WELCOME TO TRADING SYSTEM LAB® PROVIDING THE TSL ML/AI STRATEGY DESIGN PLATFORM SINCE 2006
OUR LATEST FEATURE-"DAS™" INTRO: HERE
"DAS™" EXAMPLE: HERE
ES STRATEGY EXAMPLE USING TSL-LAIMGP IN 6 MINUTES: HERE
- JULY 2020 UPDATE:
- TSL is pleased to report that TSL Machine designed trading strategies may now be ported to QuantHouse. See Demo #72 (8 minutes) to see how strategies may be designed and dropped into QuantHouse with no programming. QuantHouse, the global provider of end-to-end systematic trading solutions including innovative market data services, algo trading platform and infrastructure products and part of Iress (IRE.ASX), today announced that Trading System Lab® (TSL) has added their machine learning capabilities as part of the QuantFactory cloud backtesting suite. The QuantFactory cloud backtesting suite provides a fully configurable environment in which clients can develop, backtest, optimise and implement quantitative trading strategies that can later be executed in a standalone, live-trading environment. Machine learning outputs from TSL are integrated into the QuantDeveloper module of QuantFactory. See Press Releases for more information on this development.
- Sheltering in place has given us some well needed time for research and that has resulted in yet more breakthroughs in applied ML. At the present time we are integrating these breakthroughs into TSL and evaluating them further. With these breakthroughs, the OOS Reward to Risk numbers that are coming out of the engine are higher than we have seen before.
- We are offering an immediate 1/3 off on all new TSL Licenses and this offer will remain in place as long as this global Covid-19 pandemic continues. We will also increase the duration of all new licenses to 4 years.
- News Flash: We are testing a new look for our web site here: Go to new TSL web site » Comments would be appreciated!
- Many of us have been required to stay at home to avoid the spread of Covid-19. Our hearts go out to those affected by this terrible disease. We all hope this pandemic will be over soon and life can return to normal. Everyone please take precautions and stay well.
- See our latest Demo #70 showing how 20 simultaneous instances of TSL running on one CPU produce design rates of over 80,000 strategies per second with less than 50% CPU and memory utilization on a DELL/ALIENWARE Workstation. TSL is indeed fast and light!
- News Flash: A multi month review of TSL Seasonal Strategies auto-written by our code generator show superior Double Out of Sample performance over a large range of markets. This "Seasonal Code Writer" is included with, but is not a direct part of the TSL Platform. However integration work of specific seasonal characteristics into the TSL ML/AI Learning Machine has begun.
- Portfolio construction within TSL is multidimensional. Rather than simply place a portfolio of symbols in front of a strategy that is trading all symbols with the same direction of trades, TSL, as in single market design work, inverts the problem starting first with a desired goal, say Sharpe Ratio or Total Return/Flat Time, and converges to a solution that will trade each symbol with different directionality. For example, in a simple Portfolio of DIA, SPY and QQQ, TSL may be short DIA, long SPY and flat QQQ at the same time. Thus there is an element of hedging inherent in this Machine Learning paradigm. The specific way each strategy trades the entire portfolio is used to converge to the total desired goal. Strategies are then exported in a variety of languages with no programming required.
- The overnight markets have been rather volatile lately. TSL has numerous custom session and activity features allowing you to capture this volatility yet limit your total market exposure. With TSL you may use the overnight market information and preprocessed facts to trade the day session, use the day session information to trade the overnight sessions or participate in the entire 24 hour market entering and exiting at various times throughout the 24 hour sessions.
- See the latest demos on TSL, including runs on the latest Core i9 9980XE from Alienware/Dell. We also demo how to have multiple instances of TSL running on one PC. The new Dell Alienware Area 51 Core i9 9980XE CPUs provide for a substantial boost to system design speed over the older Core i7's. All Strategy design types benefit from this higher throughput including single market systems, pairs, portfolios and daytrading systems.
- Data is very important in the design of Trading Algorithms. Without enough data or without good quality data the design process cannot even begin. TSL and CSI have teamed up to provide excellent Daily data to you and CSI has positioned the TSL recommended portfolio of data as a Vendor Portfolio directly downloadable from the CSI UA application. IntraDay data can be derived from TradeStation directly or from several Third Party intra day data Vendors. Importation of Third Party Data into platforms such as TradeStation and MultiCharts can be trickey, but thanks to the TSL helper files, this process is simple. To learn how to import this data into TradeStation see Flash Demo Number 66.
- With TSL you can design Trading Strategies with just a few clicks and a few minutes of time. The strategies emerge ready to trade and this process requires no programming making it perfect for those who want to trade powerful Machine Designed trading strategies, but who are not programmers. The TSL platform has undergone many upgrades and enhancements since its creation in 2006 making TSL one of the most mature Machine Learning based Trading Strategy design engines in the world.
- TSL has just released a Cyclical-Seasonal Code Writing Algorithm based in part on the Ruggiero-Barna Seasonal index. See Cybernetic Trading Strategies by Murray Ruggiero for more information. This Code Writer allows for an advanced preprocessor creator for the TSL Preprocessing Fact Set. The results of the basic Code Writer have been remarkable. For either Daytrading or EOD systems, Profit Factors in excess of 3 result. Out Of Sample results are excellent and the algorithm works well on many markets including index markets and commodity markets. A filter finder with 27 filters is included if additional filtering of generated systems is desired for even further performance. This algorithm allows you to select a minimum acceptable ProfitFactor, Number of Trades, Net/Drawdown and FlatTime for the generated systems. Code is then simply pasted into PowerLanguage or Easy Language. Work on this approach is continuing and is shown to be particularly advantageous to the Expectation based Learning inherant in the TSL architecture. For Seasonal work the biggest issue is lack of data. If a pattern occurs once per year and you have 30 years in your data base then you have only 30 events to attempt to draw a conclusion about. This may not be enough information to provide robustness moving forward, so caution with seasonal work is indicated. The base TSL package, without seasonal effects can produce many strategies with 100 to 1 or higher Trade to Parameter Ratios, giving much greater confidence in robustness moving forward.
- In addition to the TSL related Platform known as Quant Systems Lab(TM), TSL has a number of Research and Development projects underway. Names like CET, White Light, FLY, WARM, DG, DNA and SPLIT are code names for these internal projects. DAS(TM), EVORUN(TM), Super Buffer, Null Reject and Post Design Eval, were created and then subsequently included into TSL using this same development process. If you are interested in becomming a early user of these advanced features contact TSL ASAP. These new and creative features and capabilities are well beyond any other software tool available.
- TSL provides Machine Learning based Trading Strategy designs complete with trade ready code. No programming is required although programmers can add value to their project with TSL, as one example, by adding their own material as a Custom Preprocessor. TSL can design Trading Strategies using Machine Learning in three very different ways. The first way is that a simple Multi-Run can be accomplished where the Platform starts, runs and stops, then exports the final Trading System. The second way is to use Evorun(TM) which performs a batch approach to design say 100,000 strategies on each setup chosen to be iterated over. You can select to vary the Bar Size, or Preprocessor, or Fitness Function or other variables. A Grid is then displayed which shows you setups that were Machine Designed that are of particular interest. The third way to use TSL is through what we call DAS(TM). DAS allows you to change the area of interest to be explored WHILE the Platform is running, in "Design Time". For example, you might start with entering on a Stop using Fitness Objective 3, Preprocessor 4, with Forward 20% Out Of Sample. Then "During the same run" you might switch to entering on a Market, using Preprocessor 1, Fitness Function 33 and "Back" 30% Out Of Sample. In this case you never stopped the run, but used DAS, which allowed you to "Reach Into" the live running engine and alter parameters "During" the Design. DAS is just one of the amazing features that have been added to this industry leading Platform. Feel free to watch as many videos as you wish to see for yourself how amazing and simple this Platform is to use.
- TSL is Artificial Intelligence and Machine Learning at its best! TSL is a Machine Learning algorithm that automatically writes Trading Systems. No programming is required. TSL is a remarkable Platform given the fact that the Trading Systems designed by the TSL machine over 10 years ago have never been reoptimized or altered in any way and are still top 10 rated by Futures Truth. TSL employs a Patented Automatic Induction of Machine Code with Genetic Programming engine capable of very high speeds and TSL produces production code, reducing or eliminating the need for trading system programming efforts and technical analysis expertise for stock markets, futures markets or any other tradable market. The Executive Brief and Demo located above will give you a overview of this powerful trading strategy production tool. It is important to note that TSL designs an unlimited number of Trading Strategies on any market, any time frame, day trading or end of day, as well as portfolios, pairs and options, again, with no programming required. Clients range from beginners to PhD level Quant researchers and developers, domestic and international, as well as CTA's/CPO's, Hedge Fund's and Prop shops. Now, with 13 years of experience serving trading customers, TSL can bring that expertise to your trading operation. TSL provides one-on-one training and consulting at no additional cost to clients, to help ensure clients get the most out of the TSL engine.
- We have completed a Demo on Bitcoin market data. You can view the demo here: Go to Bitcoin Demo »
- TSL produces completely OPEN CODE machine learning based trading strategies requiring no programming on the part of the user. TSL is not a Black Box. The math, variables, logic, signal generation, preprocessing, etc., are exported in OPEN CODE. Many of the systems come out of the evolutionary process extremely simple with the core GP code being only 7-15 lines of code, using perhaps 3-5 variables. See our Las Vegas Traders Expo PPT for an example of a system that used only one (1) parameter here: Go to the LVTE Power Point » The process within TSL results in simple, high performance trading strategies, and simpler is better.
- TSL is very easy to use which is why we have clients ranging from beginners in Technical Analysis and Trading Strategies to PhD's in Computer Science, Economics, Machine Learning and AI. Our 6 minute demo summarizes how easy TSL is to use. If you can accomplish these three steps, you can use and be productive with TSL. Go to the TSL demo »
- In the Latest Issue of Futures Truth, TSL's ML designed Trading Strategies remain highly rated, even after 10 years of Forward Walk testing on Sequestered Data, the harshest form of Out of Sample testing. Two of the top 10 eMini S&P systems are TSL systems for example. Go to the Futures Truth website » Additional historical reports may be found in Futures Truth's reports as well as in TSL presentation material. Read the opinion letters from Futures Truth and other developers and traders here: Go to the Futures Truth Opinion Letter »
- Numerous new features for 2016 have been added to TSL including In-Sample/Out-of-Sample Scatter plots with Wilcoxon tests, Design-Time Adjustable Solutions(DAS), DayTrade Discrete Bars(DTDB), SuperBuffer increases, SubSystem Usage Reports and a soon to be announced options testing integration feature. Please take a look at our latest Flash Demos: Go to the TSL Flash Demos »
- TSL is pleased to announce that CSI COMMODITY SYSTEMS, INC. and TSL have formed an agreement to provide to our clients a portfolio of commodity data, specifically engineered for TSL Machine Learning. To obtain this data a CSI data subscription is required. No other vendor provides this specifically engineered data. This daily data will allow for improved Trading Strategy design using TSL and is the result of many years of research and development of data requirements. Without proper data, robust Trading Strategy designs are very difficult to accomplish. These data portfolios are downloaded and installed as part of the CSI data application. Helper files such as .DOP's and Attributes.INI files are preassembled by TSL to allow for easy data import into TradeStation. Other platforms that can read ASCII, MetaStock or CSI price data may load this data as well for use with TSL. Contact TSL to learn more about this new Trading System design data. CSI has been shown to have the most accurate commodity data available. Go to the CSI data report »
- Arguments regarding the use of Out of Sample (OOS) data are generally centered around the possible accidential use of this held out data in the development processs. If this happens, then the blind data is no longer blind, it has been corrupted. To eliminate this possibility, TSL submitted machine designed strategies for testing on Sequestered Data. What this means is that the strategy performance measurement occurs in the future. Since the held out data does not exist when the strategies were designed, there is no way that this evaluation data can be accidentally used in the development process. Strategies produced by the TSL Machine have been tested on Sequestered Data by the independent third party, Futures Truth and are top rated, beating most other "Human" or Manually designed Trading Systems.
- NEW! Here is how you use TSL evolved systems in a C++ or C# OMS/EMS: View the TSL C# Brief: »
- The free period is over for the new Kindle Book containing our article titled: "Machine Designed Trading Systems", however you can download this inexpensive Kindle Book here: Download the Kindle Book »
- TSL is now officially on the Silicon Valley Map!. Silicon Valley Map and TSL location(6 o'clock position)»
- TSL is capable of using multi-data DNA within its preprocessors. See Demo #48 where we use the CBOE Volatility Index (VIX) to Machine Design a eMini S&P Trading System. This type of design work is simple to accomplish in TSL since the preprocessor is completely customizable using your unique patterns and indicators in a single or multiple data stream design. Enhanced Preprocessors have been shown to offer an additional boost to Trading System performance.
- How did the "TSL Software that writes Software Machine" out-design other human submissions to FT with no programming required? How do Machine Designed Trading Systems actually work? Our development chronology is well covered in our White Papers and Flash Demos available on the TSL web site.
- The 2015 OUANTLABS WEBINAR can be found here: Go to the 2015 QUANTLABS WEBINAR »
- The 2014 OUANTLABS WEBINAR can be found here: Go to the 2014 QUANTLABS WEBINAR »
- What is the Optimum Bar Size to trade? 100 tick, 15 minute, daily??? TSL's new EVORUN module allows strategies to be Machine Designed while iterating over Bar Size, Trade Type, Preprocessor, Trading Frequency and Fitness Function in one multirun. EVORUN and TSL Version 1.3 Demos 51 and 52 are now available here: Go to TSL Demos »
- ALL TSL STRATEGIES ARE FULLY DISCLOSED IN OPEN CODE.
- WANT TO READ A BOOK ON THE TSL GENETIC PROGRAM? Frank Francone co-authored the university textbook "Genetic Programming: An Introduction" (The Morgan Kaufman Series in Artificial Intelligence).
- TSL is a stochastic, evolutionary, multirun, Trading Strategy autodesigner that produces and exports portable code in a variety of languages. This is a complete end to end Trading System design platform and will autodesign High Frequency Trading Systems, Day Trading, EOD, Pairs, Portfolios and Options Trading Systems in a few minutes with no programming. See Theses, White Papers, PPT Presentations and other documentation under the Literature Link at the left. Watch the Flash Demos at the left for a complete briefing on this new technology. The TSL Platform produces "Machine Designed", Trading Strategies at ultra high rates thanks to "register level" evaluations. No other trading strategy development platform on the market provides this level of power. The LAIMGP-Genetic Program within TSL is one of the most powerful algorithms available today and operates at rates much faster than competing algorithms. With TSL, trading systems and code are written for you in languages including C, JAVA, Assembler, EasyLanguage, and others through translators.
- Frank Francone, President of RML Technologies, Inc. has prepared a flash demo titled "Genetic Programming for Predictive Modeling". RML produces the "Discipulus" Genetic Programming engine that is used within TSL. This tutorial is an excellent way to learn about Discipulus and will provide a basis for your continued understanding of TSL's Auto-Design of Trading System Paradigm Shift. TSL simplifies the data import, preprocessing and design of Trading Systems using Trading System performance as fitness. Make sure you watch the TSL demos as the TSL platform is specifically targeted for Trading System design. Download the Discipulus tutorial »
- The technology used in Trading System Lab is 60 to 200 times faster than other algorithms. See White Papers on speed studies at SAIC here: Go to white papers »