Page 1 of 1
Imagining the Dream Testing Software?
Posted: Thu Oct 16, 2003 9:33 pm
I often find it useful to think of how things might be if the limitations which constrain our possibilities no longer existed.
Turning this thought to trading system research, if testing software were somehow magic and could work the way you wanted, how would you like to test trading systems? I'm not thinking so much about features for software as the processs.
How would things be different than they are today?
What really bugs you about the testing software you use? What do you really like?
- Forum Mgmnt
Posted: Fri Oct 17, 2003 9:29 am
Why should I have to completely specify the trading system, down to the very smallest detail, myself; let the test software do that for me. I would say Gimme a reversal system that never goes flat, makes about two round-trip trades per month, and trades when the shorter timeframe trend diverges from the longer timeframe trend. Then the software does the rest and presents the Metastock code & the backtest results, along with a half dozen recommended portfolios. Just because the test apparatus is a computer, doesn't mean I should have to program it.
Posted: Fri Oct 17, 2003 5:18 pm
My dream testing system:
Iâ€™d like to see a framework put around system testing that allowed me to plug and play different pre created components into a complete system. With the flexibility to add more from the public and create my own.
I would be able to:
â€¢ Pick from a list of entries and modify the parameters. Choose as many as I want. Or go to a design sub element where I can invent a new one.
â€¢ Pick one or more filters to either filter the entries or confirm them.
â€¢ Pick from a list money management systems or combine several. Or go to a separate module for inventing my own or plugging in new ones created by my friends on the web.
â€¢ Pick from a list of exit statagies. This area would need to be infinitely adjustable and customizable but easy to implement.
â€¢ Pick your list of markets, or pick from several pre-chosen groups. Or pick a dynamic stock universe. **All Stocks with that crossed above the 200 day moving average yesterday in the S&P 500** etc.
It should work equally as well for futures and stocks
It should have a robust set of reports and graphs with the flexibility to design my own.
Additionally, After I found a system I like and had tested it sufficiently, I could hook up to a paper trading system and try it for a few months. If I still liked it, I could go live at one of several highly competitive brokers and have the system traded for me.
Okay, Iâ€™m sure there is more, but itâ€™s time to wake up from my nice little dream.
Posted: Fri Oct 17, 2003 7:57 pm
I'd like to have a trading system test software with very flexible money management testing abilities. Giving me room to invent knew mm possibilities. I'd want the dollar win ratio to be able to influence my exits and govern which set ups are optimal at the given time.
Dream Software Feature
Posted: Thu Oct 23, 2003 7:12 am
Portfolio selection would have a lot of visual aids. I would find it useful to have a graph of the instrument over the testing period. If I wanted to conduct some test for the 'trendiness' of an instrument, just having a visual representation would help in thinking up a measurement. A visual comparison of graphs, one instrument next to another, would be a quick way to get an idea of correlation between them.
It might also be nice to see the systems actions plotted on the graph of the instrument. Instrument specific statistics would also be a good way to get an idea of the impact of a specific instrument on the system.
A key feature of the testing software of my dreams is the capability to search through my library of test results by a variety of criteria.
I could tell the program to bring up all the tested systems which included an MAR ratio > 1.7. Or it could give me all the tests run between date x and date y that tested N stops between 2 and 2.25 and had exits greater than 18 days. Or all the results screens that included any of the top 10 performing systems as measured by CAGR. Or by instruments in the portfolio...
In other words, a very flexible search function that could pull past tests by the parameters I put in or by the measurement of the system in the results.
Posted: Sun Oct 26, 2003 2:20 am
Ideally, my dream software would have the ability to see five bars into the future in any time frame.
Posted: Sun Oct 26, 2003 3:16 pm
An optimization engine with a genetic algorithm that lets one enter the results they want , 60% winners 100% return 20% draw down 200 trades per year ect. and the software finds the parameters that come closest to achieving this.
Posted: Mon Oct 27, 2003 3:23 am
Market selection could be an interesting area
with trendiness and correlation measurements available for individual instruments and in comparison with each other in a portfolio.
Systems and personality
Posted: Mon Apr 12, 2004 5:33 pm
Interesting question, it comes down to your key proces. Making the best research software in order to give traders confidence to follow a system.
Anyone wants the most profit with the least drawdowns.
The system should give an answer on questions like:
I want x% profit per year and maximal drawdowns of y%, what system(s) should i trade. It is not so important if you trade moving averages or breakouts....
Or the system should give you an Ansatz (starting point) and then you can pick it up and adjust it to your personality (how much pain can you sustain?). And then: as i understand, you should have as much as reliable data as possible. It would be a very strong issue if you would get e.g. 20 years of reliable data included. When you buy a car and no fuel is in it you cannot drive, even not to the first filling station.....
I'm seriously planning to buy Veritrader 1.5 or 2. And i will trade my own account, and later for clients (I'm a managed accounts trader, but with short term trendfollowing systems).
Perhaps it is on other places in this forum, but what is a minimum account for trendfollowing? I think about 200K?
Posted: Wed Apr 14, 2004 11:57 am
Correlations of the results from an individual trading system applied to individual trading instruments. Correlations of the different instrument groups.
This would aid in market selection.
Ability to easily specify trade summary information, like STD, averages, max 5 losses.
Zeke's past test library would be awesome. How about autolibrary of functions/strategies as they are changed? When I do research on a system, sometimes it is difficult to remember all the little changes I've made.
Built in and reviewable comments section independent of the system building screen.
Ability to see the life cycle of individual trades in a graph, typical winning/losing trades.
The ability to toggle between different stop regimes.
And of course, some guy to sit around and instantly program my smallest ideas.
Posted: Wed Apr 14, 2004 12:04 pm
Mike Sankowski wrote:Zeke's past test library would be awesome. How about autolibrary of functions/strategies as they are changed? When I do research on a system, sometimes it is difficult to remember all the little changes I've made.
Try using a version control system (like RCS, CVS or subversion). It actually helps on tracking what, who, how and when.
Posted: Mon Apr 19, 2004 6:32 am
My personal dream testing system:
Is capable of backtesting and realtime trading (forex, futures, equities)
Can collect/store/process tick by tick and daily data
Filters for cleaning tick by tick data (realtime)
Multiple timeframes per instrument based on tick or daily data
Synthetic timeseries for instruments (forward adjusted, ratio adjusted, ..) for calculating signals
Trading Systems are composed of reuseable rules (entry/exit, initial stop, trailing stop, ...)
- single system on a single market,
- single system on many markets
- multiple systems on a single market
- multiple systems on many markets
Portfolio Allocation for systems/markets
Supporting different currencies
Considering Commision, Slippage, Rollovers (as real Trades)
Monte Carlo simulation for equity curves
Customizeable backtesting reports (in plain ascii and html)
Database for quering/storing backtests
Visual support tools:
Charting tool for showing entries, exits, signals, ...
Charting tool for showing equitycurves and parameter sensitivities
The good thing is i have already implemented 85% of the above features, the bad thing is there are still 15% to do...
Posted: Mon Apr 19, 2004 2:24 pm
I think I've entered a parallel world where people are friendly instead of surly.
Thanks. I'll take a look at it and report back...
Posted: Mon May 24, 2004 11:03 am
One dream for dream testing software would be the ability to copy and paste some type of price development, insert the price data into a "language decoder" in which it would break down the price action into a variety of different code translations based on a variety of price relationships, indicators etc. Say 25 different ways or perspectives to say the same thing in quasi code. ATR moves, perecentage moves, support and resistance, types of momentum thrusts etc and countless things i cant think of (thats the rub
From those automatic translations, i would try to find some similarities through a variety of similar but different looking price patterns to see if there is a universal and codable way to represent a certain price development..
Posted: Mon May 24, 2004 11:07 am
This might be more possible than you might think.
- Forum Mgmnt
Posted: Mon May 24, 2004 11:39 am
That sounds a litlle like Safir x at http://www.sirtrade.com
pricey, but interesting.
Posted: Sun May 30, 2004 9:04 pm
William, are you thinking of something such as fourier analysis but using different indicators???? Are you thinking of using such structures for pattern identification?
Posted: Sun Aug 15, 2004 12:47 am
Speaking of which, has anybody tried it? Is anybody using it? It's near impossible to find someone that uses it. The developer is supposed to implement real-time calls on his site, but nothing yet...
PS: Moderator, should we move this one to a new thread?
Posted: Wed Apr 08, 2009 12:52 pm
Mark_et_Lizard wrote:An optimization engine with a genetic algorithm that lets one enter the results they want , 60% winners 100% return 20% draw down 200 trades per year ect. and the software finds the parameters that come closest to achieving this.
Genetic optimization engine is a great thing. Tradestation just
added this feature as of the latest update. Perhaps to
compete with Multicharts who have offered this feature
for awhile now.
TS still lacks portfolio level backtesting by itself (adaptrade.com
offers a solution with their MSA software though).
I hope that TB will offer this feature. It would really speed things
up (I'm getting about a 10x speed increase with genetic
optimization vs. brute force).
Seeing that TS is now stepping on the toes of vendors who
have offered genetic opt. via add-ons, it's just a matter of
time before they offer portfolio backtesting.