating a binary V variables (0 - 1) just by detecting there is a new given text file would help a lot, as this would provide the information, there is a trade ongoing or not from the ninjatrader/broker platform
5:29 pm:<LS_Chad> Gotta run guys...talk to you later
5:29 pm:<F_trader> bye bye Chad