Changelog

Every release, by version and by product

Releases are listed newest first, each grouped by product. Per-product version numbers started with 1.1.3; before that every product carried the release number. Behaviour change means strategy results can differ from the previous version. Entries are written from the repository’s own changes; where we could not check a detail, we left it out.

  • 1.0 to 1.1.6
  • By version and by product
  • Behaviour changes are labelled
Index

Find a release or a product

Jump to a release:

1.1.6 Installer 1.1.2 1.1.5 1.1.4 1.1.3 1.1.2 1.1.1 1.1 1.0

Release 1.1.6 · 2026-11-20

Hotfix: SQX grids show what SpecSmith and GapSmith wrote, zero spread floor, commission on entry and exit

ProductVersionIn this release
GapSmith–The Data tab shows the broker profile of a new copy right away; it kept showing SQ default until StrategyQuant X was reloaded
SpecSmith–The grids refresh after Apply, Undo and the timezone fix; a measured spread of 0 is written as 0.1 pip; new option Commission on entry and exit; clearer copy or overwrite choice in Fix timezones

All other packages (and the installer, now 1.1.3) were rebuilt with the release date and have no change in what they do. The updater will download them again for that reason only.

SpecSmith

  • Fixed: values that looked unrounded or unequal. SpecSmith writes to the broker profile on the server, and the Instruments grid in StrategyQuant X kept the values it had loaded, so a spread or swap that had been written correctly could still look unrounded or unequal until you reloaded. The grids now refresh after Apply, Undo and the timezone fix.
  • Spread of 0. When MetaTrader 5 measures a spread of 0 (typically a raw-spread account that charges commission instead), SpecSmith writes a floor of 0.1 pip, so a test never runs with a free spread, and replaces a stale figure. Set Spread to Leave as it is to keep your own value.
  • New option, Commission on entry and exit. StrategyQuant X charges a commission once, at the open. By default SpecSmith follows your closed trades: the MetaTrader 5 export now counts, per symbol, the entry and exit deals of the connected account and how many carried a commission, and the commission is entered as a round trip only when both sides were charged. With no closed trade that carried a commission it follows the broker’s rule. You can also choose the broker’s rule or the per-side figure as reported. The row’s detail shows the deals counted. This replaces the advanced round-turn checkbox.
  • The Fix timezones dialog shows the two choices, create a copy or overwrite, as two clearly separated options.

Manual →

GapSmith

  • Fixed: after Fix gaps created a copy, the Data tab kept showing the copy on SQ default until StrategyQuant X was reloaded, although GapSmith had already given it the original’s broker profile. The Data tab now shows the profile at once.

Manual →

Installer 1.1.3

The installer’s licence text now names release 1.1.6. Nothing else changed; installer 1.1.2 users do not need to run it again.

Installer update · 2026-11-20

Installer 1.1.2: MetaTrader 5 allow-list fix

This is an installer-only update. No product package changed, so the updater has nothing new to download. To get it, download the installer again from the download page and run it again; the installer can be run again on a machine that already has the products. Have your licence key to hand, as it asks for it first.

  • Fixed: the MetaTrader 5 allow-list was not set on current MetaTrader 5 builds. The installer sets Allow WebRequest for listed URL for Alternative Data (api.stlouisfed.org, publicreporting.cftc.gov, api.eia.gov) by driving the terminal’s own Options dialog. On current builds it could stop early because a closed pop-up still counted as an open dialog, could freeze with the Options dialog left open, and edited the first row of the list whatever row it meant. It now ignores hidden windows, gives up on a stuck window instead of waiting, and moves through the list with the keyboard. Entries you already had are kept and a second run adds nothing.
  • New: a progress window after Install now. When the daily updater asks and you click Install now, a window lists each package as it installs and ends with what to restart, instead of a silent install and a notification you may miss.

If the allow-list still cannot be set on your terminal, the log at %LOCALAPPDATA%\AWFeedKeeper\keeper.log says why, and the manual route is Tools > Options > Expert Advisors.

Release 1.1.5 · 2026-11-19

Edge Finder Studies and Seasonality, SpecSmith spread and equal swap, locked data fixes, block names ending in AW

ProductVersionIn this release
Edge Finder–Optional Studies in Market Bias and Style Fit; a Seasonality tab that Autopilot also runs; 5-minute, 2, 4 and 6-hour slots in Time Windows. Existing results are unchanged
NN Lab–Autopilot can compare network families and training methods (optional, slow); the MetaTrader 5 indicators update incrementally. Models and their values are unchanged
GapSmith–Synthetic fill can overwrite the symbol; the dialog explains how the synthetic bars are built; a symbol being fixed is locked and listed in the Grid engine overview
SpecSmith–Behaviour change: long and short swap are written equal by default (the average); typical spread from price history; new AWSpecFeed service; the timezone fix is locked per symbol
Institutional Indicators–Wyckoff PseudoCode export fixed; Cumulative Delta faster on long histories with an optional on-disk cache; block names in StrategyQuant X now end in AW. Blocks give the same values
Alternative Data–News export warns when no price-data timezone is configured; block names now end in AW. No other change
News–Same timezone warning on the News export; block names now end in AW. Existing News blocks give the same values

The numbers are each product’s own version. A product whose number did not change in a release is not downloaded again by the updater. NameSmith was rebuilt with the release date and has no change in what it does.

Block names now end in AW

StrategyQuant X blocks from Institutional Indicators, Alternative Data and News now carry an AW ending in their names (for example AltData became AltDataAW), so they cannot clash with a block of the same name from another source. The values they return are the same as before.

  • The installer checks your saved strategies and your MetaTrader 5 sources for the old names and offers to patch them. Originals are copied to a backup folder first, and it does not run while StrategyQuant X is open.
  • Restart StrategyQuant X once after the update: it recompiles the blocks and lists them under the new names.

GapSmith

  • Synthetic fill can now overwrite the selected symbol as well as create a _synth copy. The dialog describes how the bars are built: a random price path between the real prices either side of each gap, scaled to the volatility the symbol shows at that time of week.
  • A symbol being fixed is locked. Scanning or fixing it again while a fix runs is refused, queued symbols are held too, and reopening the dialog shows the running fix. While the grid runs, the fix is listed in the Grid engine overview and stopping it there ends the wait.

Manual →

SpecSmith

Behaviour change: long and short swap are equal by default

A strategy builder that sees different long and short swap can favour one side for that reason alone. The new Long and short swap option makes both sides equal. The default is the average of the two; you can also choose the worse side, or each side as reported. Plans and applies now write equal swap unless you choose each side as reported, so the swap in a project can differ from what you applied with 1.1.4.

  • Spread is a typical value measured from price history by the new MetaTrader 5 service AWSpecFeed, which also exports your broker’s symbols at start and daily. It is used by default, and the Spread column shows both the broker value and the one used.
  • The timezone fix is locked per symbol like the gap fix. Restart the MetaTrader 5 terminal for the feed service to start.

Manual →

Edge Finder

  • Studies (optional) in Market Bias and Style Fit, with four methods you tick.
  • A Seasonality tab: weekday, day of the month, week and month of the year. Autopilot runs it as a step.
  • Time Windows gains 5-minute, 2-hour, 4-hour and 6-hour slots.
  • Interface cleanup, including the duplicate Help button in Market Bias.

Manual →

NN Lab

  • Autopilot can also compare network families and training methods. It is optional and slow.
  • The neural network indicators in MetaTrader 5 update incrementally and reuse working memory, and recognise a licence bound to your terminal in the Strategy Tester.

Manual →

Institutional Indicators

  • The PseudoCode export of 45 Wyckoff blocks was missing commas between parameters; fixed.
  • Cumulative Delta is faster on long histories, and indicator results can be cached on disk (folder IndicatorCache under LocalAppData/Alphasworth).
  • MetaTrader 5 indicators recognise a licence bound to your terminal when run in the Strategy Tester.

Installer and updater

  • After you click Install now, the updater shows a progress window that lists each package as it installs and ends with what to restart. Customers who installed before 1.1.5 get it when they next run the installer, because the updater lives in the installer.
  • The installer sets the MetaTrader 5 Allow WebRequest for listed URL list for Alternative Data (api.stlouisfed.org, publicreporting.cftc.gov, api.eia.gov) by driving the terminal’s own Options dialog, and keeps your existing entries. It was made reliable on current MetaTrader 5 builds in this release. If it cannot, the manual route is Tools > Options > Expert Advisors.
Release 1.1.4 · 2026-11-16

Alt Data MetaTrader 5 stamps, NN Lab typed model ID

ProductVersionIn this release
NN Lab–Behaviour change (model chosen by a typed ID, no list in the blocks); no sync, compile or restart for a new model after the one restart that loads this update; Name for ID lookup
Alternative Data–Behaviour change: the MetaTrader 5 feed service now stamps rows with the same publication rules as StrategyQuant X; backtests on affected series may change

The numbers are each product’s own version. A product whose number did not change in a release is not downloaded again by the updater. Products not listed in this table were rebuilt with the new release date and have no change in what they do.

Alternative Data

Behaviour change: MetaTrader 5 feed stamps follow the StrategyQuant X publication rules

Every alternative-data row carries the date it became publicly knowable, and a strategy sees a row only from that date on. The StrategyQuant X side already used release-aware stamps. The MetaTrader 5 feed service (AWAltDataFeed) used older, simpler stamps for the rows it downloaded itself:

  • FRED series were all stamped the day after the observation date. That is right for daily closes (VIX, 10-year yield, yield curve, breakeven, ECB rate, implied-volatility indices) but too early for CPI, unemployment, the fed funds rate and the broad dollar index. Example: the July 2026 unemployment value was stamped 2026-07-02 on the MetaTrader side; it is stamped 2026-08-12 on the StrategyQuant side and now on the MetaTrader side. CPI for the same month: 2026-07-02 before, 2026-08-18 now.
  • CFTC and EIA rows were stamped without allowing for US federal holidays that move the release. Example: the CFTC report dated 2026-06-30 (Friday 3 July is a holiday) was stamped 2026-07-04 and is now stamped 2026-07-07; EIA week ending 2026-01-16 went from 2026-01-22 to 2026-01-23.
  • Effect: in a backtest or live run that used rows written by the MetaTrader service, those series could be read days to weeks before they were published (look-ahead). Daily-close series were not affected.

The service now applies the StrategyQuant rules: the 12th (unemployment) or 18th (CPI) of the following month, next business day plus two days (fed funds rate), the Wednesday after the Monday release (dollar index), and the holiday-adjusted CFTC and EIA release dates, with the same eleven US federal holidays.

  • Cached raw series are re-stamped when the service loads them; only the published column changes.
  • Feature rows that the service already appended are not rewritten. If the service reports that stamps were corrected, re-export the alternative-data files from StrategyQuant X (Data Manager > Alternative data) to rebuild the history from the corrected rules.
  • The service checks its stamp code against a table generated from the StrategyQuant code (about 1,500 cases plus the holiday calendar) when it starts, and does not start if a case differs. The same check runs at the top of the AWAltDataParityCheck script, and a test on the StrategyQuant side fails if either side changes without the other.
  • Backtests that use these columns may change: macro (CPI, unemployment, fed funds minus ECB, dollar index), and the CFTC and EIA columns in holiday weeks. Results can look worse than before; the earlier results used data that was not yet public. Feature files exported only from StrategyQuant X are not affected.
  • Not modelled: ad-hoc closures such as a government shutdown can delay a release past its stamp.
  • The Petroleum block help text said “Oil instruments only”; the EIA columns are carried on every catalogue instrument, and the text now says so.

Manual →

NN Lab

Behaviour change: the model is chosen by a typed ID, with no list in the blocks

In 1.1.3 the ModelId parameter of the NN blocks was a list of your models that SQX read from the block source when it started, so a new model needed Sync model list, and an SQX restart, before the Builder showed it. In 1.1.4 ModelId is a plain number again: the model’s ID (the Id column on the Models tab), typed as a Fixed value. The blocks hold no list of models, so training, retraining or deleting a model never changes them.

  • Upgrade note: restart SQX once after updating, so the changed block sources load. After that, a model you have just trained can be used at once: no sync, no compile and no restart.
  • The readable name is looked up from the ID. The Models tab has a new Name for ID box (type 21 to see, for example, MLP_XAUUSD_H4_ab12 (MLP, ID 21, uid ab12)), the result of Sync model list lists every ID with its name, and the help text of each NN block says how to read an ID. The Builder itself shows the number, not the name.
  • Sync model list is now optional housekeeping: it gives older models a uid, refreshes the list of IDs and names, and frees the IDs of deleted models. It no longer rewrites any block, and the Compile snippets now button was removed from the Models tab.
  • An ID with no model behind it, or one outside 1 to 999, still outputs 0 (neutral); the block never falls back to untrained weights.
  • The builder range of ModelId is 0 to 0 again, so an unfixed block generates 0 (the block default, not a trained model) instead of varying the choice across every listed model. Always set ModelId to a Fixed value.
  • A strategy still stores the same number as before, and MetaTrader 5 still takes it as the input InpModelId. Strategies and templates that already name a model keep working. A project whose Blocks configuration was saved with 1.1.3 holds a copy of the old list: reset its NN blocks once.
  • New: Deploy to Builder on the Models tab. Pick a project, a Build task and a model: it ticks the Neural Network indicator and its eight condition blocks on that task’s Building blocks page and sets ModelId (the ID) and Network type as fixed values. It changes only that page, never unticks a block, and reloads the Builder window showing the project. The step-by-step how-to blocks on the Models tab are gone, since the button does those steps.
  • NN Lab now also records each new model in the ID ledger (slots.tsv) when it is trained, so the check that a model file under an ID is the model you chose no longer waits for a sync. A deleted model’s ID is kept back until the next sync.

Manual →

Other products

Institutional Indicators, Edge Finder, News, NameSmith, GapSmith and SpecSmith: rebuilt with the release date, no change to what they do. The installer is unchanged.

Release 1.1.3 · 2026-11-16

Model picker, synthetic gap fill, per-product versions

ProductVersionIn this release
Institutional Indicators–Behaviour change (CumulativeDelta thresholds, Rising/Falling Three Methods); no change to results code for TWAP (help text only); precompiled libraries and a Wyckoff speed-up described below
NN Lab–Behaviour change: models are named automatically with a slot and a uid, picked from a list after Sync model list, and the first sync rewrites older models in a new file format that an older install cannot read (the originals are backed up); a stored slot number keeps reading its model
Edge Finder–Maintenance: shared data-clock code refreshed; no change to Edge Finder’s own code
Alternative Data–Behaviour change: a data series that has gone stale now reads as absent instead of repeating its last value; fixes listed below
News–Fix in the MetaTrader 5 calendar feed’s clock handling (events are matched by event id). We have not measured how this changes event times, so check your blackout times after updating
NameSmith–Unchanged
GapSmith–Optional marked synthetic fill, and an adjustable maximum gap length. Behaviour change: the default flat fill is the same bar as before, but it is no longer capped at 15 bars
SpecSmith–Timezone check added, and an optional overwrite mode for Fix timezones (off by default); the Timezone check only reports, and Fix timezones writes data only when you press Fix

The numbers are each product’s own version. A product whose number did not change in a release is not downloaded again by the updater. NameSmith has no change in this release and keeps its previous number.

Institutional Indicators

Precompiled libraries

Part of the algorithm code now ships as precompiled libraries in user\libs (AWCumulativeDeltaMath, AWPatternsMath, AWStructureMath, AWMovingAverageMath, AWExtendedFlowTwapMath, AWWyckoffMath). The snippets are now thin wrappers around them. The aim is the same logic as before; this release does not publish a bar-by-bar comparison for these libraries, so treat the move as a change of packaging that we intend not to alter outputs.

Behaviour change: CumulativeDelta condition blocks

Compared with 1.1.2, the fixed thresholds inside six blocks are lower, so the blocks can fire on ordinary data:

  • AbsorptionDetected: 0.25 to 0.05
  • InitiativeDetected: ±0.25 to ±0.10
  • AbsorptionThenInitiative (Down and Up): absorption 0.25 to 0.03, initiative ±0.25 to ±0.05
  • ControlRetestLong and ControlRetestShort: absorption 0.15 to 0.03

On our S&P 500 H1 test sample the old thresholds gave no signals for these blocks. Strategies that use these six blocks will trade differently. BullishDivergence and BearishDivergence are unchanged. The builder search ranges of these blocks (LookbackPeriod, Level) were also changed. The MT5 and other-platform templates carry the same thresholds.

Behaviour change: Candle pattern defaults and Rising and Falling Three Methods

Rising and Falling Three Methods (patterns 13 and 14) detect more bars: the inner candles’ real bodies must be contained within the first candle’s range, the close must be beyond the first close, and at least 2 of the 3 inner candles must be counter-coloured. Their default SmallBodyFraction in the blocks is now 0.40 (was 0.30), and the Harami Cross blocks now default to 0.10. For the candle blocks, the builder now holds TweezerToleranceATR and ATRPeriod at their defaults instead of searching them. Strategies that use these blocks can trade differently. The other patterns’ detection rules are unchanged.

No change to results code: TWAP condition blocks

No logic change. The help text now says that Level is a raw price on the chart’s scale.

Wyckoff condition blocks do less work per bar

The aim is unchanged outputs. For the four core Wyckoff indicators the repository’s MetaTrader 5 parity run compares every buffer of every closed bar over 10 parameter variants (tests\mt5\README.md); we have not re-published that comparison for this release.

  • FailedStructure and SignificantBarInversion no longer rescan the chart on every bar, and the condition blocks skip the point-and-figure count. We have not published speed figures for StrategyQuant X.
  • Known trade-off: a strategy that uses a Wyckoff condition block together with the AW WES indicator block with identical parameters now builds two engines, which can make it slower than before.
  • SpikeScope value 2 is now used internally by the condition blocks.

NN Lab

Behaviour change: models get a name, a uid and a slot

You no longer type a model number. NN Lab names each model <family>_<ticker>_<timeframe>_<uid> (for example MLP_EURUSD_H1_3fa9), stores a short uid in the model file, and gives it the next free slot (1 to 999). Training again never replaces an earlier model.

  • In the Builder, ModelId is a list of your models. After training or deleting a model, press Sync model list on the Models tab, compile the snippets, and restart SQX: Compile all alone does not refresh the Builder’s block list. After a Compile all, restart before you use NN Lab or GapSmith. Superseded in 1.1.4: ModelId is a typed ID again and a new model needs no sync, compile or restart.
  • Behaviour change: the default threshold of the Neural Network Lower and Cross Down blocks changed from 0.5 to -0.5. Existing strategies that rely on the default will behave differently.
  • Always set ModelId to a Fixed value. A saved Blocks configuration keeps an old copy of the list, so reset the NN blocks in an existing project.
  • A strategy stores the slot number, and MetaTrader 5 takes it as the input InpModelId. Models trained before this version keep their old number as their slot.
  • The first sync gives older models a uid and rewrites them in the current file format (v5 for the SQX file, v2 for the MetaTrader 5 file). An older install cannot read the rewritten files; the originals are copied to a backup-pre-v5 folder first. A legacy model that cannot be read is left out of the list, unchanged, with a warning.

Manual →

Help text: what a block does with no model

The NeuralNetwork block help now states that when there is no readable trained model for the slot and architecture in user\extend\Models\AWNN, or during warm-up, the outputs stay 0 (neutral). The help text also says that zero is a value: a rule such as Signal lower than 0.4 is still true on an empty slot, while higher-than rules are false. Train and save a model in NN Lab first.

GapSmith

The Fix gaps dialog now has a fill method: Keep gaps flat (the default, exactly as in GapSmith 1.0) or Fill missing bars with synthetic data (saved as a separate _synth symbol). The flat fill puts the same bars as before; what changes for both fills is the length limit.

  • Behaviour change: no 15-bar cap. A new Maximum gap length (bars) field is blank by default, which means no limit (a safety ceiling of 100,000 bars always applies). Any gap that is not a weekend, a holiday or a scheduled closure is now filled, flat or synthetic. Overwriting the original stays limited: blank means 15 there, and it is never unlimited. A new Minimum gap length (2 by default) applies to the synthetic fill. Edge Finder and NN Lab keep their 15-bar flat fill.
  • The synthetic fill is a seeded Brownian bridge from the previous close to the open of the next real bar, volume 0. It is one random draw, not what happened, and it uses the next bar, so it is for repairing history only.
  • It is not for quiet minutes, only for genuine outages. A missing minute in tick-derived history usually means nothing traded, and flat is the faithful record of that.
  • The result is a separate symbol with the _synth suffix (the synthetic fill never overwrites), a sidecar log lists every synthetic bar, and the dialog shows the warning “Synthetic data is not market data; results on filled bars are not evidence of anything.” and asks you to tick that you understand.
  • Accuracy is claimed only as far as the test harness measured it: hidden runs of 2 to 240 bars of ordinary bars on three currency pairs and one index. In those tests one draw had 0.94 to 1.01 times the volatility of the real bars (a straight line: 0.03 to 0.39, flat: 0), while its close had about 1.35 to 1.5 times a straight line’s error. The hidden bars were not real outages. Nothing longer than 240 bars was measured. Figures and limits.

Manual →

SpecSmith

New in the options dialog: a check of the timezone you typed against the weekend reopening in a currency pair’s M1 history, assuming a 17:00 New York reopening. Verdicts: CONFIRMED, MISMATCH, DST-SHIFTED and CANNOT VERIFY. MetaTrader’s live offset is shown only as a cross-check for the day of the export. The check never changes your setting. The separate Fix timezones button, which writes a corrected _tzfix copy and can optionally overwrite the symbol after an explicit confirmation, is documented in the manual. Manual →

Alternative Data

  • Behaviour change: in the MetaTrader 5 feed service, a source series that has gone stale now reads as absent instead of repeating its last value.
  • The InstrumentOverride setting has a new NAS100 option, which carries the implied-volatility series. In the MetaTrader 5 feed, a series whose newest value is older than 14 days (daily), 21 days (weekly) or 100 days (monthly) reads as missing.
  • Implied-volatility zero handling fixed.
  • InstrumentOverride help corrected: override values 1 and 2 apply only on charts without identity.
  • The block-support constructor is now public so SQX can load the AltData blocks.

Edge Finder, News, NameSmith

Edge Finder: shared data-clock code refreshed, no change to Edge Finder’s own code. News: the MetaTrader 5 calendar feed now matches anchor releases by event id before the English name, so a terminal that is not set to English can measure its calendar clock; we have not measured how this changes event times. NameSmith: unchanged.

Quality checks

A signal-coverage harness and parity tests were added. Every condition block was checked to be able to produce signals on sample data; the SQX Indicator Tester run covers 268 rows; the Java-versus-MT5 bar-by-bar checks are stricter.

Known limits:

  • The first 100 or so bars of a series differ because of warm-up.
  • NeuralNetwork parity is verified on fixture models, not on customer-trained models.
  • Results on one sample (S&P 500 H1, NAS100) say nothing about trading results.
Release 1.1.2 · 2026-11-12

SpecSmith COT and cost rounding, GapSmith overwrite option

SpecSmith

  • Added: a COT field. Compare and apply can set the CFTC contract code of an instrument (a forex pair uses the non-dollar currency’s future, so EURUSD gives 6E). A symbol with no clear futures market is left empty, and an existing COT value is never cleared. COT is a tick box under what to write.
  • Added: Round swap and spread up to the worse side, on by default. MetaTrader reports one day’s swap and spread; each is rounded up to the next step of 1, 1.25, 1.5, 2, 2.5, 3, 4, 5, 6, 8, 10 (times a power of ten), so a backtest is not kinder than the broker is likely to be. Commission is never rounded. Behaviour change for new applies: written swap and spread values can be higher than before unless you untick it.

GapSmith

  • Added: the Fix gaps dialog offers Create a copy (the default, a <symbol>_patched symbol) or Overwrite the selected symbol. Overwrite stays disabled until you tick that you understand the original history is replaced and cannot be restored from SQX.

Documentation

  • The manual pages for Edge Finder, NN Lab, Alternative Data, News, SpecSmith, GapSmith and the workflow were revised.
Release 1.1.1 · 2026-11-11

Edge Finder costs, sessions and clock; Deploy fix; Alt Data family blocks

Edge Finder

Fixed

  • Deploy emptied the task’s building blocks. On SQX 1.45 the Deploy step serialised the Build task with a compact XML writer that cut a large task to about 200 characters, so the Builder page then saved it back with every block unticked. Deploy now writes the full task. Checked in a running SQX 1.45: the Build task kept its blocks after Apply.

Changed

  • Behaviour change: Deploy and the .sqb and .cfx exports now use the style’s own order types, entries and exits, the same result as loading the style’s .sqb by hand. Before, the target task’s own were kept. The result line lists the entries and exits the build will use.
  • Costs from the Data tab (spread, slippage, commission, swap) are now carried beside the measurements: Market Bias shows cost and net columns, Time Windows a Net per trade column and Block Edge a cost line. They apply only if the Data tab holds those values. The bias itself stays a statement about the market, with no cost charged into it.
  • The trading session named on the Data tab is respected: a signal whose entry bar opens outside the session is counted but not treated as a trade, and a position held across the session’s end is closed at the last bar inside it.
  • Time Windows now measures the clock of a dataset from its weekend reopenings (17:00 New York) and reports whether the declared timezone agrees, instead of relying on the label alone. Rules recognised: fixed offset, European, US, Australian and New Zealand daylight saving.
  • Best timeframe at the Thorough depth now also runs Time Windows on each intraday timeframe, on the side that timeframe’s own Market Bias leans. D1 and above report that they have no time of day.
  • Each analysis result is shown in its own results view rather than inside the settings tab.

NN Lab

  • Added: a Results view for the last training run: model, sample counts (train, validation, purged at the boundary), class balance, losses and a confusion matrix. Earlier these figures were computed but not shown.

Alternative Data

  • Added: five family blocks next to the all-series AltData block: AltDataVolatility (VIX, own implied volatility), AltDataRatesDollar (dollar, 10-year, curve, breakeven, Fed-ECB gap), AltDataPositioning (CFTC leveraged funds), AltDataPetroleum (EIA crude stocks, refinery use) and AltDataMacro (CPI, unemployment). They use the same feature numbers as AltData and its MetaTrader template.

News

  • Added: the MetaTrader 5 feed service measures the chart-bar clock from the bars themselves (the same weekend-reopening method) and writes it as bar_tz, marked as measured. A bar_tz you set in StrategyQuant is never overwritten; if the measurement disagrees, the service says so.
  • Australian and New Zealand daylight-saving rules were added to the News blocks’ timezone handling.
  • The documentation now states that the MetaTrader calendar dump is written at one fixed offset (the terminal’s offset on the day of the export), so an old dump can be an hour out for part of the year.

SpecSmith

  • Added: a Fix timezones button. Check reads a pair’s M1 history and finds weeks whose weekend reopening sits one hour off the label; fixing writes a corrected copy of the symbol.

GapSmith, NameSmith

  • No change to either product’s own code; the packages were rebuilt.
Release 1.1 · 2026-11-08

SpecSmith, Time Windows, GapSmith tab, product rename

SpecSmith

  • New product. A Data Manager tab that reads what your MetaTrader 5 terminal reports for each symbol (tick size, point value, minimum distance, order size step, swap, commission, trading session, optionally spread), compares it with an SQX broker profile row by row, applies the rows you tick, and can undo an apply. It can also create a new broker profile from the terminal’s symbols.
  • It ships a read-only MetaTrader 5 script, AWSpecExport, which writes two CSV files to the common data folder and changes nothing in the terminal. Exact commissions need MetaTrader 5 build 6060 or later.

Edge Finder

  • Added: a Time Windows tab: which stretches of the day and week carry a directional tendency, tested with a correction sized for the whole search (a sign-flip permutation over every window and weekday set tried), with one observation per trading date. Hours are reported on the data clock first.
  • Added: a clock audit, which converts each weekend reopening to New York time under the declared timezone and reports weeks that sit one hour off, or says it was skipped when no stable anchor exists.
  • Added: Deploy can write the recommended window to the target Build task’s time range, and shows whether it agrees with Market Bias.

GapSmith

  • Added: a GapSmith tab in the Data Manager that lists every stored symbol in one grid, so gaps can be fixed in many symbols at once. The Fix gaps button inside Analyze data quality stays.

Alternative Data, News

  • The Alternative Data and News tabs were rebuilt in the Data Manager’s own structure. Alternative Data got ribbon buttons for update, rebuild offline, export to MetaTrader, stop, keys and settings, features, mapping a broker symbol and the log; settings, features, mapping and the log open as dialogs.
  • The News tab shows a coloured verdict on its toolbar for whether the calendar has run out of scheduled releases.

NN Lab

  • Fixed: Autopilot’s grid search could fail when a current value sat near the edge of a setting’s legal range (the candidate fell below the minimum). Candidates are now clamped to the legal range and de-duplicated.
  • The training result chart draws long and short signals as arrows instead of 2-pixel dots.

Institutional Indicators

  • Renamed from Wyckoff Indicators to Institutional Indicators. The product’s contents (blocks, strategy templates, MetaTrader 5 indicators) are unchanged by the rename; its package and licence feature id is now institutional.
  • Blocks now carry NinjaTrader 8 code-export templates, so SQX 1.45 no longer warns about missing ones. No NinjaScript indicator ships for these blocks, so exported code that uses them does not compile in NinjaTrader.

Documentation

  • The online manual was added: a page per product plus getting started, workflow and FAQ.
Release 1.0 · 2026-10-27

First release, and the updates published under the 1.0 label

The first release went live on 2026-10-27. Between 2026-11-02 and 2026-11-06 the packages were published again under the same 1.0 label, so each dated entry below names the build it first appeared in. Per-product version numbers did not exist yet.

Institutional Indicators (then named Wyckoff Indicators)

  • Condition blocks, strategy templates and example strategies for StrategyQuant X, and MetaTrader 5 indicators.
  • 2026-11-02: the MetaTrader 5 Wyckoff indicators skip a flat one-tick bar (open, high, low and close equal) the way the SQX blocks do, so MetaTrader 5 and SQX agree after holiday bars. In our MetaTrader 5 tester run against the Java blocks on S&P 500 H1 and M15, the buffers compared showed no differences.
  • 2026-11-03: in the SQX blocks, every closed bar, including a flat one, is stepped exactly once; price ties that a decimal quote can leave one unit apart are compared with a small relative tolerance in the blocks, the MetaTrader 5 indicators and the NN label code.

NN Lab

  • A module for training neural networks inside StrategyQuant X (Autopilot, Features, Labels, Models, Signals, Topology and Training tabs), with neural-network blocks and MetaTrader 5 indicators for the MLP, LSTM, CONV and CONVLSTM families.
  • 2026-11-03: training measures how much of each block input is absent over the bars it trains on; above 50% it refuses and names the block, above 5% the result carries a warning. An unreadable model manifest is kept as models.json.bad instead of being overwritten by the next save.
  • 2026-11-05: iRprop- is the default optimizer (Adam and SGD stay selectable). Autopilot now considers every feature and every indicator block, runs ranking and sweeps as grid jobs, and recovers from unusable blocks instead of stalling. Inputs SQX already provides (RSI, Stochastic, Momentum, and the two COT percentile columns where SQX ships COTCpispec) are hidden and no longer ticked.

Edge Finder

  • A module with Market Bias, Style Fit, Best timeframe, Block Edge, Autopilot, Deploy and Runs tabs.
  • 2026-11-03: the Benjamini-Hochberg false-alarm grade ignores undefined p-values when counting the family; before, two undefined values in a four-test family raised every q-value. Edge Finder and NN Lab share one job queue.
  • 2026-11-05: block, indicator, timeframe and style scans run as grid jobs instead of one after another on the request thread.
  • 2026-11-06: Apply refreshes the Builder’s cached copy from the saved task; Block Edge scans the whole block registry including price-level indicators; the Building blocks tab was removed; a task with no order type ticked falls back to the style’s entries; Deploy has no task selector (Apply goes to the project’s Build task) and the Edge Finder and NN Lab projects are hidden from the target list.

Alternative Data

  • A Data Manager tab, blocks, MetaTrader 5 indicators, a feed service and a keys script.
  • 2026-11-03: three publication-time leaks fixed: COT on a holiday Friday is released on Monday, EIA after a Monday holiday on Thursday, and the BLS jobs report can be scheduled as late as the 10th; bars before the release no longer see the print. Monthly derived features (CPI year over year, unemployment change) look up the comparison month by date, not by row position. A last value older than 14 days (daily), 21 days (weekly) or 100 days (monthly) reads as absent instead of being carried forward.

News

  • A Data Manager tab, timing blocks, strategy templates and example strategies, a MetaTrader 5 indicator, feed service and export script.
  • 2026-11-01: the calendar clock is measured by MetaQuotes event id first and English name second, so German, Russian or Chinese terminals can measure it without a typed timezone.

NameSmith

  • A databank action that renames strategies from a template, with live preview and undo.

GapSmith

  • A Fix gaps action inside Analyze data quality: short holes in M1 history are filled with flat, zero-volume bars at the previous close and imported as a new <symbol>_patched symbol.

Installer, updater and licensing

  • A signed installer, an updater with rollback, and licence keys issued after checkout.
  • 2026-11-01: licence server fixes: a refund or chargeback that arrives before the paid-order event now leaves the key disabled, and a repeated delivery of the same order returns the same key instead of a second one.
  • 2026-11-03: the installer defaults to the standard x64 build and warns that an optimized build must be supported by the machine before you choose it; wizard labels no longer crop.
Updates

How update notes appear, and how to go back

In the updater. The updater does not carry release notes. Its popup names the products that are ready to install and says which of StrategyQuant X and MetaTrader 5 to close first, with Install now and Later. When it finishes, a notification and the file %LOCALAPPDATA%\AWUpdater\notice.txt say which products were updated and that SQX needs a restart. Read this page for what changed; look for Behaviour change before you update a machine that trades.

Rolling back. The updater keeps one previous version per product. With SQX and MetaTrader 5 closed, run AWUpdater.ps1 -Rollback <product> in PowerShell from the updater folder inside your Alphasworth install folder, where the product is the package name the updater uses: institutional, nnlab, edgefinder, altdata, news, autorename (NameSmith), gapsmith or specsmith. AWUpdater.ps1 -Status lists the installed packages and which have a rollback copy. Updating, in the manual →

Risk

Trading leveraged instruments carries a high risk of loss. Software output is not investment advice and says nothing about future results. Validate every strategy yourself before risking capital.