Brian, I've done a some prototyping of the duplicating symbol process for running multiple strategies. It turns out, as Angelo pointed out in an earlier post, that duplicating and renaming exported symbol text files for import is not much of an issue. I've found several programs that can rename all the files at one time (much like I could do using wild chars in DOS 20 years ago...). Importing the symbols back into AB is also fairly easy because multiple import files can be specified at once. Thus, the most time consuming task is the manual entering the symbol information like margin, tick size, and point value since unlike stocks, with futures most of the symbols have different values. I'm not going to dwell on how best to duplicate symbols too much because it's generally a do once process. By the time I can automate the process, I can probably do it manually.
BTW, I'm still verifying the idea of combining multiple systems using duplicate symbols. I'm testing it with 2 systems and so far the results appear to be correct. Regards, David --- In [email protected], "brian_z111" <brian_z...@...> wrote: > > David, > > If you could rename tickers internally would it help? > > I am pretty certain that the only way to internally rename tickers is via OLE > .... I think there was some discussion and a post or two. > Possibly Dingo was involved in posting the CBT code? > > I think you could also run an ATC and just add OHLCVOI, to each psuedo > ticker, for one symbol only. > > --- In [email protected], "ang_60" <ima_cons@> wrote: > > > > --- In [email protected], "dbwyatt_1999" <dbw451@> wrote: > > > > > > ang_60, > > > > > > Duplicating the symbol data for each model is actually a very interesting > > > idea. In my case that would be 12 x 27 = 324 symbols and 12 watchlists. > > > Is there an easy way to duplicate a symbol? If I export, copy file, > > > rename file, import, manually type in the symbol information (tick size, > > > margin, point value), and manually add to a watchlist, I think I'm > > > looking at about 2 minutes labor per symbol. That would be about 11 > > > hours of setup where it's easy to make a mistake... One little mistake > > > can lead to much more time trying to locate mis-typed symbol information > > > or missing symbol in watchlist... > > > > > > In addition to the 12 watchlists, I assume there would need to be a > > > super-watchlist with all the symbols that would be specified for use by > > > the portfolio back-tester. > > > > > > Your excellent suggestion actually gives us a way to run multiple systems > > > simutaneously on the same instruments (via different symbol names) > > > without worrying about scaling-in and out or simultaneous long and short > > > trades. I now have to decide if I want to take on the labor to set it > > > all up and verify... I've read posts in the past of people running > > > back-tests and explorations with hundreds of symbols so I assume AB can > > > handle 324 symbols (I know it will be hardware dependent). AB uses so > > > little resources with 27 symbols, I'm thinking 324 should not be a > > > problem. > > > > > > Many Thanks, > > > > > > David > > > > > > > > > > David, > > > > every solution ha s its pro's and con's. > > As you know, the csv file you export from AB, is not instantly readable > > from MSA, you need to label every column, and this takes time too (MY time > > and not machine time).... and should be repeated for every test and every > > file. > > > > Copying data requires time too, but.... I mostly use Metastock data format > > (also intraday) and use two utilities-handling data, so > > > > - creating copies of present directories takes minutes not hours; > > - it is au automated task, and so it doens't bother me how many time it si > > required. > > > > The boring part is chianging the names, but this could be automated too > > (people thinking to write a C program, like the person starting this thread > > should be able to do it for breakfast) and - after all - is "una tantum". > > > > But.... as I said, my method is just a stupid workaround.... the definitive > > solution has yet to come. > > > > For the rest, I don't understand the need of a superwatchist with all the > > symbol, as this is simply the "ALL" symbol in the database, ... neither the > > need to use the CBT (nothing against with the CBT, but here it is not > > needed). > > > > As for AB, I had no problem in testing 300+ symbols with more than 20 years > > of daily data (simple code, no loops that are time consuming). I have some > > problems in testing large amount of intraday data, but as I'm not sure what > > is the cause, I will retain myself for any further comment on this matter. > > > > Greetings, > > > > Angelo. > > >
