Documentation · Workflow

The workflow in practice

Data, decide, build, trade: one worked path through all eight modules, and the checks that keep the backtest and the live terminal reading the same numbers.

  • DATA
  • DECIDE
  • BUILD
  • TRADE

Four stages, one set of numbers

Data, decide, build, trade. Each stage hands the next one something it can check. The same path, shorter and with the tabs named, is on the workflow page.

StageQuestionModule(s)What it hands on
DATADoes the symbol in SQX describe what the terminal trades?SpecSmith, GapSmithA broker profile that matches the terminal, and a symbol without short M1 holes.
DECIDEIs there anything to build here, and in which shape?Edge FinderA measured side, symmetry, style, block list and optional time range, written into a Build task.
BUILDWhich rules survive?StrategyQuant X (Builder, retest, walk-forward matrix, robustness checks, export), NameSmith, and the input modules that add blocksStrategies that are named, built on inputs the terminal will also have.
TRADEDoes the terminal read what the backtest read?MetaTrader 5The same history, model weights, calendar and alt-data rows.

You do not need every module. The sections below show the whole path so you can see where each one fits.

1. Data: make the symbol match the terminal

  1. Export the broker’s specifications

    The AWSpecFeed service in the terminal logged in to your broker does this by itself once the terminal has connected, and then once a day. It places no order and writes two CSV files to Common\Files\AWSpecs. If it is not running, drag the AWSpecExport script onto a chart instead.

  2. Compare and apply

    In the SpecSmith tab of the Data Manager, choose the export and a broker profile, compare, tick the rows you want and apply. The values they replace are saved first, and Undo restores them. SpecSmith does not guess the broker’s clock, so set the timezone yourself; its Timezone check then tests your value against a currency pair’s weekly reopening and reports CONFIRMED, MISMATCH, DST-SHIFTED or CANNOT VERIFY, without changing it. Its separate Fix timezones button checks M1 history whose weekly reopening is one hour off its label.

  3. Fill the holes

    Check the M1 history under Analyze data quality. GapSmith fills genuine gaps (never weekends, holidays or closures), flat by default, up to the maximum gap length you set, into a _patched symbol and leaves the original alone; an optional synthetic fill goes into a separate _synth symbol. Decide which symbol you will build on and use the same one from here on.

2. Decide: measure before you build

  1. Fix the data and reserve bars

    In Edge Finder’s Data tab choose the symbol, timeframe and a range of several years, and add an out-of-sample range under Data range parts. A Build task uses the same control, so what you reserve here is reserved there.

  2. Run Autopilot, or each tab

    Choose Everything to find the best timeframe, or Current timeframe if you have already chosen one. Read the q grades and the Held out column before the means. Time Windows needs a data clock; read the line at the top of its result before you trust an hour.

  3. Deploy to a Build task

    On Deploy, pick the project, confirm the side, symmetry, style and time range, and Apply. The project must not be running.

Treat the result as a place to look, not a forecast. Costs are applied only if the Data tab holds them, and stops, targets and sizing are not modelled.

Add inputs before you measure or build

Block Edge scans every block SQX has registered, so install the input modules first if you want their blocks measured.

Structure and flow

Tick the Wyckoff and other Institutional blocks (state, event and level blocks) in the Build task. Use shift 1 so the rules read closed bars.

Non-price series

In Alternative Data, paste the FRED and EIA keys, press Update all, and tick the AW AltData blocks. Percentile and range blocks make good regime filters. If the series will matter live, leave the export on and run the AWAltDataFeed service on the trading terminal.

Release timing

In News, start the AWNewsFeed service in MetaTrader, press Import once in SQX (the file is copied back to MetaTrader automatically), and check the measured clock and the forward window. Tick a blackout block on an M1 to M15 chart, or the news-bar blocks on higher timeframes. Set Price data timezone if your data has no declared timezone.

Trained models

In NN Lab, train on the same symbol and timeframe with the same reserved range. On the Models tab press Deploy to Builder for the Build task (it ticks the NN blocks and sets ModelId to the model’s ID as a fixed value) and Deploy to MetaTrader 5. A model you have just trained needs no sync, compile or restart.

3. Build: let StrategyQuant do the searching

  1. Run the Build task as usual. The suite adds inputs and a starting configuration; it does not change how SQX evaluates or ranks strategies.
  2. Keep the task’s out-of-sample range inside the window you reserved in Edge Finder and NN Lab. A model or measurement that saw those bars invalidates the validation.
  3. Run SQX’s own retest, walk-forward matrix and robustness checks on the survivors. The suite measures and supplies; whether a strategy holds up is StrategyQuant’s verdict.
  4. Use NameSmith to name the survivors before you export, so the symbol, timeframe and style are in the name.

4. Trade: hand over to MetaTrader 5

  1. Export the strategy

    Export as a MetaTrader 5 EA. Only MT5 and PseudoCode are functional targets for these blocks.

  2. Compile StrategyQuant’s indicators once

    The installer already put the AW indicators in the terminals you ticked. What it cannot supply are StrategyQuant’s own Sq* files: copy them into MQL5\Indicators and compile them once (see Install and licence).

  3. Deploy NN models, if you use them

    Press Deploy to MetaTrader 5 on NN Lab’s Models tab so the model files are in Common\Files. The alternative-data files and events.csv reach Common\Files without a step from you: the feed services write them, and SQX copies the News calendar there after each import.

  4. Check the services

    AWAltDataFeed, AWNewsFeed and AWSpecFeed are registered by the installer and start with the terminal. On a terminal that must keep itself current, confirm they are running under Navigator → Services. The Alternative Data hosts are added for you when you tick the service.

  5. Run the exported EA in the Strategy Tester first

    Compare its trades with the SQX backtest on the same history before going live (the checklist below says what to compare).

The parity checklist

When a backtest and a terminal disagree, the cause is almost always one of the following. Work down the list.

CheckWhat to compareWhere
Same instrument descriptionTick size, point value, swap, commission and session.SpecSmith
Same historyBars, gaps and the data source. A hole in M1 history changes where a stop fills.SQX Analyze data quality; GapSmith
Same clockThe broker timezone of the price data, and the calendar clock measured by News. An hour of error moves every blackout.News tab calibration line; Price data timezone
Same filesThe alt-data feature files, events.csv and the model files are the same bytes SQX used, and are current.Alternative data tab; AWAltDataMonitor; NN Lab Models tab
Same indicatorsEvery AW indicator and Sq* file the EA calls is in MQL5\Indicators and compiled. This matters most on a terminal the installer did not touch, such as a VPS.Terminal Experts log
Closed barsRules read shift 1, not the forming bar. The AW indicators step closed bars only.Block parameters
Frozen versus live dataThe feed services do not run in the Strategy Tester, so a test reads frozen files; live they extend the files forward.Alternative Data service
Costs and fillsSpread, commission, swap and slippage assumptions in SQX versus the account the EA trades.SQX Build task; MT5 account
LicenceThe computer the terminal runs on is covered, and the updater task is enabled so the licence keeps renewing.Your licence

What stays outside the suite

  • Costs, sizing and risk rules belong to the strategy and to your account. If you trade a funded account with a daily or trailing drawdown rule, the suite does not model it; set it up in SQX and in the EA.
  • Robustness testing is StrategyQuant’s. Run its retest, walk-forward matrix and the other checks on the strategies the builder produces.
  • Data supply. The history you trade on comes from your data provider and broker. The suite checks and, where the rule allows, repairs short gaps; it does not supply history.
Risk

Matching the backtest to the live terminal removes one class of surprise. It does not make a strategy profitable, and no part of the suite predicts results. Trading leveraged instruments carries a high risk of loss.