Program Trading Software: Free Options and 6 Checks

Contents

1. What Is Program Trading Software? The Three Things It Does
2. What Free Program Trading Software Is There, and How Far Does It Go?
3. Do You Need to Code to Use Program Trading Software?
4. Can Your Broker's Own Software Be Used for Program Trading?
5. Cloud, Desktop or Your Own Code: What Differs
6. How to Choose Program Trading Software: Six Checks to Run in Order Before You Pay
7. Frequently asked questions
8. References

What Is Program Trading Software? The Three Things It Does

Program trading software is a tool that hands your written trading rules to a computer to execute. Its functions fall into three layers: strategy generation, backtesting and validation, and order execution. One product usually covers only one or two of them. Before you choose, confirm that it reads your market data, backtests with commission and slippage, offers out-of-sample validation, and exports to the platform you actually trade on.

Think of it as running a restaurant. The first layer is developing the dishes: coming up with rules, deciding parameter ranges, and searching across many combinations. The second layer is tasting and costing: replaying the rules on historical bars, deducting commission and slippage, then checking against data the rules have never seen. The third layer is serving: connecting live quotes, sending orders to your broker, and managing positions. A cook is not always a good bookkeeper, and a bookkeeper may not serve tables. Software works the same way.

Each layer needs one thing from you. Strategy generation needs your idea, or at least a direction, such as “trend-following after a breakout” or “buy a pullback to a moving average.” Backtesting needs your historical data, including the correct sessions and contract specifications. Order execution needs a broker account and a platform that can connect to it. If you have none of the three layers yet, TAIEX futures (TX) traders can start with a no-coding route into TAIEX futures program trading, from zero, then come back and compare against the checks in this article.

What Free Program Trading Software Is There, and How Far Does It Go?

Free options fall into roughly four types. Each usually covers only one layer, either backtesting or execution. Mass strategy generation and systematic out-of-sample validation mostly require paid tools or your own work. This describes the types in general, not any specific product.

1. The strategy tester in a free charting platform. You write a simple script on the chart, the platform runs it once over historical data, and you get a report. You can usually set commission and slippage, but it tests one strategy at a time, and you split the out-of-sample data and keep the records yourself.
2. Writing it yourself in an open-source language or a spreadsheet. Python or Excel can both run backtests. The upside is full transparency. The downside is that data cleaning, session settings, the cost model and the out-of-sample split are all yours to write, and one mistake produces a good-looking but wrong report.
3. Your broker's own software. It mostly provides live quotes, order entry and simple conditional orders. Backtesting is not its focus, and it usually has no strategy generation or out-of-sample validation. The next section covers this in more detail.
4. An API from the exchange or broker. An API is an interface that lets your own program send orders and receive quotes directly. It covers only the execution layer, so you still prepare the first two layers yourself.

Type of free option	Strategy generation	Backtesting	Out-of-sample validation	Order execution
Strategy tester in a free charting platform	Usually none	Usually yes	Do it yourself	Usually none
Your own code in an open-source language or spreadsheet	Do it yourself	Do it yourself	Do it yourself	Do it yourself
Your broker's own software	Usually none	Usually none	Usually none	Usually yes
An API from the exchange or broker	Usually none	Usually none	Usually none	Usually yes

So free software is not unusable. You just need to know where it stops. A common route is to get one idea running in a free tester, make sure you can read the report, and then decide whether to pay for the first two layers. The six checks in the last section of this article go through, in order, what to look for in a paid strategy generator.

Do You Need to Code to Use Program Trading Software?

Not necessarily. Depending on the software, there are three ways to tell the computer a rule: settings-based, drag-and-drop, and code-based. You can test rules without coding, but reading the exported code still helps, because whether the order logic and cost settings are correct is ultimately checked in the code or on the settings page.

Settings-based means choosing an indicator from a menu and filling in numbers, for example picking a moving average, entering 20, and choosing “buy on a close above.” Drag-and-drop means connecting condition blocks, like joining pipes. Code-based means writing the platform's language directly, such as PowerLanguage, Pine Script or MQL. All three test the same rule. The difference is where your time goes: settings-based work goes into options, code-based work into syntax.

Even if you don't code, you still need to check two things. First, the order logic: a signal is generated at a bar's close, and the fill should come on the next bar, not at the close of the same bar, because you did not know that price at the time. Second, the cost settings: whether commission, transaction tax and slippage are entered for every trade, and whether they match your broker's actual charges. If you can read the exported code, you can confirm both again before live orders. If you haven't written any strategy yet, the 30-day route to your first algorithmic trading strategy puts these steps in order.

Can Your Broker's Own Software Be Used for Program Trading?

Yes, for the last layer. Taiwan brokers' own software suits order execution and live quotes, and most also offer simple conditional orders such as price-touch and stop loss orders, but it usually does not include strategy generation or out-of-sample validation. A common approach is to finish testing in research software, export the strategy to an order platform that can connect to your broker, and execute through your broker account.

The reason is a division of labor. Broker software is designed to let you see quotes quickly, send orders and check your account, and it does all three thoroughly. Research software is designed to find, among thousands or tens of thousands of candidate rules, the few that pass validation, and turn them into code that can run. The middle layer that joins the two is an order platform that supports broker APIs.

In practice, first ask your broker two questions: which order platforms or API connections it supports, and which products those interfaces cover. The answers decide which language your research software must export. Choosing the execution side first and working back to the export format is much easier than the reverse.

Cloud, Desktop or Your Own Code: What Differs

The three approaches differ in who decides each setting. A cloud platform saves installation and setup time, but the platform decides the data range, sessions and cost model. Desktop software leaves you in control of data and computing power. Writing your own code is the most flexible, but you are responsible for data cleaning and execution differences. Which one to pick depends on which layer you want to spend your time on.

Item	Cloud platform	Desktop software	Writing your own code
Setup time	Shortest: open an account and use it	Install the software and import data	Longest: you build both the environment and the data
Who decides the data source and range	The platform	You	You
Computing power	Allocated by the platform	Your computer, depending on its CPU or graphics card	Your computer or a rented server
Who decides commission and slippage settings	Fields the platform provides	You, on the software's settings page	You, in your code
Export to an order platform	Depends on the platform	Depends on the languages the software supports	You write the conversion
Skills needed	Reading reports	Reading reports and settings pages	Programming, data handling, statistics

A simple rule of thumb: if data matters most to you, for example you need to test the TAIEX futures night and day sessions separately or use a continuous contract you prepared yourself, desktop software or your own code fits better. If you only want to validate one or two ideas first, a cloud platform is fastest. If you want to test a large number of candidate strategies at once, computing power becomes the bottleneck. Desktop software then runs as fast as your computer's CPU or graphics card allows, and your own code runs as fast as you write it efficiently.

How to Choose Program Trading Software: Six Checks to Run in Order Before You Pay

Run six checks in order before you pay. If one fails, don't look at the next, because software that cannot read your data is of no use however many other features it has. The order is: market data support; backtests that include commission and slippage and fill on the next bar; out-of-sample and walk-forward validation; export to the platform you actually trade on; that platform connecting to your broker; and reports you can keep and compare.

1. Support for your market data. Confirm the software can load bars for the products you trade, such as TAIEX futures, ES and NQ, and can set the correct session. What passing looks like: after you import a stretch of data, the number of bars per day matches the session you expect. What failing looks like: the night and day sessions are stitched into one daily bar, or the time zone is off by several hours. You can convert each exchange's sessions to Taiwan time with the free futures trading hours lookup; for sessions that cross midnight, such as US index futures, how to set up a backtest for the Mini Dow futures night session in Taiwan time has a full side-by-side conversion.
2. Backtests include commission and slippage and fill on the next bar. Slippage is the gap between your actual fill price and the price the backtest assumed. Next-bar fill means that once a signal is confirmed at this bar's close, the order fills on the next bar. What passing looks like: the settings page has commission and slippage fields, and they are deducted in every evaluation. What failing looks like: costs are added separately after the backtest finishes, or the fill price is the close of the signal bar.

   Hypothetical example

   Suppose a Mini-TAIEX (MXF) strategy is backtested over ten years with an average gross profit of NT$600 per trade across 2,000 trades. Suppose round-trip commission and futures transaction tax total NT$150, and slippage is one tick, NT$50, per side, or NT$100 round trip. Cost per trade is then NT$250, leaving NT$350 net. If slippage doubles, cost per trade becomes NT$350 and net falls to NT$250. For the same strategy, whether the cost fields are filled in can completely reverse the conclusion on the report.

3. Out-of-sample and walk-forward validation. Out-of-sample validation re-runs the rules on data that was not used to tune parameters, to see whether the results still hold. Walk-forward analysis repeats that split many times along the history, tuning parameters on one segment and validating on the next. Pardo (2008) laid out walk-forward analysis as a standard procedure, and Bailey et al. (2014) explain why the best result picked from many candidates is almost always overstated. What passing looks like: the software splits the data automatically and records each segment's out-of-sample result automatically. What failing looks like: it can only run one stretch of history and give one result.
4. Export to the platform you actually trade on. Research software should end with code that runs on your order platform. What passing looks like: the exported file loads directly on the target platform, and the entry and exit bars match the backtest. What failing looks like: you get only a report or a chart and have to rewrite the rules yourself. AlphaPrime (Valynox) currently exports to MultiCharts (PowerLanguage), TradingView (Pine Script v6), TradeStation (EasyLanguage), NinjaTrader (NinjaScript), MetaTrader 4 (MQL4) and MetaTrader 5 (MQL5), and the full list of supported export platforms is maintained on that page.
5. That platform connects to your broker. The exported code must run on a platform that can connect to your broker account. What passing looks like: the broker publicly lists support for the platform or offers a matching API. What failing looks like: the platform can backtest and chart, but you have to find another route to send orders. Ask your broker for this step, not the software vendor.
6. Reports you can keep and compare. The settings and result of every test should be saved so you can compare them months later. What passing looks like: reports for different versions of the same strategy can be viewed side by side, with the cost settings recorded. What failing looks like: the result disappears when you close the window and you can only take screenshots.

To show what to look at in a report, here is a real test. The box below is a search AlphaPrime ran on ES daily bars from early January 2019 to the end of June 2025, long and short, 1 contract, with $20 commission and $25 slippage deducted per trade. It tests an alternative rule that uses the Kaufman Adaptive Moving Average (KAMA) as the only entry and exit, with no stop loss, profit target or filters. The purpose is only to show how to read a report, not to evaluate any program trading software or the indicator itself. This result is one kept from the 1,024 candidates the search tried, ranked on return relative to drawdown. Picking the best of more than a thousand after the fact inevitably flatters the numbers, so these numbers do not show that this rule earned this much money, and readers should not expect the same result. Harvey and Liu (2015) argue exactly this: a number on a report should be discounted by how many candidates were tested behind it.

Three things in this report are worth reading. First, read net profit and maximum drawdown together. Drawdown is the deepest loss you must sit through along the way, and a result with high net profit but a drawdown close to half of that profit is hard to hold in practice. Second, there are only 65 trades. Six and a half years of daily bars never produce many trades, so this sample is too small, and any per-trade average can only be treated as a clue. Third, a time in market of 93.01% means the rule holds a position almost all the time, flipping between long and short with no stop loss, which bears directly on whether you can sleep with it. Good backtesting software puts these numbers on one page so you don't have to flip back and forth.

Here is a next step: open any backtesting software you have, find the commission and slippage fields, enter your broker's actual charges, and re-run the same strategy. Note how net profit, maximum drawdown and number of trades change. Then compare those three numbers with their positions in the report above and see whether your software puts them on one page. After this step you will know whether check 2 passes, and you can move on to check 3.

Frequently asked questions

What free program trading software is there?

Free options fall into roughly four types: the strategy tester in a charting platform, writing your own in Python or Excel, your broker's own software, and an API from the exchange or broker. The first two lean toward backtesting, and the last two toward order execution. Mass strategy generation and systematic out-of-sample validation mostly have to be filled in with paid tools or your own work.

Do you need to code to use program trading software?

Not necessarily. Settings-based software has you pick indicators from a menu and fill in numbers. Drag-and-drop software connects conditions with blocks. Code-based software means writing a language such as PowerLanguage, Pine Script or MQL. You can test rules without coding, but you still check for yourself which bar a signal fills on and whether costs are entered.

Can your broker's own software be used for program trading?

Yes, for the order execution layer. Taiwan brokers' software mostly provides live quotes, order entry, account information and simple conditional orders, but it usually has no strategy generation or out-of-sample validation. A common approach is to finish testing in research software, then export the strategy to an order platform that can connect to your broker and run it there.

What is the difference between backtesting software and automated trading software?

Backtesting software applies rules to historical data, deducts costs and checks the out-of-sample results. It answers “did this rule hold up in the past?” Automated trading software connects to live quotes, sends orders to your broker and manages positions. It answers “should this order be sent now?” A product may do only one of the two, so before choosing, find out which layer it covers.

How do you choose program trading software?

Run six checks in order: it reads your market data; it backtests with commission and slippage and fills on the next bar; it offers out-of-sample and walk-forward validation; it exports to the platform you actually trade on; that platform connects to your broker; and reports can be saved for comparison. If one check fails, skip the next ones. This saves a lot of trial time.

References

• Bailey, D. H., Borwein, J., López de Prado, M., & Zhu, Q. J. (2014). Pseudo-Mathematics and Financial Charlatanism: The Effects of Backtest Overfitting on Out-of-Sample Performance. Notices of the American Mathematical Society, 61(5), 458–471.
• Harvey, C. R., & Liu, Y. (2015). Backtesting. The Journal of Portfolio Management, 42(1), 13–28.
• Pardo, R. (2008). The Evaluation and Optimization of Trading Strategies (2nd ed.). John Wiley & Sons.