RealTest: What It Is and What It Actually Does

RealTest is Windows software for backtesting systematic trading strategies at the portfolio level. It is written and maintained by Marsten H. Parker, a full-time systematic trader of more than twenty years who appears in Jack Schwager's Unknown Market Wizards. The software began as his own research tool and is still developed by one person.

That last fact explains most of what follows. RealTest is not built to look impressive in a demo. It is built to answer one question, which is whether a set of rules would have survived a long stretch of real market history once costs, capacity and portfolio limits are applied.

What portfolio-level testing means

Most retail backtesters answer a narrower question, which is how one set of rules performed on one symbol. You get an equity curve for AAPL, then another for MSFT, and nothing tells you what happens when both signals fire on the same morning and you only have cash for one.

RealTest starts from the portfolio. You give it a universe, a capital allocation and a set of limits, and it walks forward bar by bar deciding which signals it can actually take. When ten stocks qualify and you allow five positions, it ranks them, fills what fits and skips the rest. The equity curve you get at the end is the one a single account would have produced.

The same structure lets several strategies run inside one account. The vendor describes it as combining strategies that differ by side, strategy type, markets traded and bar size. In practice that means a long mean reversion system, a short system and a monthly rotation can share one pool of capital, compete for it, and be measured together. Correlation between your own strategies is visible in the result rather than assumed away.

What else it does

Walk-forward optimisation

RealTest treats parameters as first-class objects through a Parameters section, and its optimiser can run walk-forward, meaning it fits on one window and tests on the next unseen window, repeatedly. This matters more than any other feature on the list, because a single-pass optimisation over the whole history tells you almost nothing. We covered why in the five backtesting mistakes that fake an edge.

Monte Carlo and trade-level analysis

The software includes Monte Carlo analysis, interval testing, and trade plots covering scatter distributions, MAE and MFE distributions, and a cumulative trade-level profit curve. Those plots answer the question a summary table cannot, which is whether the result came from the whole sample or from six good trades.

Order generation

RealTest produces the next session's order list in text or CSV, with templates for the Interactive Brokers Basket Trader and The Chartist Smart API. It can also import your actual fills from an Interactive Brokers Flex Query and compare them against what the backtest expected. That comparison is the only honest way to find out whether your live slippage assumption was right.

Survivorship handling

With Norgate data imported, index membership becomes a testable variable. Variables like InSPX, InRUI and InNDX tell you whether a stock was in the index on that date, not whether it is in the index today. Testing on today's S&P 500 members back to 2000 removes every company that failed, without telling you, which inflates results by a few percent a year on large caps and considerably more on small caps.

What RealTest does not do

It is not a charting package. It draws price charts with your trades marked on them, and that is where the visual work ends. It is not a broker and it does not place orders by itself. It has no scanner for discretionary setups, no news feed, no social features, and no cloud version. It runs on your machine and reads data you supply.

It also has no opinion about whether your idea is any good. RealTest will happily test a strategy with fourteen optimised parameters and report a beautiful curve. The discipline is yours.

How it compares with the alternatives

The honest summary is that TradingView and RealTest are not competitors. Plenty of people run both, using RealTest to decide what is worth trading and TradingView to watch it during the session. Amibroker is the closest comparison, since it also does portfolio-level testing on Windows with its own language. Python gives you everything and hands you the bill in build time, which is fine if writing infrastructure is part of what you enjoy.

The scripting language

RealTest scripts are plain text files made of named sections with tab-indented items. The vendor is careful to say it is not a programming language, and that framing is accurate. There are no loops to write and no state to manage, because the engine already knows how to walk bars, rank candidates and manage positions. You describe the strategy and it supplies the machinery.

Here is a complete, runnable script. It is the 50/200 day moving average crossover on SPY, chosen because it is the most familiar rule in the business and there is nowhere to hide in it.

Settings:
	DataFile:	SPY.rtd
	StartDate:	Earliest
	EndDate:	Latest
	AccountSize:	100000

Data:
	Fast:	MA(C, 50)
	Slow:	MA(C, 200)

Strategy:	SMACross
	Side:	Long
	MaxPositions:	1
	QtyType:	Percent
	Quantity:	100
	Commission:	Max(1, 0.005 * Shares)
	Slippage:	0.001 * FillPrice
	EntrySetup:	Cross(Fast, Slow)
	ExitRule:	Cross(Slow, Fast)

That is the entire system, and it runs as written. On SPY from 29 January 1993 to 10 July 2026 it produced 15 trades, a 10.02% annualised return, a 33.30% maximum drawdown, a MAR of 0.30, a 0.78 Sharpe ratio and a win rate of 86.67%, with the account exposed to the market 74.68% of the time. Thirteen winners out of fifteen sounds excellent until you notice that a third of the account went missing along the way, which is the sort of thing a summary table hides and a drawdown figure does not.

Three details in there matter more than the crossover rule itself. Commission and Slippage are ordinary formula items rather than an afterthought in a settings dialog, so a costed test is the default rather than something you remember to switch on. Entry timing defaults to the next open, so the signal is calculated on the close and filled the following morning, which is what you could actually have done. And the Data section is calculated once for every bar before the test starts, which is why a universe of thousands of symbols stays fast.

Scaling that same script from one symbol to a universe is where the portfolio machinery appears. You add an Import section naming a Norgate watchlist, raise MaxPositions, and add a SetupScore formula to rank candidates when more of them qualify than you have room for. The strategy logic does not otherwise change.

The learning curve is real but short. Most people who already think in rules are writing working scripts within a week. If you want a finished script to read rather than a blank file, our free RealTest strategy is a complete file with the entry logic, sizing and cost handling in place.

What it costs

As of August 2026, a new licence is $389 USD, which covers lifetime use of the current version plus every release and update for the following twelve months, and allows activation on two computers for the same person. After that year, a $159 USD extension adds another year of updates. Letting it lapse does not disable the software, it stops the updates. There is a 30-day free trial with no payment required. Current figures are on the RealTest purchase page.

Data is separate and usually costs more than the software over time. Norgate is the common pairing for US, Australian and Canadian stocks, largely because of the point-in-time index membership. RealTest also imports from CSI, Tiingo, EOD Historical Data, Yahoo, MetaStock and plain CSV files, so you can start on free data and move up once the strategy is worth the subscription.

What you need to run it

RealTest is a 64-bit Windows application. It runs on a Mac through Boot Camp or Parallels with a full Windows installation, and on a cloud virtual machine, though the documentation states that API-level emulators such as Wine and Crossover are not supported. A screen of at least 1920x1080 is expected. Memory is the constraint that bites, since a set of strategies covering all US common stocks over ten years runs in 4GB, while going further back or including delisted symbols wants 16GB.

Who it is for

RealTest is a good fit if you trade a rules-based system on stocks or ETFs, you care about what a portfolio of strategies does rather than what one symbol does, and you are willing to write a script instead of clicking a strategy builder. It suits people who have already been burned by a backtest that looked perfect and then lost money, because most of the software's design is aimed at exactly that failure.

It is a poor fit if you trade discretionarily, if you need a chart-first workflow, or if you want a hosted service that runs strategies for you. Those are real needs and other tools serve them better.

How to start

Take the 30-day trial and import a small universe first, something like twenty liquid ETFs on free data, rather than the full Russell 3000. The import step is where most first attempts stall, and it is much easier to debug on twenty symbols. Write one simple strategy with three or four parameters and read the trade list rather than the summary statistics. Then run the same script with commission and slippage set to zero and compare, which tells you immediately how much of the result was ever available to you.

Once the mechanics are familiar, the useful next step is validation rather than more ideas. Walk-forward and Monte Carlo are built in, and they are the tools that separate a strategy from a coincidence. If you want to see the process applied end to end, we walked through it in the guide to starting algorithmic trading, and the failure modes to watch for are in how to tell when a strategy has stopped working.

We build and publish RealTest strategies for a living, with the scripts, the backtest reports and the assumptions included, so you can read the code rather than trust a screenshot. The full set is in the RealTest strategy collection.