Founding beta · invite only

One verdict, from a rulebook that doesn't bend.

Delta scores every setup on sweep, reclaim, SMT and session, runs structure, macro and risk gates, then publishes one verdict with its evidence. You place the trade.

The form asks for your email address and nothing else. If a tester seat opens for you, the invitation comes by email.

From the dashboard · tester view · replay of 2026-07-01

  • TerminalThe verdict, the four checks behind it, the structure, macro and risk gates, and — only when it raises a GO — the plan if it fills.
  • Live TapeA running record of recent events — sweeps and whether they were reclaimed, SMT reads, structure shifts, macro news, vetoes and verdict changes. Each is time-stamped, filterable by kind and searchable.
  • JournalWhat Delta raised and any decision recorded against it — one shared record, not your own trades. Delta does not verify fills.
  • DisciplineA check-in before the session, the rules on screen during it, a debrief after, and each day graded on process. It never touches a verdict. Operator only — testers do not get this screen.
The Delta Terminal on a phone in light mode, tester view, replay of 2026-07-01: a GO SHORT verdict on MNQ, labelled research only and not a live signal, and the start of the plan if it fills. On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta.The Delta Terminal on a phone in dark mode, tester view, replay of 2026-07-01: a GO SHORT verdict on MNQ, labelled research only and not a live signal, and the start of the plan if it fills. On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta.The Delta Terminal in light mode, tester view, replay of 2026-07-01: a GO SHORT verdict on MNQ, the plan if it fills with the target named by its level, the three gates, the evidence, and each level by name with no chart. On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta. The status line reads not a live signal.The Delta Terminal in dark mode, tester view, replay of 2026-07-01: a GO SHORT verdict on MNQ, the plan if it fills with the target named by its level, the three gates, the evidence, and each level by name with no chart. On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta. The status line reads not a live signal.

No live track record yet. Forward test open — 1 of 100 trades resolved for the pre-registered MES 1-minute configuration.

On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta.

How a verdict is made

Four checks, three gates, one verdict.

Every closed bar goes through the same rulebook, and every verdict comes with the reason behind it.

MNQ · 15m · replay 2026-07-01 · 10:45 ETEvidence
  • METLiquidity sweepMKO · Market Open level · 10:45 ET
  • METSweep reclaimedprice traded back through the level
  • METSMT supportiveSMT bearish · candidate short
  • METTradeable sessionNY AM
  • PASSStructure gateshort setup: MKO swept and reclaimed
  • PASSMacro gateno caution window, impact score below 60
  • PASSRisk gatestop 56.75 pts, inside the 150-pt cap · before 15:30 ET

GO SHORTall four checks met · all three gates pass

The Terminal's evidence for one replayed bar (MNQ, 2026-07-01). A replay, not a live signal.On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta.

The four-point table

  1. 01Liquidity sweep

    A bar trades through a named level after the previous bar closed on the near side: PDH/PDL, PWH/PWL, PMH/PML, P2WH/P2WL, the Asia and London highs and lows, the Midnight Open, the 08:30 NY Open or the 09:30 Market Open. A run above a level points the setup short; a run below points it long.

  2. 02Sweep reclaimed

    A bar closes back through the level, either the sweep bar itself or one of the next three. No reclaim, no setup.

  3. 03SMT supportive

    At the sweep, Delta checks the partner contract (MES for MNQ; MNQ for MES and ES) at its own copy of the same level, the same day. If the partner did not take its level, that divergence confirms the direction. Without that confirmation there is no GO, so MGC, which has no partner, cannot reach GO.

  4. 04Tradeable session

    The bar has to open inside London (02:00–05:00 ET), NY AM (09:30–12:00 ET) or NY PM (13:00–15:30 ET). Any other time is not a tradeable window: Asia, the lunch hour, the gaps between sessions and weekends.

The three gates

  1. 05Structure gate

    It passes when the most recent sweep happened within the last 12 bars and was reclaimed, and there is an entry to use: the midpoint of the latest open fair value gap in the trade's direction, or else the swing broken by the last structure shift. SMT points against the setup when the partner took one of its own levels that day and this contract did not take its copy. That is a veto: the verdict is STAND DOWN however good the rest looks.

  2. 06Macro gate

    Delta reads one file written by a separate macro monitor: an alert label, an impact score from 0 to 100 (one for index futures and one for gold when the file gives both) and any open caution window. A score above 80, or an open caution window that covers your market, makes a setup STAND DOWN. A score from 60 to 80 is caution, and caution means NO TRADE. If the file is missing, unreadable or more than a day old, its scores are not used: the gate passes and says why. A caution window it explicitly set still vetoes until its stated end.

  3. 07Risk gate

    The stop goes beyond the swept extreme, at least a quarter of the stop's distance past the swept level. It can be no tighter than a noise floor set from recent bar ranges and no wider than a fixed cap in points. The targets are the nearest named levels in the trade's direction, and the second must sit at least twice the stop distance from the entry. A bar that opens at 15:30 ET or later cannot pass.

The verdict

  1. 08One verdict

    GO LONG or GO SHORT comes only when all four checks are met and all three gates pass, with the entry, the stop and the target levels attached. Otherwise: STAND DOWN when SMT or the macro gate vetoes a setup; NO TRADE when a setup fails the risk gate, misses a check or meets macro caution; SCANNING while no setup has formed. When the market is shut, the dashboard says MARKET CLOSED. Each verdict carries the engine's own reason.

Why most sessions end with no trade

Every rule has to hold on the same bar. A sweep with no reclaim, a reclaim with no SMT, a setup that forms at 15:45, a stop wider than the cap: each ends without a trade, and Delta shows you which rule stopped it. Most sessions end that way. Standing aside is still a verdict.

Delta does not place orders. You decide on your own chart whether to take a GO and whether its entry is still reachable.

Top-down

Top-down, read the same way every time.

You check Daily, 4H and 1H before you take a setup. Delta reads them by fixed rules as bars close and shows the read beside the verdict. It's there as context and doesn't get a vote.

The Top-down panel: the frames drawn as a stack of plates, Daily on top, and one row per frame with its trend, last break of structure and zone.The Top-down panel: the frames drawn as a stack of plates, Daily on top, and one row per frame with its trend, last break of structure and zone.
The Top-down panel as an invited tester sees it: the engine's read of real MNQ bars at the close of the 10:45 ET bar on 2026-07-01, with the 4H built from the real 1H bars.
Frames
Daily, 4H and 1H, whichever sit above the analysis frame, then the analysis frame itself, then 5m below it when the analysis frame is longer than 5m. The operator picks the analysis frame: 1m, 5m, 15m or 1h. On the default 15m, that's all five.
On each frame
The trend (bullish, bearish or ranging) from the last two swing highs and lows. The last break of structure (a BOS, or an MSS when it breaks against the trend) and when it happened. Whether the last close sits in premium, discount or equilibrium of the range between the latest swing high and swing low.
Higher-frame bias
When more than half of the higher frames with a readable trend are bullish, or more than half are bearish, that is the bias, shown with how many agree. Otherwise it reads mixed. When there's a setup, the read names the frames it runs with and the ones it runs against.
As bars close
Only closed bars count. The read updates when the analysis frame closes a bar, and each other frame refreshes after its own bar closes: 5m every five minutes, 1H hourly, 4H every four hours, Daily once a session. A frame that falls behind is marked stale.
Context only
It never changes the verdict, the gates or the tier. Whether trading with the higher frames helps was put to a pre-registered backtest. It didn't clear the bar set in advance, so the read stays context.

Invited testers see the same trends, breaks, zones and times. The price levels behind them stay with the operator's licence.

Funded-account rules

Built around funded-account rules

Delta does not place orders. It doesn't read your balance or your firm's rules. Four of these are written into its rulebook and run as each bar closes. The fifth is yours: a log that grades how you traded.

  • 15:30 ET

    A 15:30 ET cutoff

    No bar that opens at 15:30 ET or later can produce a GO for the rest of the trading day, and the time is fixed in the code, not a setting.

    risk gate

  • 150 pts

    A stop cap per instrument

    Every instrument has a stop cap on file, and a setup whose stop runs wider gets no GO; on MNQ, the contract the cap was set for, that's 150 points, and the other three contracts use that same figure.

    risk gate

  • MIS · windows

    Macro caution windows

    Delta reads a file written by a separate macro monitor the operator runs: an open caution window on your market, or an impact score above 80 for your market, stands a setup down, and from 60 up nothing gets a GO. A file more than a day old stops counting, except a caution window it explicitly set.

    macro gate

  • 1 setup

    One verdict per setup

    Delta knows a setup by its direction and entry, so a GO that holds across several bars is recorded once, even while its stop is refined; a new direction or entry is a new setup.

    verdict engine

  • Your process

    Discipline grades process, not P/L

    You check in before the session and debrief after it, and the grade comes from your own answers on things like size, stops, revenge trades and stand-downs; your P/L counts only if the day's loss runs more than 15% past the daily loss limit you set.

    Discipline

During the beta, Discipline opens on the operator account only.

The desk

Everything on the desk.

Seven screens, and two more where the operator sets alerts and controls. Where a tester sees less than the operator, the line says so.

Terminal. The verdict with its evidence: four checks, three gates, and on a GO, the plan if it fills.

Operator: the chart. Tester: each level by name — swept or not, reclaimed or not, price above or below.

The Terminal screen in light mode, tester view, replay of 2026-07-01: a GO SHORT verdict on MNQ, labelled not a live signal, with the plan if it fills, the three gates and the evidence. On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta.The Terminal screen in dark mode, tester view, replay of 2026-07-01: a GO SHORT verdict on MNQ, labelled not a live signal, with the plan if it fills, the three gates and the evidence. On MNQ the terminal marks the verdict research only; MES is the only instrument enabled for the beta.

Live Tape. Sweeps, SMT, structure shifts, vetoes, cutoffs and GOs as the engine logged them. Filter by kind, or search.

Holds the latest few dozen events; the engine keeps a longer record. Testers see level names where the operator sees prices.

The Live Tape screen in light mode, tester view, replay of 2026-07-01, with the Structure filter selected: time-stamped SMT reads, sweeps with whether they were reclaimed, the structure shift and the level maps, and the line saying 8 of 20 events are hidden by this filter.The Live Tape screen in dark mode, tester view, replay of 2026-07-01, with the Structure filter selected: time-stamped SMT reads, sweeps with whether they were reclaimed, the structure shift and the level maps, and the line saying 8 of 20 events are hidden by this filter.

Feed Health. Live or replay, connection, bar freshness, SMT partner feed — each reported separately, with what you can rely on.

The Feed Health screen in light mode, tester view, replay of 2026-07-01: the data source reads REPLAY, and the bar stream, partner series, order-flow hub and clock are each reported separately.The Feed Health screen in dark mode, tester view, replay of 2026-07-01: the data source reads REPLAY, and the bar stream, partner series, order-flow hub and clock are each reported separately.

Journal. What Delta raised and whether the operator took or passed each, kept apart from the operator's workbook.

One shared record for the desk, not a tester's own trades. The workbook opens for the operator only. Delta does not verify fills.

The Journal screen in light mode, tester view, during the replay of 2026-07-01: the tester's default tab, This app's record, is empty with nothing recorded yet, and the workbook tab is marked operator only.The Journal screen in dark mode, tester view, during the replay of 2026-07-01: the tester's default tab, This app's record, is empty with nothing recorded yet, and the workbook tab is marked operator only.

Discipline. Check in before the session, keep your rules on screen, debrief after. Days are graded on process.

Operator only. Kept in this browser, and it never touches a verdict.

The Discipline screen in light mode, operator view (operator only; testers do not get this screen): the pre-session check-in, question one of five. Its date, 2026-09-24, is a sample day, not the replay of 2026-07-01.The Discipline screen in dark mode, operator view (operator only; testers do not get this screen): the pre-session check-in, question one of five. Its date, 2026-09-24, is a sample day, not the replay of 2026-07-01.

Research. Seven studies, including a negative one: what each found, how strong the evidence is, and what limits it.

Research is not a track record. Per-trade rows open for the operator only.

The Research screen in light mode, tester view: the note that research is not a track record, the line that these summaries are the whole research surface for a tester account, and the key to reading each study.The Research screen in dark mode, tester view: the note that research is not a track record, the line that these summaries are the whole research surface for a tester account, and the key to reading each study.

Macro. The macro file's impact score, caution windows and gate — flagged when the file is over a day old.

Operator and testers.

The Macro screen in light mode, tester view, during the replay of 2026-07-01: a banner saying the macro file is 18 days old so its scores are not used, the macro impact gauge marked not applied, the file's alert headline, and the gate on the last verdict (PASS).The Macro screen in dark mode, tester view, during the replay of 2026-07-01: a banner saying the macro file is 18 days old so its scores are not used, the macro impact gauge marked not applied, the file's alert headline, and the gate on the last verdict (PASS).
  • Terminal

    The verdict with its evidence: four checks, three gates, and on a GO, the plan if it fills.

    Operator: the chart. Tester: each level by name — swept or not, reclaimed or not, price above or below.

  • Live Tape

    Sweeps, SMT, structure shifts, vetoes, cutoffs and GOs as the engine logged them. Filter by kind, or search.

    Holds the latest few dozen events; the engine keeps a longer record. Testers see level names where the operator sees prices.

  • Feed Health

    Live or replay, connection, bar freshness, SMT partner feed — each reported separately, with what you can rely on.

  • Journal

    What Delta raised and whether the operator took or passed each, kept apart from the operator's workbook.

    One shared record for the desk, not a tester's own trades. The workbook opens for the operator only. Delta does not verify fills.

  • Discipline

    Check in before the session, keep your rules on screen, debrief after. Days are graded on process.

    Operator only. Kept in this browser, and it never touches a verdict.

  • Research

    Seven studies, including a negative one: what each found, how strong the evidence is, and what limits it.

    Research is not a track record. Per-trade rows open for the operator only.

  • Macro

    The macro file's impact score, caution windows and gate — flagged when the file is over a day old.

    Operator and testers.

Alerts · Settings. Operator only: GO alerts by browser or Discord-compatible webhook; instrument, timeframe, pausing new GOs, locking this browser.

Data path

Where the data comes from.

Closed bars, the SMT partner's bars and one macro file go into the rulebook. One verdict comes out, to you, and the line stops there.

  1. ProjectX / TopstepX bars

    MES by default and in the beta · MNQ, ES or MGC selectable, labelled research only

    Closed bars for the chosen contract, over TopstepX's REST API. The bar still forming never reaches the rulebook.

  2. Partner contract

    MES for MNQ · MNQ for MES and ES · none for MGC

    Read for SMT only. A GO needs the partner to confirm; divergence against the setup stands it down. MGC has no partner, so it never reaches GO.

  3. Macro file

    an alert label · a 0–100 impact score · any caution window

    Written by a separate macro monitor that runs apart from the engine, not continuously. A score over 80 or a caution window for your market stands setups down; 60 to 80 blocks a GO.

  4. The rulebook

    four checks · three gates

    Sweep, reclaim, SMT, session; then the structure, macro and risk gates. A GO needs all four checks and all three gates. The rules are fixed in code and do not learn from results.

  5. You

    one verdict, with its evidence

    You read it beside your own chart and place the trade yourself, or don't. Delta does not place orders.

There is no order route. The server has no endpoint that sends an order, and its TopstepX connections carry market data only.

When the feed drops

  • A poll that fails passes nothing on. Delta never invents a bar or fills a gap it did not receive.
  • The next poll asks again. Each request covers enough recent history to span a short outage, so the closed bars it missed arrive in order, each one once.
  • If the newest bar is too old for an open market, the response is held back and the rulebook never sees it. Feed Health judges bar age against the exchange clock and marks the feed lagging or stale.
  • When ProjectX rate-limits the account, the bar polls pause: 5 seconds, then 10, doubling to a 2-minute cap, or longer if the server asks. A clean poll resets the wait.
  • When its connection drops, the trade tape, context only, reconnects after 1, 2, 5 and 10 seconds, then every 30 seconds, signing in afresh each time.

Context, not input

Two reads sit beside the verdict and never feed it: the top-down read, whose bias comes from the higher frames (Daily, 4H and 1H on the default 15m), and the trade tape from TopstepX's market hub, with every print's size and the side that crossed the spread. Quotes are top of book only, with no depth. Volume bursts from the tape show on the Live Tape, marked context only.

The top-down read on the default 15m. Each frame refreshes after its own bar closes: in an hour of open market, 5m twelve times, 15m four times, 1H once, 4H and Daily at most once. The read never changes the verdict.

The night desk

The chart, with the lights down.

The bars the engine reads, drawn as red and green candles in colours of their own, kept apart from the green that marks a GO.

The Delta chart in dark mode, widened to the full row: red and green MNQ 15-minute candles with the session bands underneath.
The operator's chart, dark theme, widened; the price scale is cropped out · MNQ 15m, replay of 2026-07-01.
  • Pan and zoom

    Drag to pan, scroll to zoom at the cursor. With the chart focused, the left and right arrows pan one bar, plus and minus zoom, and Home returns to live and resets the view.

  • Drag it taller

    Pull the grip under the chart to set its height, or focus the grip and use the arrow keys. Double-click puts the default height back.

  • Widen

    When the desk is wide enough for three columns, Widen gives the chart the full row. Narrow gives Evidence and Recent changes their columns back.

  • Light or dark

    The moon/sun control switches the desk between dark and light and remembers your choice in this browser. Dark is the default.

Candles update live only on the engine's own timeframe. The operator can ask the engine for another timeframe's closed bars, one request per switch, and that view doesn't update. Where there's no current engine read to ask, the timeframe row is a readout marked display only.

The chart is the operator's view. Invited testers get no candles and no level prices. In its place they get a level ladder: each level, which side of it price is on, whether it was swept, from which side and when, and whether it was reclaimed.

A chart to read, with no order ticket on it. Delta does not place orders.

Research

Research is not a track record.

Seven studies are indexed on the dashboard's Research screen, and the negative and null findings are listed with the rest. Every number in the research index is copied from the cited report, and each study card on that screen shows its sample size and what limits it. This page lists each study's name, what it found and how it was designed. The figures stay on the dashboard, next to their samples.

01 EDGE Strategy-Input Backtest — 6 inputs, 1 year Pre-registered · only one of the six inputs held, and on a small holdout
02 MIXED Enhanced Order Block (EOB) — pre-registered study Pre-registration on file
03 NO EDGE EOB 50-point floor — wide-stop counterfactual Protocol frozen before the re-walk, but the rule was chosen after seeing the losses · adopted by operator decision, not validated
04 NO EDGE Infinity Loop — self-refining search, 15m/5m/1m Pre-registration on file
05 EDGE Delta Model — 30-day no-lookahead walk Rulebook written before the walk; in-sample, and a small sample — directional support, not proof
06 NEGATIVE Composed A+ System — 13-month grade study No pre-registration
07 DESCRIPTIVE 6-Month Multi-Instrument Baseline No pre-registration

Turn it with the buttons or the arrow keys; every study is listed in full below.

EDGE — the tested rule beat its own null · MIXED — part of the rule held, part did not · NO EDGE — measured, and it changes nothing either way · NEGATIVE — the tested rule is worse than doing nothing · DESCRIPTIVE — a measurement of the data, not a test of a rule

  • Strategy-Input Backtest — 6 inputs, 1 year

    EDGE

    Pre-registered · only one of the six inputs held, and on a small holdout

  • Enhanced Order Block (EOB) — pre-registered study

    MIXED

    Pre-registration on file

  • EOB 50-point floor — wide-stop counterfactual

    NO EDGE

    Protocol frozen before the re-walk, but the rule was chosen after seeing the losses · adopted by operator decision, not validated

  • Infinity Loop — self-refining search, 15m/5m/1m

    NO EDGE

    Pre-registration on file

  • Delta Model — 30-day no-lookahead walk

    EDGE

    Rulebook written before the walk; in-sample, and a small sample — directional support, not proof

  • Composed A+ System — 13-month grade study

    NEGATIVE

    No pre-registration

  • 6-Month Multi-Instrument Baseline

    DESCRIPTIVE

    No pre-registration

The dashboard marks its study figures as simulated research fills that assume the limit filled. None of those figures are repeated here.

No live track record yet. Forward test open — 1 of 100 trades resolved for the pre-registered MES 1-minute configuration.

Access

Request access

At launch, Delta runs for one operator and a small group of invited testers. Each tester signs in with a personal access code, not an account, and receives Delta's read: the verdict, the levels by name, the structure it found, and — on a GO — its own entry and stop. Exchange prices — candles, level prices, the target price — stay with the operator's licence, so keep your own chart open beside it.

The form asks for your email address and nothing else. If a tester seat opens for you, the invitation comes by email.

Have an access code? Enter it on the app's own sign-in screen, never on this site. How a code is used

FAQ

Questions.

Does Delta place trades?

No. Delta does not place orders. It has no order route: it reads market data and publishes a verdict, and you place every trade you take yourself.

Does it run by itself?

The read does. Delta reads every closed bar on its own and updates its verdict when the read changes. The decision and the order stay with you: Delta has no order route.

Which markets?

Four futures contracts: MES, MNQ, ES and MGC, on 1-minute, 5-minute, 15-minute or 1-hour bars. The operator picks the instrument and timeframe for the desk. MES is the default instrument and the only one enabled for the beta; on MNQ, ES or MGC the terminal marks the verdict research only.

What does a tester see?

Delta's read, not the market's data. That means the verdict and its gates, the levels by name, the structure it found, and the verdict's own entry and stop, with the target named by its level. Candles, level prices, the target price and the planned R:R stay with the operator's licence, so keep your own chart open beside it.

Is there a track record?

No live track record yet. Forward test open — 1 of 100 trades resolved for the pre-registered MES 1-minute configuration. The studies above are research, not a track record.

What does it cost during the beta?

Nothing is charged. Access is by invitation during the beta, and there is no plan, no price and no card. Request access with your email to be considered.