Abstract
Project Montauk is a systematic swing‑trading framework that applies a single‑position, all‑in or all‑out approach to the Direxion Technology Bull 3× Shares ETF (TECL). From 01/01/2009 through 06/30/2025, an initial $1,000 compounded to approximately $582,600 across 11 trades, outperforming buy‑and‑hold TECL ($108,000) and QQQ ($14,000) while reducing maximum drawdown from 77% to 57%. The strategy achieves a profit factor of 5.94 and a monthly Sharpe ratio near 0.98 .
Introduction
The objective of Project Montauk is to magnify the long‑term structural growth of United States mega‑cap technology equities while sidestepping extended bear phases. The engine exploits three structural advantages:
Leverage with discipline: entries occur only during established uptrends; exits trigger promptly on verified weakness.
Cap‑weighted index mechanics: the S&P technology sector continuously replaces laggards with leaders, mitigating individual issuer risk.
Trend duration focus: trades target multi‑month impulses rather than short‑term price fluctuations.
Rationale for Long‑Horizon Leverage
Daily reset tailwind: over multi‑year horizons, triple leverage compounds the sector’s net drift despite short‑term decay.
Quality filter: S&P eligibility (≥ $15 billion float‑adjusted market cap and GAAP profitability) biases the basket toward market leaders.
Structural inflows: recurring retirement contributions create a persistent bid for cap‑weighted index funds, indirectly lifting TECL.
Technology growth runway: secular adoption of cloud infrastructure, generative AI, and edge computing sustains double‑digit revenue expansion.
Asset Selection
Alternative funds (FNGU, SOXL, TQQQ) were rejected due to right‑to‑call clauses, narrow industry scope, or shorter price history. TECL offers deep liquidity, index continuity to the late 1990s, and diversified exposure across the S&P technology sector.
Methodology
Data: daily TECL bars from Yahoo Finance imported into TradingView.
Walk‑forward: 70% development window, 30% out‑of‑sample, rolled forward annually.
Assumptions: Position size fixed at 100 % of equity; no additional margin.
Validation: each rule revision must improve return‑to‑risk after costs.
Strategy Evolution (01/01/2009 – 06/30/2025)
Across the full test window the strategy advanced through five primary revisions. Version 1.0 paired an ATR stop with a MACD thrust, generating 42 trades, a 19% compound annual growth rate (CAGR), and a 67% maximum drawdown. Version 3.0 added a 2,000‑day EMA regime gate, halving trade count to 21 while lifting CAGR to 31% and trimming drawdown to 54%. Version 5.0 adopted a three‑lane EMA stack; trades fell to 15 and CAGR climbed to 43%, with drawdown easing to 44%. Version 6.4 compared the price line with a 500‑day EMA, recording 14 trades, a 47% CAGR, and the strategy’s lowest drawdown, 41%. The final Version 7.7 blended the ATR logic with a slope‑aware exit, cutting trades to 11 and nudging CAGR toward 50%, albeit at the cost of a wider 57% drawdown. Early builds prioritised noise suppression and regime detection, whereas later iterations leaned on adaptive EMA hierarchies and slope‑based exits to curb slippage and focus entries on sustained momentum.
Earlier iterations emphasized noise suppression (buffers, cool‑downs) and regime detection (ultra‑long EMAs). Later versions replaced MACD logic with adaptive EMA stacks and slope‑aware exits, reducing trade frequency and slippage.
Current Algorithm (Version 7.7)
lenFast = 15
lenMid = 30
lenLong = 500
lenSlope = 70
emaFast = ta.ema(close, lenFast)
emaMid = ta.ema(close, lenMid)
emaLong = ta.ema(close, lenLong)
emaSlope = ta.ema(close, lenSlope)
atrStop = strategy.position_size > 0 ? strategy.position_avg_price - 3 * ta.atr(40) : na
upTrend = emaFast > emaMid and emaMid > emaLong and ta.rate(emaSlope, 10) > 0.1
enter = upTrendcrossDn = ta.crossunder(emaFast, emaLong)
slopeDn = ta.rate(emaFast, 5) < -1.3
exitSig = (close < atrStop) or (crossDn[1] and crossDn) or slopeDn
strategy.entry("Long", strategy.long, when = enter and strategy.opentrades == 0)
strategy.close("Long", when = exitSig)
lenMid = 30
lenLong = 500
lenSlope = 70
emaFast = ta.ema(close, lenFast)
emaMid = ta.ema(close, lenMid)
emaLong = ta.ema(close, lenLong)
emaSlope = ta.ema(close, lenSlope)
atrStop = strategy.position_size > 0 ? strategy.position_avg_price - 3 * ta.atr(40) : na
upTrend = emaFast > emaMid and emaMid > emaLong and ta.rate(emaSlope, 10) > 0.1
enter = upTrendcrossDn = ta.crossunder(emaFast, emaLong)
slopeDn = ta.rate(emaFast, 5) < -1.3
exitSig = (close < atrStop) or (crossDn[1] and crossDn) or slopeDn
strategy.entry("Long", strategy.long, when = enter and strategy.opentrades == 0)
strategy.close("Long", when = exitSig)
Performance Summary (01/01/2009 – 06/30/2025)
Net equity: $582,603 from $1,000 initial capital.
Number of trades: 11 (10 wins, 1 loss).
Profit factor: 5.94.
Peak drawdown: 57 percent, versus 77 percent for buy‑and‑hold TECL.
Median intra‑trade give‑back: 8.4%.
Risk Management Features
ATR‑based trailing stop (3 × 40‑day ATR).
EMA hierarchy to confirm regime alignment.
Quick‑slope filter to catch rapid breakdowns.
Two‑bar cool‑down to suppress immediate re‑entry.
Limitations
Reliance on daily bars excludes intraday dynamics and may miss gap risk.
Back‑test history limited to post‑2009 bull cycle; performance in prolonged sideways markets is untested.
Strategy assumes uninterrupted ETF operation; extreme market events could trigger fund restructuring.
Future Work
Probabilistic position scaling: transition from binary exits to gradient de‑risking.
Volatility‑adaptive stops: widen in calm regimes, tighten in turbulence.
Cross‑symbol rotation: periodic comparison with SOXL to capture semiconductor‑specific momentum.
Composite trend scoring: integrate EMA slope, MACD distance, and relative strength into a unified entry threshold.
Conclusion
Project Montauk demonstrates that a disciplined, rule‑based framework can harness triple leverage in mega‑cap technology equities while materially reducing downside exposure. Historical testing indicates substantial alpha generation relative to passive alternatives. Continued refinement will focus on adaptive risk controls and rotational overlays to preserve outperformance in evolving market conditions.