SPYQQQSMHNVDAAAPLTSLAMSFTAMZNGOOGLMETACOINMSTRCRCLIRENTQQQSOXLSPXLBTCETHSPYQQQSMHNVDAAAPLTSLAMSFTAMZNGOOGLMETACOINMSTRCRCLIRENTQQQSOXLSPXLBTCETH
Doc: VANTERA-ENGINE-BLUEPRINTRev. 2026-09-11

Every engine,
drawn to spec.

Eight quantitative engines, the rules each one runs, and the last trade each one actually took in backtest — drawn on a TradingView chart with entry, stop loss, take profit and timeframe.

Project
Deltalyn
Drawing
System blueprint
Engines
08 declared · 04 drawn
Data
To close 2026-09-11
Exported
2026-09-12 02:58 UTC
Status
Research · places no orders
  • Entry
  • Stop loss
  • Take profit
  • Activation · members
Sheet 01 / 05Doc: ENG-INPUTS-01
01 — Read the chart

Price, volume, indicators, structure.

Every engine starts from the same four readings and turns them into a single vote — long, or abstain with a written reason. The vote goes to the risk gate; nothing here ever reaches a broker.

Price
OHLC candles, 1D or 4H
Volume
Confirms or vetoes a break
Indicators
SMA · EMA · RSI · ATR · z-score
Structure
Channels · prior highs · regime
Engine · decide()
Pure function of closed bars
Vote
Long · or abstain, with the reason
Risk gate
Last door · fail-closed

Engine inventory

08 engines · source of each drawing below
  1. E-01 · ROTATION·LEV
    Leveraged rotation
    TF · 1D
    TQQQ · SOXL · SPXL
    Reads: Underlying vs SMA50/200 · realised-vol percentile · 20-day high · volume
    Backtest tradeLast trade of the engine's own backtest
  2. E-02 · DONCHIAN·4H
    Crypto Donchian channel
    TF · 4H
    BTC · ETH (long)
    Reads: Donchian 20 · EMA20 · SMA50/200 · RSI 14 · ATR 14 · volume
    Replayed backtestLast long of a bar-by-bar replay of its decide()
  3. E-03 · STAT-ARB
    Pairs mean reversion
    TF · 1D
    ETH/BTC · COIN/ETH · MSTR/ETH · CRCL/COIN
    Reads: Spread log(A) − β·log(B) · z-score · AR(1) half-life
    Backtest tradeLast trade of the engine's own backtest, in z units
  4. E-04 · PROBABILITY
    Setup evidence engine
    TF · 1D
    21 graded assets × 4 setups
    Reads: Donchian 20/55 · trend pullback · RSI-60 momentum · Wilson 95% interval
    Graded evidenceGrade table + last graded trade, replayed
  5. E-05 · TJL
    Trend Join Long
    TF · Intraday + 1D
    US equities
    Reads: Prior-day high · premarket high · SMA200 · weekly trend · relative volume
    No dataNo in-engine backtest; scans only 10:00–15:30 ET
  6. E-06 · HMM REGIME
    Hidden-state regime detection
    TF · 1D
    ETH · BTC
    Reads: Baum-Welch HMM on daily log returns · forward-only filtering
    State · membersNo entries by design · members see today's state
  7. E-07 · RISK GATE
    Portfolio risk gate
    TF · Per ticket
    Every ticket
    Reads: Distance to stop · correlation cluster · monthly drawdown · open heat
    No dataEvaluates private positions — not published
  8. E-08 · ETH-DCA
    ETH accumulation
    TF · 1D
    ETH
    Reads: Drawdown percentile · Fibonacci zones · exchange net-flow
    No dataOutput depends on private capital settings — not published
X: 010.00Deltalyn · System blueprint
Sheet 02 / 05Doc: ENG-CONTEXT-02
02 — Context before prediction

Explain the market before calling it.

Before an engine votes, it writes the state of the market down: trend, key levels, momentum, volatility — and what would prove it wrong. Members see today's memo for every engine and the exact level that would activate it.

Members only

Today's memos and activation levels are on the desk.

Each engine's context memo — trend, key levels, momentum, volatility, invalidation — and the exact price that would activate it, drawn on today's chart. Included with the Analyst and Full Board plans.

  • TQQQ
  • SOXL · via SOXX
  • SPXL
  • BTC 4H
  • ETH 4H
  • ETH/BTC z
  • HMM state
  • TJL scanner
X: 020.00Deltalyn · System blueprint
Sheet 03 / 05Doc: ENG-RULES-03
03 — Rules become code

If it can't be written down, it isn't a rule.

Every engine is a function over closed bars: the same inputs always return the same vote. These are the parameters each one runs with today, read from its own configuration — plus a Pine mirror you can load on TradingView.

ROTATION·LEV

Leveraged rotation

TF · 1D
  1. Regimeunderlying close > SMA50 > SMA200
  2. And20-day realised vol ≤ 60th percentile of the last year
  3. And3× daily compounding tracks the underlying within tolerance
  4. Triggerclose > prior 20-day high · volume ≥ 1.2× average
  5. Stopentry − 1.5 × ATR14
  6. Targetentry + 3 × ATR14 · min R:R 1.5
  7. Timeexit after 30 sessions
Pine Script v6 · Leveraged rotation
Indicator only · paste into TradingView's Pine editor
//@version=6
indicator("Deltalyn · Leveraged rotation — blueprint", shorttitle = "VNT ROT", overlay = true)

// Daily chart of TQQQ, SOXL or SPXL. Set the matching unleveraged index.
underlying = input.symbol("NASDAQ:QQQ", "Underlying (QQQ / SOXX / SPY)")
smaFastLen = input.int(50, "Trend SMA fast")
smaSlowLen = input.int(200, "Trend SMA slow")
volLen     = input.int(20, "Realised-vol lookback")
volPctLen  = input.int(252, "Vol percentile window")
volPctMax  = input.float(60, "Max vol percentile")
brkLen     = input.int(20, "Breakout lookback")
volMult    = input.float(1.2, "Volume confirmation x")
atrLen     = input.int(14, "ATR length")
stopMult   = input.float(1.5, "Stop = entry - ATR x")
targetMult = input.float(3.0, "Target = entry + ATR x")
maxHold    = input.int(30, "Max holding (sessions)")

// Regime is read on the UNDERLYING: close > SMA50 > SMA200, realised vol in
// its lower percentiles. (The engine's tracking-drift check is not replicated.)
[uClose, uFast, uSlow, uVolPct] = request.security(underlying, "D", [close, ta.sma(close, smaFastLen), ta.sma(close, smaSlowLen), ta.percentrank(ta.stdev(close / close[1] - 1, volLen), volPctLen)], lookahead = barmerge.lookahead_off)
regimeOk = uClose > uFast and uFast > uSlow and uVolPct <= volPctMax

// Trigger on the leveraged ETF: close above the PRIOR 20-day high on volume.
priorHigh = ta.highest(high, brkLen)[1]
volumeOk  = volume >= ta.sma(volume, brkLen)[1] * volMult
atrVal    = ta.sma(ta.tr(true), atrLen)
signal    = regimeOk and close > priorHigh and volumeOk

// One position at a time; exit on stop, target or time, like the backtest.
var float entry = na
var float stop = na
var float target = na
var int entryBar = na
exited = false
if not na(entry)
    if low <= stop or high >= target or bar_index - entryBar >= maxHold
        entry := na
        stop := na
        target := na
        entryBar := na
        exited := true
if na(entry) and not exited and signal
    entry := close
    stop := close - atrVal * stopMult
    target := close + atrVal * targetMult
    entryBar := bar_index

plot(priorHigh, "Activation: 20-day high", color = #E3C68C, style = plot.style_stepline)
plot(entry, "Entry", color = #C2934A, linewidth = 2, style = plot.style_linebr)
plot(stop, "Stop loss", color = #FB7185, style = plot.style_linebr)
plot(target, "Take profit", color = #34D399, style = plot.style_linebr)
plotshape(entryBar == bar_index, title = "Entry signal", style = shape.triangleup, location = location.belowbar, color = #C2934A, size = size.small)
DONCHIAN·4H

Crypto Donchian · longs

TF · 4H
  1. Regimeprice > SMA200 on 4H closes
  2. Mode 1within 2% of the lower band · RSI < 40 turning up · rejection candle → SL min(band − 1.5%, entry − 0.6%) · TP upper band
  3. Mode 2close > upper band + 0.3% · RSI 55–78 · 6-bar momentum up · volume ≥ 1.2× → SL min(band − 3%, entry − 0.6%) · TP 1.5R
  4. Mode 3price > SMA50 > SMA200 · touch EMA20 (±0.5%) and reclaim · RSI 40–82 → SL min(4-bar low × 0.997, entry − 0.6%) · TP 2R
  5. Gateno trade unless R:R ≥ 1.5
Pine Script v6 · Crypto Donchian 4H
Indicator only · paste into TradingView's Pine editor
//@version=6
indicator("Deltalyn · Crypto Donchian 4H (longs) — blueprint", shorttitle = "VNT DON", overlay = true)

// BTCUSD or ETHUSD on the 4H chart. Every input reads the CLOSED bars before
// the current one and uses this bar's close as the price — what the engine sees.
len        = input.int(20, "Donchian period")
minRR      = input.float(1.5, "Minimum R:R")
stopBuffer = input.float(1.5, "Stop buffer below band %")
minRisk    = input.float(0.60, "Minimum risk %")
edgeTouch  = input.float(2.0, "Edge touch %")
microN     = input.int(6, "Micro-trend bars")
rsiLow     = input.float(40, "RSI oversold")
boRsiMin   = input.float(55, "Breakout RSI min")
boRsiMax   = input.float(78, "Breakout RSI max")
boBuffer   = input.float(0.3, "Breakout buffer %")
boStopPct  = input.float(3.0, "Breakout stop below band %")
boVolMult  = input.float(1.2, "Breakout volume x")
pbEmaLen   = input.int(20, "Pullback EMA")
pbTouch    = input.float(0.5, "Pullback touch %")
pbRsiMin   = input.float(40, "Pullback RSI min")
pbRsiMax   = input.float(82, "Pullback RSI max")
pbRR       = input.float(2.0, "Pullback R:R")
pbStopPct  = input.float(0.60, "Pullback minimum stop %")

// The engine's RSI: plain average of the last 14 gains and losses.
engineRsi(src, n) =>
    d = src - src[1]
    ag = math.sum(math.max(d, 0), n) / n
    al = math.sum(math.max(-d, 0), n) / n
    al == 0 ? 100.0 : 100 - 100 / (1 + ag / al)

price   = close
upper   = ta.highest(high, len)[1]
lower   = ta.lowest(low, len)[1]
sma200  = ta.sma(close, 200)[1]
sma50   = ta.sma(close, 50)[1]
emaF    = ta.ema(close, pbEmaLen)[1]
atrV    = ta.sma(ta.tr(true), 14)[1]
rsiNow  = engineRsi(close, 14)[1]
rsiPrev = engineRsi(close, 14)[2]
micro   = close[1] - close[1 + microN]
bullish = price > sma200
nearLower = math.abs(price - lower) / lower * 100 <= edgeTouch
nearUpper = math.abs(price - upper) / upper * 100 <= edgeTouch
brokeUp   = price > upper * (1 + boBuffer / 100)

// Mode 1 — reversal at the lower band, confirmed by RSI turn + rejection candle.
rng       = high[1] - low[1] == 0 ? 1.0 : high[1] - low[1]
lowerWick = (math.min(close[2], close[1]) - low[1]) / rng
rejection = rsiNow < rsiLow and rsiNow >= rsiPrev and (close[1] > close[2] or lowerWick >= 0.35)
mrStop    = math.min(lower * (1 - stopBuffer / 100), price * (1 - minRisk / 100))
longMR    = nearLower and bullish and micro > -atrV and rejection and (upper - price) / (price - mrStop) >= minRR

// Mode 2 — breakout through the upper band with RSI, momentum and volume.
boVolOk = volume[1] >= ta.sma(volume, 20)[2] * boVolMult
longBO  = not nearLower and brokeUp and bullish and rsiNow >= boRsiMin and rsiNow <= boRsiMax and micro >= atrV and boVolOk
boStop  = math.min(upper * (1 - boStopPct / 100), price * (1 - minRisk / 100))

// Mode 3 — pullback to the EMA inside an uptrend, mid-channel only.
upTrend = price > sma50 and sma50 > sma200
longPB  = not nearLower and not nearUpper and not brokeUp and upTrend and low[1] <= emaF * (1 + pbTouch / 100) and price > emaF and rsiNow >= pbRsiMin and rsiNow <= pbRsiMax and micro >= atrV
pbStop  = math.min(ta.lowest(low, 4)[1] * 0.997, price * (1 - pbStopPct / 100))

var float entry = na
var float stop = na
var float target = na
exited = false
if not na(entry)
    if low <= stop or high >= target
        entry := na
        stop := na
        target := na
        exited := true
if na(entry) and not exited
    if longMR
        entry := price
        stop := mrStop
        target := upper
    else if longBO
        entry := price
        stop := boStop
        target := price + minRR * (price - boStop)
    else if longPB
        entry := price
        stop := pbStop
        target := price + pbRR * (price - pbStop)

plot(upper, "Donchian upper", color = color.new(#98A2B3, 45), style = plot.style_stepline)
plot(lower, "Donchian lower", color = color.new(#98A2B3, 45), style = plot.style_stepline)
plot(emaF, "EMA 20 (pullback activation)", color = #E3C68C)
plot(entry, "Entry", color = #C2934A, linewidth = 2, style = plot.style_linebr)
plot(stop, "Stop loss", color = #FB7185, style = plot.style_linebr)
plot(target, "Take profit", color = #34D399, style = plot.style_linebr)
STAT-ARB

Pairs mean reversion

TF · 1D
  1. Spreadlog(A) − β·log(B), β from 90 days
  2. Scorez over the last 60 days
  3. GateAR(1) half-life between 3 and 60 days
  4. Entry|z| ≥ 2 · R:R ≥ 1.2
  5. Stop|z| reaches 3.5
  6. Target|z| back to 0.5
  7. Timeexit after 30 sessions
Pine Script v6 · Pairs z-score
Indicator only · paste into TradingView's Pine editor
//@version=6
indicator("Deltalyn · Pairs z-score — blueprint", shorttitle = "VNT PAIRS", overlay = false)

// Daily. Spread = log(A) - beta * log(B); z-scored with TODAY's beta over the
// window, as the engine does. (Its AR(1) half-life gate is not replicated.)
legA    = input.symbol("COINBASE:ETHUSD", "Leg A")
legB    = input.symbol("COINBASE:BTCUSD", "Leg B")
betaLen = input.int(90, "Hedge-ratio lookback")
zLen    = input.int(60, "Z-score lookback")
entryZ  = input.float(2.0, "Entry |z|")
exitZ   = input.float(0.5, "Exit |z|")
stopZ   = input.float(3.5, "Stop |z|")
minRR   = input.float(1.2, "Minimum R:R")

a = math.log(request.security(legA, "D", close, lookahead = barmerge.lookahead_off))
b = math.log(request.security(legB, "D", close, lookahead = barmerge.lookahead_off))
beta = ta.covariance(a, b, betaLen) / ta.variance(b, betaLen)
spreadMean = ta.sma(a, zLen) - beta * ta.sma(b, zLen)
spreadVar  = ta.variance(a, zLen) + beta * beta * ta.variance(b, zLen) - 2 * beta * ta.covariance(a, b, zLen)
z  = (a - beta * b - spreadMean) / math.sqrt(spreadVar)
rr = (math.abs(z) - exitZ) / (stopZ - math.abs(z))
signal = math.abs(z) >= entryZ and math.abs(z) < stopZ and rr >= minRR

plot(z, "Spread z-score", color = #98A2B3, linewidth = 2)
hline(0, "Mean", color = color.new(#98A2B3, 60))
hline(entryZ, "Entry band +", color = #E3C68C, linestyle = hline.style_dashed)
hline(-entryZ, "Entry band -", color = #E3C68C, linestyle = hline.style_dashed)
hline(exitZ, "Take profit +", color = #34D399, linestyle = hline.style_dotted)
hline(-exitZ, "Take profit -", color = #34D399, linestyle = hline.style_dotted)
hline(stopZ, "Stop +", color = #FB7185)
hline(-stopZ, "Stop -", color = #FB7185)
plotshape(signal, title = "Entry signal", style = shape.circle, location = location.absolute, color = #C2934A, size = size.tiny)
TJL

Trend Join Long

TF · Intraday
  1. Dailyclose > prior-day high · prior close > SMA200
  2. Intradayclose > premarket high and > today's high so far
  3. Andweekly close > weekly SMA20 · relative volume ≥ 1.5×
  4. Stopentry − 1.5 × ATR14
  5. Targetentry + 3 × ATR14
  6. Window10:00–15:30 ET, Mon–Fri
Pine Script v6 · Trend Join Long
Indicator only · paste into TradingView's Pine editor
//@version=6
indicator("Deltalyn · Trend Join Long — blueprint", shorttitle = "VNT TJL", overlay = true)

// Intraday chart (e.g. 5m) with extended hours ON. Same entry filters as the
// scanner: daily breakout over SMA200, premarket-high break, weekly trend, volume.
atrLen        = input.int(14, "ATR length")
stopMult      = input.float(1.5, "Stop = entry - ATR x")
targetMult    = input.float(3.0, "Target = entry + ATR x")
minRelVolume  = input.float(1.5, "Min relative volume (vs 20 bars)")
requireWeekly = input.bool(true, "Require weekly close > weekly SMA20")

prevDailyHigh  = request.security(syminfo.tickerid, "D", high[1], lookahead = barmerge.lookahead_off)
prevDailyClose = request.security(syminfo.tickerid, "D", close[1], lookahead = barmerge.lookahead_off)
dailySma       = request.security(syminfo.tickerid, "D", ta.sma(close, 200)[1], lookahead = barmerge.lookahead_off)
weeklyClose    = request.security(syminfo.tickerid, "W", close[1], lookahead = barmerge.lookahead_off)
weeklySma20    = request.security(syminfo.tickerid, "W", ta.sma(close, 20)[1], lookahead = barmerge.lookahead_off)

dailyBreakout = close > prevDailyHigh and prevDailyClose > dailySma
weeklyOk      = not requireWeekly or weeklyClose > weeklySma20
volumeOk      = volume > ta.sma(volume, 20) * minRelVolume

var float pmh = na
var float hod = na
if ta.change(time("D")) != 0
    pmh := na
    hod := na
if session.ispremarket
    pmh := na(pmh) ? high : math.max(pmh, high)
hodBefore = hod
if session.ismarket
    hod := na(hod) ? high : math.max(hod, high)
intradayBreakout = session.ismarket and not na(pmh) and not na(hodBefore) and close > pmh and close > hodBefore

atrVal = ta.atr(atrLen)
signal = dailyBreakout and intradayBreakout and weeklyOk and volumeOk

var float entry = na
var float stop = na
var float target = na
exited = false
if not na(entry)
    if low <= stop or high >= target
        entry := na
        stop := na
        target := na
        exited := true
if na(entry) and not exited and signal
    entry := close
    stop := close - atrVal * stopMult
    target := close + atrVal * targetMult

plot(prevDailyHigh, "Activation: prior-day high", color = #E3C68C, style = plot.style_stepline)
plot(pmh, "Premarket high", color = color.new(#98A2B3, 40), style = plot.style_circles)
plot(entry, "Entry", color = #C2934A, linewidth = 2, style = plot.style_linebr)
plot(stop, "Stop loss", color = #FB7185, style = plot.style_linebr)
plot(target, "Take profit", color = #34D399, style = plot.style_linebr)
PROBABILITY

Setup evidence engine

TF · 1D
  1. SetupsDonchian-20 · Donchian-55 · trend pullback · RSI-60 momentum
  2. Entrynext bar's open — never the triggering close
  3. Stopentry − 2 × ATR14
  4. Target2R · cap 40 sessions
  5. Tiestop and target in one bar → stop counts
  6. Gradeexpectancy at the low end of the Wilson 95% interval
HMM REGIME

Hidden-state regime

TF · 1D
  1. Model2-state Gaussian HMM, Baum-Welch fit on daily log returns
  2. Liveforward-only filtering — no smoothing, no future bars
  3. Outputstate probabilities that scale the crypto engines; never an entry
RISK GATE

Portfolio risk gate

TF · Every ticket
  1. R01size from the distance to the stop
  2. R02≤ 2 positions per correlation cluster
  3. R05−8% monthly circuit breaker halts new entries
  4. R06heat alert above 4.0R open risk · fail-closed on a bad price
X: 030.00Deltalyn · System blueprint
Sheet 04 / 05Doc: ENG-BACKTEST-04
04 — No backtest, no confidence

The last trade each engine took.

Real candles, the engine's own code, and its most recent closed trade in backtest — entry, stop loss and take profit drawn to the cent. Every chart here is history, not a signal.

▲ BUY entry▼ SELL · hit target▼ SELL · stopped out▼ SELL · closed on time
ROTATION·LEV · engine backtest

Leveraged rotation

Each 3× ETF trades only while its own index is in a calm uptrend. The trigger is a close above the prior 20-day high on volume.

TQQQ · Nasdaq-100 3xTF · 1D
Backtest · not a live result
Trade spec
Entry
35.98
close · 2024-06-12
SL
34.08
entry − 1.5×ATR14
TP
39.79
entry + 3×ATR14
Exit
Take profit hit
2024-07-05 · 15 sessions · +2.01R
Trades
17
Win rate
52.9%
Sample 2017-12-15 → 2024-07-05
Avg per trade
+0.58R
total +9.94R
Median hold
11
days

Run with the engine's own backtestPair(): reproduced trade for trade, and identical to its scan of 2026-09-11 17:49 ET.

SOXL · Semiconductors 3xTF · 1D
Backtest · not a live result
Trade spec
Entry
58.08
close · 2026-01-15
SL
52.83
entry − 1.5×ATR14
TP
68.58
entry + 3×ATR14
Exit
Take profit hit
2026-01-28 · 8 sessions · +2.00R
Trades
25
Win rate
44.0%
Sample 2017-10-18 → 2026-01-28
Avg per trade
+0.32R
total +8.01R
Median hold
10
days

Run with the engine's own backtestPair(): reproduced trade for trade, and identical to its scan of 2026-09-11 17:49 ET.

SPXL · S&P 500 3xTF · 1D
Backtest · not a live result
Trade spec
Entry
227.57
close · 2025-12-11
SL
218.27
entry − 1.5×ATR14
TP
246.17
entry + 3×ATR14
Exit
Stop loss hit
2025-12-12 · 1 session · -1.00R
Trades
19
Win rate
47.4%
Sample 2017-10-04 → 2025-12-12
Avg per trade
+0.42R
total +7.98R
Median hold
9
days

Run with the engine's own backtestPair(): reproduced trade for trade, and identical to its scan of 2026-09-11 17:49 ET.

DONCHIAN·4H · replayed backtest · longs

Crypto Donchian channel

Dotted lines are the 20-bar channel the engine saw at each close; the solid grey line is its EMA20.

BTCUSD · Donchian 20TF · 4H
Backtest · not a live result
Trade spec
Entry
78,854.14
long pullback · 2026-08-30 16:00 UTC
SL
77,711.10
min(4-bar low × 0.997, entry − 0.60%)
TP
81,140.23
entry + 2R
Exit
Stop loss hit
2026-08-30 20:00 UTC · 1 bar · -1.00R
Trades
25
Win rate
36.0%
Sample 2024-11-02 → 2026-08-30
Avg per trade
+0.05R
total +1.29R
Median hold
17
4H bars

The engine keeps no backtest of its own. This replays its decide() once per closed 4H bar over 4,379 bars of Coinbase BTC-USD, 1H candles aggregated to UTC 4H — longs only, exit at SL or TP, both in one bar counts as SL. Its channel matches the engine's live scan within 0.02%.

ETHUSD · Donchian 20TF · 4H
Backtest · not a live result
Trade spec
Entry
2,471.60
long pullback · 2026-09-01 00:00 UTC
SL
2,426.12
min(4-bar low × 0.997, entry − 0.60%)
TP
2,562.56
entry + 2R
Exit
Stop loss hit
2026-09-01 16:00 UTC · 4 bars · -1.00R
Trades
40
Win rate
40.0%
Sample 2024-11-06 → 2026-09-01
Avg per trade
+0.26R
total +10.33R
Median hold
11
4H bars

The engine keeps no backtest of its own. This replays its decide() once per closed 4H bar over 4,379 bars of Coinbase ETH-USD, 1H candles aggregated to UTC 4H — longs only, exit at SL or TP, both in one bar counts as SL. Its channel matches the engine's live scan exactly. 1 simulated position still open at the end of the data is not drawn — it would read as a live signal.

STAT-ARB · engine backtest

Pairs mean reversion

The priority pair is drawn; the table holds the engine's backtest for all four pairs, losers included.

ETH vs BTC · spread zTF · 1D
Backtest · not a live result
Trade spec
Entry
z -2.43
2026-09-03 · long ETH · short BTC
SL
z -3.50
|z| reaches the stop band
TP
z -0.50
|z| back to the exit band
Exit
Take profit hit
2026-09-10 · 7 sessions · +1.80R
Trades
99
Win rate
23.2%
Sample 2017-03-10 → 2026-09-10
Avg per trade
-0.17R
total -16.54R
Median hold
9
days

Reported in z-score units, never in price — the engine's own convention. Exits are scored in the entry-day frame (β, mean, σ held fixed).

PairTradesWin rateAvg / tradeTotalLast tradeReproduced
ETH vs BTC9923.2%-0.17R-16.54R2026-09-03 · TP +1.80R✓ engine code
COIN vs ETH3342.4%+0.08R+2.67R2026-08-19 · TIME +0.18R✓ engine code
MSTR vs ETH7538.7%+0.49R+36.69R2026-07-10 · SL -1.00R✓ engine code
CRCL vs COIN333.3%+0.55R+1.66R2026-07-01 · SL -1.00R✓ engine code
PROBABILITY · public evidence output

Setup evidence engine

Grades every setup from the pessimistic end of a 95% interval. A small, flattering sample gets a D.

Grade A
19
≥ +0.15R with n ≥ 50
Grade B
14
≥ +0.05R with n ≥ 30
Grade C
7
> 0
Grade D
44
no demonstrable edge
Coverage
84
21 assets × 4 setups · Wilson 95%
Grade matrix · cell = grade · n cases · pessimistic RGraded 2026-09-07
AssetDonchian-20Donchian-55Trend pullbackPullback v2RSI-60 momentumTJL daily
AAPL
An 260 · +0.32R
Bn 185 · +0.12R
Dn 146 · -0.12R
—
An 118 · +0.35R
—
AMZN
Dn 216 · -0.21R
Dn 139 · -0.37R
Dn 128 · -0.08R
—
Dn 124 · -0.00R
—
COIN
An 75 · +0.15R
Bn 45 · +0.32R
Dn 50 · -0.79R
—
Dn 35 · -0.27R
—
ETHUSD
An 237 · +0.22R
An 172 · +0.24R
Dn 210 · -0.01R
—
An 139 · +0.30R
—
GE
An 189 · +0.33R
An 150 · +0.35R
Dn 131 · -0.09R
—
An 104 · +0.22R
—
IREN
An 84 · +0.29R
An 51 · +0.35R
Dn 59 · -0.11R
—
Bn 39 · +0.23R
—
LCID
Dn 37 · -0.59R
Dn 12 · -0.86R
Dn 40 · -0.57R
—
Dn 28 · -0.61R
—
MSTR
Dn 164 · -0.14R
Dn 98 · -0.23R
Dn 134 · -0.24R
—
Dn 92 · -0.30R
—
NVDA
An 238 · +0.28R
An 177 · +0.35R
Cn 179 · +0.01R
—
An 114 · +0.17R
—
NVDL
Cn 72 · +0.02R
Cn 50 · +0.04R
Dn 43 · -0.09R
—
Dn 27 · -0.36R
—
PLTR
Dn 111 · -0.01R
Dn 74 · -0.06R
Dn 68 · -0.11R
—
Dn 63 · -0.01R
—
QQQ
An 325 · +0.31R
An 258 · +0.25R
Dn 153 · -0.25R
—
An 123 · +0.46R
—
SMCI
Dn 147 · -0.08R
Dn 101 · -0.16R
Dn 127 · -0.40R
—
Bn 100 · +0.06R
—
SMH
Bn 277 · +0.09R
Bn 202 · +0.11R
Bn 173 · +0.10R
—
Dn 127 · -0.06R
—
SOXL
Dn 238 · -0.05R
Dn 163 · -0.05R
Dn 125 · -0.12R
—
Cn 127 · +0.03R
—
SOXX
Bn 274 · +0.06R
Bn 202 · +0.11R
Cn 160 · +0.03R
—
Dn 129 · -0.03R
—
SPXL
Cn 318 · +0.04R
Dn 244 · -0.09R
Dn 133 · -0.20R
—
Bn 123 · +0.09R
—
SPY
Bn 342 · +0.10R
Dn 269 · -0.01R
Dn 145 · -0.09R
—
Bn 123 · +0.14R
—
TQQQ
An 293 · +0.18R
Bn 220 · +0.09R
Dn 151 · -0.23R
—
An 117 · +0.47R
—
TSLA
Cn 181 · +0.00R
Bn 123 · +0.15R
Dn 148 · -0.39R
—
Dn 91 · -0.18R
—
TSLL
Dn 53 · -0.26R
Dn 34 · -0.11R
Dn 32 · -0.82R
—
Dn 26 · -0.59R
—
ETHUSD · Donchian-20TF · 1D
Backtest · not a live result
Trade spec
Entry
2,251.68
next open · signal 2026-08-19
SL
2,103.79
entry − 2×ATR14
TP
2,547.46
entry + 2R
Exit
Take profit hit
2026-08-21 · 2 sessions · +2.00R
Grade
A
Donchian-20 · ETHUSD
Cases
237
Hit rate
47%
CI 41%–54%
Expectancy
+0.40R
pessimistic +0.22R

Levels rebuilt from the engine's published rules — next-bar open, 2×ATR14 stop, 2R target, 40-session cap — and checked against its own recorded outcomes: 73/73 ETHUSD trades match exactly.

No publishable backtest

Engines with nothing to draw

Shown on purpose: an empty frame beats an invented trade.

No data · TJL

Trend Join Long keeps no in-engine backtest and scans only on intraday bars, 10:00–15:30 ET. Drawing a trade here would mean inventing one. Its rules are mirrored in Pine on sheet 03.

No data · HMM regime

Produces no entries by design: it infers the market's hidden state and scales the other engines. No entry, stop or target exists to draw. Members see today's state on sheet 02.

No data · Risk gate

Its inputs are live positions, which are private and never published. The rules it enforces are listed on sheet 03.

No data · ETH accumulation

Its output depends on private capital settings, so nothing from it is published on this page.

X: 040.00Deltalyn · System blueprint
Sheet 05 / 05Doc: ENG-AUDIT-05
05 — Audit the engine against itself

Pass, review, or fail.

Seven questions every backtest has to answer. A verdict is given only where the system actually computes the check; everything else is marked not evaluated — and that alone keeps an engine from a clean pass.

Pass
0
subjects
Review
4
subjects
Fail
6
subjects
Engine · subjectLook-ahead biasOverfittingSample sizeFees & slippageWeak stopRegime dependenceExpectancyVerdict
TQQQ
Leveraged rotation · 1D
PassNot eval.FailNot eval.Not eval.Not eval.PassFail
SOXL
Leveraged rotation · 1D
PassNot eval.FailNot eval.Not eval.Not eval.PassFail
SPXL
Leveraged rotation · 1D
PassNot eval.FailNot eval.Not eval.Not eval.PassFail
BTC
Crypto Donchian · long · 4H
PassNot eval.FailNot eval.Not eval.Not eval.PassFail
ETH
Crypto Donchian · long · 4H
PassNot eval.ReviewNot eval.Not eval.Not eval.PassReview
ETH vs BTC
Pairs mean reversion · 1D
PassNot eval.PassNot eval.Not eval.Not eval.FailFail
COIN vs ETH
Pairs mean reversion · 1D
PassNot eval.ReviewNot eval.Not eval.Not eval.PassReview
MSTR vs ETH
Pairs mean reversion · 1D
PassNot eval.PassNot eval.Not eval.Not eval.PassReview
CRCL vs COIN
Pairs mean reversion · 1D
PassNot eval.FailNot eval.Not eval.Not eval.PassFail
ETHUSD · Donchian-20
Setup evidence · 1D
PassNot eval.PassNot eval.Not eval.ReviewPassReview

Sample-size thresholds are the evidence engine's own (n ≥ 50 for a pass, 30–49 for review). Expectancy is gross — no fees or slippage are modelled, which is exactly why that check reads not evaluated. TJL, the HMM, the risk gate and ETH accumulation have no publishable backtest and are not audited here.

Evidence behind every verdict +
TQQQ · Leveraged rotation
  • PassLook-ahead bias: Walk-forward: each day sees only bars up to itself; the 20-day high excludes today.
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • FailSample size: 17 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: Trades only inside its own uptrend filter; no out-of-regime split.
  • PassExpectancy: +0.58R per trade, gross.
SOXL · Leveraged rotation
  • PassLook-ahead bias: Walk-forward: each day sees only bars up to itself; the 20-day high excludes today.
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • FailSample size: 25 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: Trades only inside its own uptrend filter; no out-of-regime split.
  • PassExpectancy: +0.32R per trade, gross.
SPXL · Leveraged rotation
  • PassLook-ahead bias: Walk-forward: each day sees only bars up to itself; the 20-day high excludes today.
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • FailSample size: 19 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: Trades only inside its own uptrend filter; no out-of-regime split.
  • PassExpectancy: +0.42R per trade, gross.
BTC · Crypto Donchian · long
  • PassLook-ahead bias: decide() only sees closed 4H bars before the entry bar.
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • FailSample size: 25 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: SMA200 regime filter is built in; no split is computed.
  • PassExpectancy: +0.05R per trade, gross.
ETH · Crypto Donchian · long
  • PassLook-ahead bias: decide() only sees closed 4H bars before the entry bar.
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • ReviewSample size: 40 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: SMA200 regime filter is built in; no split is computed.
  • PassExpectancy: +0.26R per trade, gross.
ETH vs BTC · Pairs mean reversion
  • PassLook-ahead bias: Walk-forward; exits are scored in the entry-day frame (β, mean, σ).
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • PassSample size: 99 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: No per-regime split is computed.
  • FailExpectancy: -0.17R per trade, gross.
COIN vs ETH · Pairs mean reversion
  • PassLook-ahead bias: Walk-forward; exits are scored in the entry-day frame (β, mean, σ).
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • ReviewSample size: 33 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: No per-regime split is computed.
  • PassExpectancy: +0.08R per trade, gross.
MSTR vs ETH · Pairs mean reversion
  • PassLook-ahead bias: Walk-forward; exits are scored in the entry-day frame (β, mean, σ).
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • PassSample size: 75 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: No per-regime split is computed.
  • PassExpectancy: +0.49R per trade, gross.
CRCL vs COIN · Pairs mean reversion
  • PassLook-ahead bias: Walk-forward; exits are scored in the entry-day frame (β, mean, σ).
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • FailSample size: 3 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • Not eval.Regime dependence: No per-regime split is computed.
  • PassExpectancy: +0.55R per trade, gross.
ETHUSD · Donchian-20 · Setup evidence
  • PassLook-ahead bias: Entry at the next bar's open — replayed 73/73 trades exactly.
  • Not eval.Overfitting: No out-of-sample or parameter-sensitivity run is computed.
  • PassSample size: 237 closed trades · grade rules want n ≥ 50.
  • Not eval.Fees & slippage: Backtest is gross: no commission or slippage model.
  • Not eval.Weak stop: Stops are rule-based; no stop-sensitivity test is computed.
  • ReviewRegime dependence: Pessimistic R: 0.04 above SMA200, -0.21 below.
  • PassExpectancy: +0.22R per trade at the pessimistic end of the interval, gross.
X: 050.00Deltalyn · System blueprint
Sources

Daily candles from the engine's local 10-year end-of-day history; crypto 4H candles from Coinbase's public API. Backtests run the engine's own code. Evidence grades as published by the engine itself.

What this is not

Research, not advice. No page here places an order or suggests an amount. Backtests are gross of fees and slippage; a past trade drawn on a chart is not a forecast.

Revision

Data to close 2026-09-11
Exported 2026-09-12 02:58 UTC

Deltalyn is a quantitative research terminal for educational and informational purposes. We are not a registered investment adviser, and no signal, backtest, or figure on this site constitutes personalized advice to buy or sell any security. The platform does not execute orders, does not custody funds, and does not manage third-party capital. Historical or simulated ("backtest") performance does not guarantee future results; trade at your own judgment, with your own risk management, and with independent licensed financial advice if you need it.