Strategy research

Search a strategy's parameters, then test the chosen setting on data it was not chosen on. This is a research tool: it places no real orders, and no result here is evidence that a strategy is profitable. The best of many tried settings usually looks better than it is. The built-in strategy is a demonstration only: it is not the trader's strategy.

1. Strategy and parameter ranges

Strategy

DEMONSTRATION STRATEGY: illustrative only. Any result is a plumbing check, not a finding about a real strategy.

Only parameters the strategy exposes for research can be searched. Tick Search to try a range; otherwise the value in the base parameters is used. The bounds shown are sanity limits, not recommendations.

Parameter	Allowed	Values
Lookback bars (LookbackBars)	1 to 1000, whole numbers	20, 21, 22, 23, 24
Stop % of price (StopPercent)	0.05 to 20, decimal	1, 1.05, 1.1, 1.15, 1.2
Target (R multiple, 0 = none) (TargetRMultiple)	0 to 10, decimal	fixed by the base parameters
Max trades per symbol per day (MaxTradesPerDay)	1 to 20, whole numbers	fixed by the base parameters

Base parameters (JSON). Searched parameters are replaced per combination.

2. Data: symbols and periods

Each line is a symbol and the period it is traded in, as on the Portfolio page. Lines from a saved scan keep a link to that scan so its filters can be checked for look-ahead. Bars come from the local store and are downloaded from Massive only when missing.

Add the matches of a saved scan

No saved scans yet (save one on the Scanner page).

Or type symbols (comma or space separated)

From (New York date)

Nothing added yet.

Bars

Regular trading hours only

3. Simulated account, costs and objective

The values are examples, not any broker's rules or the trader's risk policy. The same shared-account engine as the Portfolio page runs every combination.

Initial capital

Position sizing

Sizing value

Max concurrent positions

Slippage (% of price)

Commission per share ($)

Minimum per order ($)

Close open positions at the end of a period

Objective: rank combinations by

Minimum closed trades to be ranked

Net P&L: realized result after all costs. Ignores risk taken and how many trades produced it. A configuration with too few trades is never ranked: doing nothing cannot win by avoiding losses.

4. How the result is checked

Warm-up: show each period the history just before it, so indicators are warm when it starts (as in real time)

Only earlier data from the same line is used, never later data. Without it every period starts cold: a strategy that needs N bars of history cannot act on the first N bars of each period. It applies only to strategies that say how much history they need.

Everything is tested on the same data it is searched on. The result only shows which setting fitted this period best. It is not evidence for anything else.

5. Size of the experiment, check, run

25 combination(s) · 25 backtest(s) in total · warm-up: 24 bar(s) before each period · limits: 2,000 combinations, 200 values per parameter, 2 at a time

Add at least one symbol and period.

No out-of-sample validation: every result is in-sample (the best fit to this data), not evidence of an edge.

Save this experiment

Saved experiments

A saved experiment keeps its definition, the engine and strategy versions, a fingerprint of the data of every line, and the results. Market data is not copied into the database. Open shows the stored result without simulating anything. Repeat runs it again on today's data and says what is the same and what changed. A partial experiment was cancelled before it finished.

No saved experiments yet.