Systematic trading models, built and audited the way an allocator would.
Forward-tested on paper capital across 41 systematic models since Dec 2025. Published by DataFortress.cloud GmbH as a passive quantitative research record.
Engineering & governance
How the book is run, and what stops it from going wrong:
Containerised execution
Every strategy runs as an isolated Kubernetes CronJob. Orders, positions and daily marks are written to PostgreSQL by the same pipeline for every model.
Trade-only, non-withdrawal keys
Broker API keys are scoped to trade-only permissions where the venue supports it, held as Kubernetes Secrets with restricted RBAC and rotated at least quarterly. Withdrawal and transfer scopes are never enabled.
Drawdown kill-switch and circuit breakers
Each strategy carries a -25%-from-seed kill switch that suspends it pending human review, alongside per-order position-size caps and daily-loss circuit breakers.
Tamper-evident audit trail
Daily portfolio marks are inserted, never updated, in an append-only table. Institutional allocators can request read-only access to the audit trail under NDA.
Model-validation approach
Quantitative models are evaluated through institutional allocator diligence rather than raw backtest metrics:
Benchmark-adjusted alpha
We compute annualised Jensen's alpha against the closest relevant benchmark (SPY, QQQ, or FTWD). A model must prove risk-adjusted excess return after isolating market beta, ensuring levered index beta is not claimed as alpha.
Statistical significance (t-stat & n.s.)
Alpha is evaluated with Student's t-distribution. Ratios with |t| < 2.0 are flagged as n.s. (not statistically significant), distinguishing repeatable structural edge from random variance on limited history.
Multi-asset correlation
Correlation is continuously benchmarked against equity and global asset indices (SPY, QQQ, FTWD). Reporting the highest correlation prevents equity-tracking strategies from misrepresenting themselves as market-neutral.
Strict live vs. backtest separation
Historical backtests serve solely as reference for design thesis and sensitivity analysis. Only automated daily forward-test marks are used for screening and performance reporting.
Model roster — forward-test record
Ranked by Sortino ratio (downside-adjusted risk premium). Every strategy is forward-tested daily against real market data from a common seed; returns are directly comparable across the table.
Reading the metrics. Max corr is the highest correlation to any of our three buy-and-hold benchmarks (SPY, QQQ, FTWD), with the closest one named. Alpha is annualised Jensen’s alpha against that closest benchmark: the return left over after the strategy’s benchmark exposure is accounted for. t is the t-statistic of that alpha; n.s. marks estimates that are not statistically significant (|t| < 2) and indistinguishable from zero at this sample size. Trades is lifetime executed trades. Allocator screen reflects minimum institutional criteria (Sortino ≥ 1.5, Calmar ≥ 0.8, Max DD < 20%, Max Corr < 0.40, with ≥ 90 days live history).
| # | Strategy | Sortino | Calmar | CAGR | Max DD | Alpha (ann.) | t | Max corr | Trades | Days live | Allocator screen |
|---|---|---|---|---|---|---|---|---|---|---|---|
| 1 | Benchmark_FTWD Benchmark | 2.59 | 3.61 | 19.26% | -5.33% | +10.9% vs SPY | 1.25 n.s. | 0.60 (SPY) | 1 | 216 | — |
| 2 | option_CollarBot | 2.46 | 18.89 | 31.08% | -1.65% | +12.7% vs GLD | 0.56 n.s. | -0.53 (GLD) | 3 | 32 | Insufficient history |
| 3 | option_WheelBot | 2.46 | 18.80 | 20.17% | -1.07% | +10.1% vs GLD | 0.76 n.s. | -0.41 (GLD) | 3 | 32 | Insufficient history |
| 4 | EarningsInsiderTiltBot | 2.45 | 4.72 | 22.84% | -4.84% | +7.4% vs SPY | 0.75 n.s. | 0.75 (SPY) | 1560 | 185 | Fail |
| 5 | TARegimeAdaptiveBot | 2.36 | 3.03 | 16.33% | -5.39% | +7.0% vs QQQ | 1.10 n.s. | 0.74 (QQQ) | 5 | 185 | Fail |
| 6 | RegimeAdaptiveBot | 2.22 | 4.24 | 9.85% | -2.32% | +2.9% vs SPY | 0.58 n.s. | 0.74 (SPY) | 1370 | 185 | Fail |
| 7 | AIHedgeFundBot | 2.10 | 2.43 | 37.28% | -15.34% | +13.1% vs SPY | 0.59 n.s. | 0.69 (SPY) | 18 | 190 | Fail |
| 8 | Benchmark_SPY Benchmark | 1.84 | 2.26 | 14.22% | -6.30% | +4.8% vs QQQ | 0.83 n.s. | 0.87 (QQQ) | 1 | 216 | — |
| 9 | AdaptiveMeanReversionBot | 1.69 | 4.21 | 30.63% | -7.27% | -4.9% vs QQQ | -0.84 n.s. | 0.97 (QQQ) | 1 | 160 | Fail |
| 10 | option_LeapCallBot | 1.48 | 11.03 | 43.60% | -3.95% | +3.6% vs GLD | 0.09 n.s. | -0.51 (GLD) | 1 | 35 | Insufficient history |
| 11 | KronosTraderBot | 1.44 | 3.52 | 5.67% | -1.61% | +0.1% vs SPY | 0.02 n.s. | 0.53 (SPY) | 981 | 151 | Fail |
| 12 | FearGreedBotQQQInverse | 1.23 | 2.41 | 17.52% | -7.27% | +2.3% vs QQQ | 0.50 n.s. | 0.97 (QQQ) | 1 | 215 | Fail |
| 13 | Benchmark_QQQ Benchmark | 1.21 | 2.28 | 16.55% | -7.27% | -2.6% vs SPY | -0.28 n.s. | 0.87 (SPY) | 1 | 216 | — |
| 14 | DeepSeekToolBot | 1.08 | 1.47 | 11.35% | -7.72% | -5.0% vs SPY | -0.49 n.s. | 0.82 (SPY) | 232 | 185 | Fail |
| 15 | option_PMCCBot | 1.07 | 8.17 | 25.91% | -3.17% | +12.6% vs GLD | 0.29 n.s. | -0.22 (GLD) | 4 | 32 | Insufficient history |
| 16 | XAUSyntheticMetalTreeBot | 0.76 | 1.38 | 16.91% | -12.25% | +5.1% vs SPY | 0.25 n.s. | 0.49 (SPY) | 43 | 215 | Fail |
| 17 | GoldenButterflyMomBot | 0.73 | 1.38 | 3.34% | -2.41% | -6.3% vs SPY | -1.63 n.s. | 0.79 (SPY) | 17 | 160 | Fail |
| 18 | SharpePortfolioOptWeeklyBot | 0.67 | 0.43 | 2.89% | -6.76% | -1.9% vs FTWD | -0.31 n.s. | 0.41 (FTWD) | 740 | 215 | Fail |
| 19 | RecursiveDecayHarvestBot | 0.52 | 0.43 | 12.54% | -28.99% | -62.3% vs QQQ | -2.37 | 0.92 (QQQ) | 3 | 160 | Fail |
| 20 | FearGreedBotQQQ | 0.45 | 0.36 | 5.08% | -14.02% | -7.1% vs QQQ | -0.89 n.s. | 0.88 (QQQ) | 53 | 216 | Fail |
| 21 | StockNewsSentimentBot | 0.43 | 0.95 | 3.94% | -4.12% | +1.6% vs BTC | 0.15 n.s. | 0.36 (BTC) | 53 | 160 | Fail |
| 22 | XAUZenbotTreeBot | 0.37 | 2.31 | 10.45% | -4.52% | +11.2% vs GLD | 1.00 n.s. | 0.24 (GLD) | 20 | 215 | Fail |
| 23 | AdaptiveMeanReversionBTCBot | 0.20 | -0.05 | -0.47% | -8.93% | +4.1% vs QQQ | 0.09 n.s. | 0.58 (QQQ) | 1 | 74 | Insufficient history |
| 24 | InstitutionalFlowBot | 0.14 | -0.09 | -0.49% | -5.56% | +2.2% vs SPY | 0.05 n.s. | 0.37 (SPY) | 65 | 35 | Insufficient history |
| 25 | SynthesizedHyperConvexityBot | 0.06 | -0.01 | -0.27% | -19.24% | -21.2% vs QQQ | -1.47 n.s. | 0.73 (QQQ) | 49 | 160 | Fail |
| 26 | SwingTitaniumBot | 0.05 | 0.10 | 0.44% | -4.36% | +0.7% vs GLD | 0.10 n.s. | 0.06 (GLD) | 2 | 216 | Fail |
| 27 | TelegramSignalsBankBot | 0.00 | 0.24 | 24.12% | -100.00% | — | — | — | 39 | 172 | Fail |
| 28 | option_CatalystCallBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 35 | Insufficient history |
| 29 | option_CreditSpreadBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 35 | Insufficient history |
| 30 | option_CrossVolBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 32 | Insufficient history |
| 31 | option_DispersionBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 32 | Insufficient history |
| 32 | option_EarningsCalendarBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 32 | Insufficient history |
| 33 | option_IndexVolBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 32 | Insufficient history |
| 34 | option_IronCondorBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 35 | Insufficient history |
| 35 | option_MispricingBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 32 | Insufficient history |
| 36 | option_MispricingScanBot | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 32 | Insufficient history |
| 37 | us_f6d63504b8fc4179bf311d74f8dd8231 | 0.00 | 0.00 | 0.00% | 0.00% | +0.0% vs SPY | 0.00 n.s. | 0.00 (SPY) | 0 | 52 | Insufficient history |
| 38 | Benchmark_GLD Benchmark | -0.09 | -0.18 | -4.58% | -25.30% | -16.5% vs SPY | -0.67 n.s. | 0.44 (SPY) | 0 | 216 | — |
| 39 | Benchmark_BTC Benchmark | -0.18 | -0.37 | -13.87% | -37.70% | -19.5% vs SPY | -0.41 n.s. | 0.23 (SPY) | 0 | 216 | — |
| 40 | GptBasedStrategyBTCTabased | -0.30 | -0.98 | -18.49% | -18.86% | -16.5% vs BTC | -0.65 n.s. | 0.40 (BTC) | 4 | 215 | Fail |
| 41 | SqueezeMomentumBot | -0.73 | -1.47 | -11.78% | -7.99% | -6.6% vs GLD | -0.71 n.s. | 0.39 (GLD) | 6 | 151 | Fail |
| 42 | EURUSDTreeBot | -0.96 | -0.87 | -3.98% | -4.59% | -4.1% vs GLD | -1.44 n.s. | 0.31 (GLD) | 106 | 216 | Fail |
| 43 | XAUZenCarryBot | -1.14 | -1.44 | -3.14% | -2.19% | -1.8% vs SPY | -0.56 n.s. | 0.81 (SPY) | 8 | 75 | Insufficient history |
| 44 | TARegimeMultiAssetBot | -1.31 | -4.98 | -5.06% | -1.01% | +1.1% vs BTC | 0.24 n.s. | 0.67 (BTC) | 5 | 39 | Insufficient history |
| 45 | TSMOMTrendBot | -2.16 | -6.77 | -17.10% | -2.53% | -0.4% vs BTC | -0.03 n.s. | 0.70 (BTC) | 38 | 39 | Insufficient history |
| 46 | option_EarningsCrushBot | -2.48 | -8.78 | -49.61% | -5.65% | -36.2% vs QQQ | -1.38 n.s. | 0.62 (QQQ) | 72 | 32 | Insufficient history |
Click any strategy name for its investment thesis, full risk-adjusted metrics, drawdown profile, and historical backtest reference.
The table is the summary. Extended analytics and the full DDQ are in the data room.
Request data-room accessInstitutional engagement & access
We offer direct institutional engagement pathways tailored to allocators, proprietary trading desks, and qualified investors:
Separately Managed Accounts (SMA)
White-glove execution directly within your institutional brokerage (Interactive Brokers, Varengold) under a BaFin-regulated liability umbrella (Haftungsdach) partnership. Client retains full custody and key ownership.
API Integration Licence
Institutional JSON/FIX research feed integration directly into your firm's order routing and risk infrastructure. Includes complete technical documentation and specifications.
KVG Fund Wrapper
Co-allocation and fund structuring via specialized German/EU Service-KVGs (UCITS / Spezial-AIF) for qualified institutional ticket sizes.
Research community
For individual practitioners and quantitative researchers, we provide daily research notes, market regime updates, and technical discussion:
Join the Research Desk Telegram
Daily quantitative research notes, strategy updates, and active research discussion directly from our desk.
Join Telegram Channel →
The Alpha-Verified Quantitative Execution Suite
Deploy quantitative execution models with radical public transparency and zero black-box claims.
- The Forward-Test Transparency Dashboard: Real-time Jensen’s alpha, benchmark correlation, Sortino ratios, and drawdown metrics updated daily from an append-only Postgres database. No hidden losses, no curve-fitted backtests.
- Self-Managed / Research Feed ($0): Daily alpha reports & market regime shifts
- API Integration ($199/mo): For professional investors keeping own custody & keys
- Strategic Partner / SMA ($999/mo): White-glove fully managed allocation & private placement
Our Risk-Free Guarantee
Every Tier 2 and Tier 3 onboarding includes a comprehensive 14-day technical due diligence audit. If our systems and performance data do not meet your rigorous quantitative standards, you receive a 100% refund of onboarding fees.
Limited Availability: Q4 2026 cohort capped at 25 Strategic Partner slots to ensure optimal execution liquidity.
How to invest
Self-Managed / Research Feed
The fastest way to get started. Perfect for retail investors and those who want to execute manually or test the waters with live insights.
- 1 free research note / day
- Community access & educational updates
- Lowest barrier to entry
Or Discuss Needs:
Book OnboardingAPI Integration
For prop shops and advanced HNWIs who maintain their own execution stacks. Access our research feed via REST/FIX API.
- Institutional JSON/FIX API feed
- Full technical documentation
- Direct integration into your trade engine
Strategic Partner / SMA
Keep custody of your assets. Our algorithms execute directly on your preferred institutional brokerage (IBKR, Varengold, etc) via a secure sub-account.
- Professional investors only (§67 WpHG)
- Full transparency of every trade
- Liability umbrella (Haftungsdach) partnership
Within every architecture, choose one strategy or Omni (a balanced portfolio of the entire book).
Strategic fit between risk appetite and deployment model is decided in your free onboarding call.
Diligence collateral
Institutional investors and family offices conducting operational and quantitative due diligence can access our institutional materials:
Due Diligence Questionnaire (DDQ)
Complete answers covering operational setup, risk management, data hygiene, and regulatory reverse-solicitation framing (§67 WpHG).
Read online DDQ →Data Room Access
Institutional access to detailed execution logs, daily NAV marks, and statistical distribution sheets.
Request data room →Institutional Deck & Factsheet
The investor deck and strategy factsheet are available on request to qualified investors.
Request deck & factsheet →Strategy families
Strategies are grouped by the market inefficiency they target. A diversified allocator typically blends families rather than selecting within one.
- Tree-based ML — Gradient-boosted and random-forest models trading FX and metals on engineered features.
- Regime-adaptive — Detects market regimes (trend vs. mean-reversion) and switches strategy per regime.
- Sentiment & news — Trades on news flow, fear-greed indices, and aggregated sentiment indicators.
- Kronos forecasting — Uses the Kronos time-series foundation model to forecast short-horizon returns.
- AI / LLM-driven — LLM-in-the-loop strategies: GPT, DeepSeek, and multi-agent hedge-fund setups.
- Momentum & swing — Classic momentum, Bollinger/Keltner squeeze, and multi-day swing strategies.
- Portfolio optimisation — Sharpe-maximising, recursive-decay, and convex multi-asset allocation.
- Event-driven — Trades earnings tilts and curated Telegram research feeds.
- Benchmarks — Buy-and-hold benchmarks (SPY, QQQ, FTWD, GLD, BTC) spanning equity, commodity and crypto beta.
Methodology
- Paper-traded forward test. Each strategy executes daily within containerized Kubernetes CronJobs using the open-source python_tradingbot_framework. Orders are simulated against live market prices using a $10,000 paper-capital seed so return metrics are directly cross-comparable; no real client capital is deployed on this portal.
- End-of-day marks. Portfolio worth is recorded daily from open positions plus cash and written to an append-only Postgres store.
- Metrics. Sharpe, Sortino, Calmar, Profit Factor, and drawdown computed from the daily return series, annualised at 252 trading days. Correlation and beta are calculated against live SPY, QQQ, and FTWD benchmark feeds.
- Live vs. backtest. Every strategy page separates the live forward-test (authoritative) from historical backtests (reference only). Backtest results are never used to rank models.
Last updated: 2026-10-30 06:32 UTC
For professional investors
Request the investor deck, DDQ, and extended analytics. Firm-gated and reviewed manually.
Request access