`RunStrategyBacktest`

Runs a compiled NinjaScript strategy through NT8's Strategy Analyzer engine for supported configurations. With optimization, it enumerates a parameter grid. This is a historical job, not deployment of a strategy to a trading account.

Property	Value
Required scope	`mcp:trade`
Execution	Asynchronous; poll `GetMcpJob` with the returned `job_id`
Account	NT8's isolated `Backtest` simulation account
Prerequisite	The strategy class must be available in NT8's compiled Custom assembly

Prepare the class

A successful CompileNinjaScript(in_memory: true) validates source in the sandbox. Use WriteNinjaScriptFile to save the approved file, then inspect compile_engine. If it is file_only, press F5 in the NinjaScript Editor and resolve errors before backtesting. A sandbox compile alone does not load a new strategy into NT8's normal compiled strategy path.

When iterating, use versioned class names such as MyStrategy_v1 and MyStrategy_v2. Older add-ons may resolve a previously loaded type after recompilation; newer versions prefer fresher assemblies, but versioned names also make the evidence easier to audit.

Single-backtest request

This is a smoke-check configuration using NT8's built-in SampleMACrossOver. It uses no commission and zero slippage. Confirm the class, instrument, trading-hours template and historical data exist in your installation. Use appropriate costs for a performance assessment.

{
  "strategy_class": "SampleMACrossOver",
  "instrument": "MES 06-26",
  "bars_period": { "period_type": "Minute", "value": 5 },
  "from": "2026-04-01T00:00:00Z",
  "to": "2026-04-30T23:59:59Z",
  "trading_hours": "CME US Index Futures ETH",
  "parameters": { "Fast": 10, "Slow": 25 },
  "fill": {
    "type": "Standard",
    "slippage_ticks": 0,
    "commission_template": ""
  },
  "account": { "initial_cash": 100000, "denomination": "UsDollar" },
  "timeout_ms": 120000
}

The current field names are strategy_class and bars_period.period_type. Put costs and fill settings in fill. The account argument is an overrides object, not an account-name string. Do not send "account": "Sim101".

Argument	Meaning
`strategy_class`	Bare name of the compiled strategy class
`instrument`	Full instrument name, including the contract for futures
`bars_period`	Primary bar type and value; optional `value2` and `market_data_type` where applicable
`from`, `to`	ISO-8601 UTC timestamps
`trading_hours`	NT8 template name; omitted means the instrument's primary template
`parameters`	Public writable strategy-property overrides; use that class's actual names
`fill`	Order-fill resolution, slippage ticks and commission-template name
`account`	Initial cash and denomination for the isolated Backtest account
`timeout_ms`	Optional elapsed-time ceiling for the job

Poll for the result

The initial call returns a job reference. Call GetMcpJob with its job_id. Pending or running is not a completed test. Read the terminal status and returned result; retain errors when a job fails or is cancelled. For a successful single run, inspect the metrics and trade detail actually returned by your installation.

Do not fill a response template with assumed profit, drawdown or trade counts. For an engine comparison, match the compiled class, data, dates, trading hours, parameters and supported fill/cost settings in NT8's UI, then compare the two observed result sets.

Parameter sweep

Add an optimization object to a complete backtest request. For a custom class declaring FastPeriod and SlowPeriod, this produces nine combinations:

{
  "optimization": {
    "fitness": "MaxNetProfit",
    "parameters_sweep": [
      { "name": "FastPeriod", "min": 8, "max": 10, "step": 1 },
      { "name": "SlowPeriod", "min": 20, "max": 22, "step": 1 }
    ]
  }
}

Those names are examples for a custom strategy. SampleMACrossOver uses Fast and Slow instead. Unknown property names return an error with the available names.

Sweeps enumerate Cartesian combinations. The optimizer name is metadata; GeneticOptimizer and StrategyGenerator do not change the loop into those algorithms. Only numeric parameters are sweepable. Fixed parameters are applied to every iteration on add-on 1.13.8 and later, with swept values taking precedence. Older versions did not reapply fixed parameters during a sweep.

Inspect the returned ranking block and each row's fitness_value. If the requested fitness cannot be driven on that NT8 build, the response identifies its NetProfit fallback. keep_best_results and instantiated_on_each_iteration do not control the Cartesian loop. Full per-trade detail is omitted from sweep results; run the selected parameter_values as a separate single backtest to inspect the trades.

Supported-settings boundary

Tick Replay is not exposed by this MCP interface. fill.type maps to OrderFillResolution. A Tick primary bar series is a tick-bar backtest. Neither that bar type nor High fill resolution enables the Strategy Analyzer's Minute-plus-Tick-Replay mode.

Backtests always use the isolated Backtest account. If that account is unavailable, the tool fails rather than falling back to Sim101 or a live account. A separate DeployStrategy call is required for forward simulation.

Failure handling

Reported condition	Next step
`nt8_build_unsupported`	Inspect the missing-symbol details and NT8/add-on compatibility
`no_isolated_backtest_account`	Restore/verify the NT8 Backtest account; do not substitute a trading account
`bad_parameter`	Match the returned list of valid strategy-property names
Timeout or `backtest_aborted`	Inspect the job result and reduce the request size or revise the elapsed-time ceiling
Missing class or historical data	Verify the NT8 rebuild and data availability before retrying

Workflow

Define rules, inspect symbols, compile the draft, approve the file write, verify the NT8 rebuild, then run a fixed test. Review trade behavior before a bounded sweep. Evaluate selected parameters on separate data. Only then consider a separately approved simulation deployment and verify it through GetDeployedStrategyState.

For prompts, the exact capture checklist and the optional Sim101 step, follow Build an AI NinjaTrader Strategy: Compile, Backtest and Simulate.

Prepare the class​

Single-backtest request​

Poll for the result​

Parameter sweep​

Supported-settings boundary​

Failure handling​

Workflow​

Related tools​