Your strategies
Unlimited strategies, walk-forward analysis, and the signal research lab.
What do you want to test?
Each option below builds a complete, runnable strategy -- indicators, signals, and rules. Everything stays editable on steps 2-4 afterwards, the same as anything added by hand.
Cointegration pairs trading
Scan a universe of tickers for pairs whose price spread is stationary (mean-reverting), then build a real long/short pairs-trading strategy on whichever pair you pick: long the cheap leg and short the rich leg when the spread's z-score gets stretched, closing both when it reverts.
Same-sector tickers are the most economically sensible candidates -- e.g. a handful of oil majors, banks, or beverage companies.
Initialize the strategy
Download price history and create the portfolio, account, and order book this strategy trades through. Everything downstream is scoped to what you set here.
The portfolio tracks positions per symbol. The account tracks cash and equity across portfolios. Rules write orders into a third store -- the order book -- which is why a backtest resets all three before each run.
Add indicators
An indicator turns price history into a new column. Signals then compare those columns.
Added so far
Strategy JSON
Add signals
A signal is a true/false column: the moment something you care about happens. There are six signal types to choose from.
Added so far
A crossover fires only on the bar the relationship changes, not on every bar it holds. That is what separates a Crossover signal from a Comparison signal -- and why an entry rule built on a comparison re-fires every day.
Add rules
A rule turns a signal into an order. Entry rules open positions, exit rules close them, and chained rules react to a parent's fill.
Added so far
Trade statistics
Price chart & signals
Most recent bars
Trade order statistics
Order management
Multi-symbol signal scan · experimental
Parameter optimization
Vary one argument of an indicator or rule across a set of candidate values, then run the full backtest once per combination. Only numeric arguments can be swept -- a column or series argument would be silently replaced by a bare number.
Robustness & walk-forward
Two questions about the optimization result. First, how much of it is luck given the number of trials. Second, whether a parameter chosen on past data holds up on the data that followed.
These diagnostics run against the trial portfolios from a parameter optimization run. Build a paramset on the Optimization screen and run it first.
Rolls a training and testing window across the whole date range: for each training period, the optimizer picks the combination with the highest net trading P&L, then applies exactly that combination out-of-sample on the following testing period.
Signal research lab
Study what a signal is worth before wrapping rules around it. For every bar the signal fired, look at the returns of the days that followed. These are post-hoc analysis tools, not signal generators.
Research tools, not signals
Every plan runs the same real backtesting engine. Paid plans add the tools that test whether a result survives outside the window it was fitted in.
Agent Mode
A multi-agent research loop: backtests your current strategy, has independent specialist agents analyze and critique it, proposes a small revision, and repeats -- then validates the best version once on data it never saw during refinement.