User guide / No trading experience needed
Learn to test
your first strategy.
Start with one example. Learn what its rules mean, run a test, and understand the result before you take the next step.
01 / Start here
What you are building
A strategy is a set of rules for when to buy, when to sell, and how much to use. AI Algo Factory lets you build those rules, test them on past prices, and watch them in a simulated account.
Entry and exit
An entry starts a trade. An exit ends it. A position is the amount you still hold.
Long and short
A long trade aims to gain when price rises. A short trade aims to gain when price falls.
Market and symbol
The market is what you test. Its symbol is the short name you select in the editor.
Bar and timeframe
A bar groups prices over a set period. With Daily selected, each bar covers one trading day.
You can read Explore and public results without an account. Sign in to save a strategy, run tests, or manage your work.
02 / First strategy
Build one example, step by step
Use the EMA 9/21 example to learn the controls. Its settings are a teaching example, not a recommendation to trade.
Before saving, know that your strategy's results will be public. Its rules stay private unless you share them.
Open New strategy and choose the example
Sign in, choose New strategy, then select EMA 9/21 example under Start from. Give it a name you will recognise, such as My first practice strategy.
Choose the markets
The example starts with AAA and BBB as placeholders. Replace both with supported symbols from the market picker, or remove the rows you do not need. Keep Daily for this exercise. Wait for the chart to load.
Read the entry and exit
EMA means exponential moving average. It smooths prices and gives recent prices more weight. The 9-period line responds faster than the 21-period line.
This example enters a long trade when the 9 EMA crosses above the 21 EMA. It exits when the 9 EMA crosses below it. A crossover is the moment the two lines change order, not every bar on which one stays above the other.
Check protection, position size, and costs
The example also has a 5% stop loss and a 10% take profit. A stop loss asks the simulation to exit after an adverse price move. A take profit asks it to exit after a gain.
Position size is how much of the account a trade can use. A 5% stop on a position is different from a 5% loss on the whole account.
Fees are charges for trades. Slippage is the difference between the expected price and the fill price. A fill is a completed order. The fields use basis points, shown as bps. 100 bps = 1%; 1 bp = 0.01%.The example's costs are sample assumptions.
Next bar openmeans a signal from one bar leads to an order at the start of the next bar. With daily bars, that means the next trading day's open.
Check the preview and Ready to test?
The preview shows one selected market at a time. Read the entry and exit markers, return, worst drop, and trade count. A rising price chart alone does not tell you whether the strategy made money.
Ready to test? checks the rule setup. Its Proof score is not a success probability or a profit forecast. Read any warnings before you continue.
Save, then open the first test
Choose Save. This records version 1 and opens its strategy page without starting tests. Open Feasibility and choose Run feasibility. Read the result before you try the rest of the tests.
You have completed this exercise when you can find your saved version and explain the first test's result. A failed test is still useful information.
Start from your own idea instead
Use Describe it instead and choose Fill the blocks. Include the markets, entry, exit, position size, protection, costs, and order timing. Read every filled block and any unsupported items before saving. You can also start with Blank, NNFX, or Krown. Those templates contain more rules.
If you change a template after editing, Replace it replaces the rules in the editor. Keep editing keeps your current draft.
03 / Read results
Understand the numbers
A backtest simulates how saved rules would have behaved on past prices. It is evidence about that test, not a forecast of future profit.
Illustration only · not a recorded strategy
- Start$1,000
- Peak$1,200
- Low after peak$1,080
- Finish$1,120
The account finished 12% above its start. Its largest fall was 10%, from $1,200 to $1,080. These describe different parts of the same result.
- Return
- The gain or loss relative to the starting value. Check which test produced it and how much time the test covers.
- Worst drop, or maximum drawdown
- The largest recorded fall from a previous account peak. It can remain large even after the account recovers.
- Equity
- The account's value, including cash and open positions. An equity curve shows how that value changed.
- Trade count
- How much trading the test recorded. A result from a few trades gives you less evidence. Check the report: paper counts may represent filled orders rather than complete entry-and-exit pairs.
- MAR ratio in this app
- The displayed return divided by the size of the worst drop. In the illustration, 12 ÷ 10 = 1.2. This app uses its displayed return, not an annualised return. A higher number means more recorded return per unit of drawdown. It does not prove a strategy is safe or steady.
- n/a
- The app does not have a usable value. It does not mean zero. MAR stays unavailable when return or drawdown is missing, or the worst drop is below 0.5%.
Check where the numbers came from
Explore uses paper-session results for the current version when available. Otherwise it uses walk-forward results, then feasibility results. A walk-forward headline can be an average return across test periods, paired with their worst drawdown.
Compare the version, dates, markets, costs, and result source before comparing ranks. A short paper session and a long historical test do not cover the same experience. The small card chart may use feasibility history when there is no paper curve, even if the headline numbers come from walk-forward testing.
04 / Run tests
Know what each test checks
Open a stage on the strategy page to see its status and report. Passedmeans the version met that stage's checks. Didn't pass means a check failed. Needs your review means you need to inspect the result.
Feasibility
A first historical test. Check that the rules produce trades and usable results after the assumed costs.
Walk-forward
Work forward through time. Choose settings on older prices, then test them on later prices that were not used to choose those settings. This helps reveal rules that fit the past too closely.
Best settings
An optional scan for supported strategies with adjustable settings. Read whether nearby settings also work, not just which combination scored highest. The web app's Run all tests action skips this optional scan. Open the Best settings stage to run it when available.
Robustness
Test worse conditions, such as higher costs or different return sequences. Open the report to see which checks raised concerns.
Paper trading
Use new market bars with a simulated account. Follow its trades and alerts without placing real orders.
Run the required tests together
Save and run all testssaves the editor's rules and starts the test chain. From a saved strategy, choose Run all tests. It runs feasibility, walk-forward, robustness, and a final evidence check before paper trading can start.
The chain stops if a required check fails, a review is needed, or a service error occurs. If it stops, the button can read Run again. Passing every required check can start paper trading automatically. It cannot start live trading.
Keep track of versions
Edit strategy opens the saved rules in the same editor. Each save creates another version. Older reports remain in the history and may still appear with their version label. Test the new version again. An existing paper session keeps the version it started with.
Back to guide contents ↑05 / Find and share
Find your work and explore others
| Open | Use it to |
|---|---|
| Explore | Browse public results. Filter by stage or market type, such as Stocks or Crypto. Rules shared finds readable rules for signed-in members. Mine appears when signed in. |
| Strategies | Find your own saved work. Use All, Needs you, Testing, Paper trading, or Stopped. |
| New strategy | Build rules in the editor, use an example, or describe an idea. |
| A strategy page | Read results. If you own it, edit the rules, run tests, review findings, or change sharing. |
| Settings | Manage test profiles, access for outside tools, and billing. |
Explore ranks by MAR ratio by default. You can also choose Return, Longest paper trading, or Newest. Newest sorts by the latest update. Use the page controls to see more results.
Understand sharing before you save
A saved strategy's results are public, including to people who are signed out. Your rules stay private unless you choose to share them with signed-in members.
| Who is viewing? | What they can do |
|---|---|
| Signed-out visitor | Read public results and market facts. |
| Signed-in member | Read public results. Read shared rules and clone them when the owner allows access. |
| Owner | Read the full workspace, edit rules, run tests, record decisions, and change sharing. |
Clone to my strategies opens a separate draft for you to save and test. It does not change the original. Turning sharing off removes access through the app; it cannot remove copies someone already made.
Back to guide contents ↑06 / When a test stops
Work out what needs your attention
Open Strategies → Needs you, then open the strategy. Read the stopped stage, its version, the report, and any error message before choosing an action.
- Accept
- Record that you reviewed the result when this action is offered. Acceptance does not repair the rules or automatically continue the test chain.
- Repair
- Record that the strategy needs changes. Use Edit strategy to make those changes, save a new version, and run the tests again.
- Discard
- Archive the strategy and keep its evidence. An archived strategy cannot start new tests.
No trades in the preview
Check that the chart loaded the intended symbol and timeframe. Check whether the entry conditions can occur together in that history. An empty trade list is different from unavailable data.
The chart or test service is unavailable
Read the message, confirm the selected market is supported, and retry later. If the message says the service is not configured, contact the person who manages the app. Changing trading rules will not fix a missing service.
A report shows an older version
Read its version label. Open the matching stage and rerun it for your current saved version. Old evidence does not verify changed rules.
Save is unavailable
Read the validation messages beside Save and correct the fields they name. If an imported or AI-generated rule does not fit the editor, resolve the listed issue before saving.
07 / Paper trading
Follow the simulated account
Open Strategies → Paper trading. A session shows the version it uses, account value, return, positions, and alerts. Its orders and fills are simulated.
The app checks for new market bars on a schedule. This is not a live tick-by-tick feed. A daily strategy will not produce a new result on every page refresh.
- Pause
- Pause session processing. Inspect the account and alerts before continuing.
- Resume
- Continue a paused session.
- Stop
- End that session. Resume does not restart a stopped session; start a new session from the strategy page when available.
The app can pause a session when it reaches its drawdown limit. Review the alert before you resume. The displayed day count is elapsed session time, not a count of trades or trading days.
Combine strategies in a portfolio
A portfolio groups paper-trading strategies. A weight is each strategy's share of the proposed allocation. For example, equal weights divide the allocation evenly.
Use New portfolio in the Paper trading view once strategies are available. Review shared markets and correlation, which measures how closely returns move together. Two strategies can have different rules and still lose at the same time. A portfolio report does not enable live trading.
Back to guide contents ↑08 / Settings and terms
Learn the extras when you need them
Gate profiles
A gate profile is a set of pass criteria for robustness tests. The available profiles include Default, Trend following, and Mean reversion. Settings shows their criteria. A different profile changes how the evidence is judged; it does not improve the underlying trading result.
Agent keys
An agent key lets an outside tool act with the access you give it. You do not need a key to use the app yourself. If you connect a tool, set its permissions, daily submission limit, and expiry in Settings. Copy the key when shown; the app shows the secret only once. Revoking it stops future access through that key.
More terms you will see
- Indicator
- A calculation from market data, such as a moving average. It helps define a rule; it is not a promise about the next price.
- AND / OR
- AND requires all conditions in its group. OR can be satisfied by an alternative condition. Check the entry and exit groups separately.
- Risk limit
- A rule that restricts position size, exposure, or losses in the model. It does not remove market risk.
- Exposure
- How much of the account is committed to positions.
- Overfitting
- Choosing rules that match past prices too closely and do poorly on other data.
- Out-of-sample
- Data kept separate from the data used to choose the settings.
For more on real order behaviour, see Investor.gov's guide to order types. A real stop order does not guarantee the eventual fill price.
Back to guide contents ↑