A Swiss-based independent Engineering service delivering systematic, model-driven stock selection — free of analyst opinion, emotional bias, and discretionary judgment. Recalibrates every 15 trading days. Validated across 20 years of unseen market history. Live alpha vs S&P 500 since 2025.
Finance Engineering. Not Finance Opinion.
Instead of an analyst telling you what to buy, a mathematical model tells you — one tested across 20 years of real market history and recalibrating every 15 trading days on live data. The majority of traditional models rely on statistics — regression, factor analysis, probabilistic inference. PVOCX does not. Stock features are extracted through a proprietary mathematical framework drawing on topology, geometry, and algebra — structural analysis that eliminates distributional assumptions and captures patterns invisible to conventional statistical methods. News sentiment scores are computed mathematically and fed into the model as one signal among many. No manual overrides, no discretionary judgment. Validated out-of-sample across every major market cycle since 2004.
The 15-day engine.
Three stages. Zero human judgment. Every 15 trading days the cycle repeats — automatically.
Stock features are extracted through a non-statistical mathematical framework drawing on topology, geometry, and algebra. This reveals structural patterns in price behaviour invisible to conventional statistical methods. News sentiment is computed mathematically and included as one input signal among hundreds.
More independent models score every stock from different structural angles. A supervisory ensemble resolves conflicts and amplifies only the selections where multiple models converge — concentrating output on the market's strongest consensus signals.
One definitive ranked list — delivered every 15 days with full transparency. Current picks, entry signals, confidence scores. No forecast. No narrative. Just the output.
Built on rigour,
not opinion.
Most investment services sell you an opinion. PVOCX sells you the output of a system. The idea behind PVOCX is to use pure data, mathematics, and the structural patterns they reveal — with no human judgment in the selection process.
A single model can be wrong. The system uses more independently trained models, each scoring every stock from a different structural angle — topology, geometry, and algebra rather than conventional regression or factor methods. A stock only makes the final selection when multiple models agree. No single model dominates, and no fragile signal gets through alone.
Before each delivery, the system runs through a wide range of possible stock combinations — not just ranking the universe and cutting at a threshold. A supervisory layer then arbitrates between competing paths, discarding fragile selections and keeping only those that hold up across the full simulation space. The output is not a ranking. It is a distillation.
Markets change. A model trained two years ago may be missing what matters today. Every 15 trading days, the system recalibrates on the most recent market data — the window that research shows is optimal for this type of structural modelling. More data introduces noise; less is insufficient. Approximately 17 recalibration cycles per year means the model is always working on the freshest possible signal.
Not all market environments are equal. The system continuously monitors structural shifts — rising volatility, sector dislocations, deteriorating breadth — and adjusts accordingly. News sentiment is recomputed mathematically each cycle and integrated as a quantitative signal. When the structural and sentiment signals align, the model's conviction is at its highest.
No model is immune to external shocks. In early 2025, the system recorded 4 consecutive periods of negative alpha — driven by macro volatility outside any model's predictive reach. This is normal and expected. From period 5 onward the system recovered fully: 9 of the following 13 periods posted positive alpha, closing 2025 with VOCX-10 at +27.8% vs the S&P 500's +13.2%. Resilience is not the absence of drawdown — it is the structural ability to recover.
PVOCX does not promise guaranteed returns. No service honestly can. What the system has consistently demonstrated across 20 years of out-of-sample data is a selection process that outperforms the S&P 500 over time — not in every single period, but in the cumulative result that counts. Every period's results — good and bad — are published in full and publicly verifiable.
Out-of-sample tested. Verified. Transparent.
Every result is based on out-of-sample evaluation — the model never touched this data during training. No curve-fitting. No look-ahead bias. No survivorship bias. Walk-forward validation across every major market cycle since 2004, with live performance publicly verifiable from 2025.
Choose your edge.
Each tier is a strict subset of the one above. VOCX-5 is always the top 5 inside VOCX-10, which is always the top 10 inside VOCX-25 — the same ranked list, zoomed in. Free subscribers see last period's VOCX-25 results in full. The live picks are inside. Subscribing reveals which ones.
- Every 15 days: last period's full VOCX-25 list with performance vs S&P 500.
- VOCX-25 picks — last period evaluation
- 15-day holding period cycle
- Performance chart & YTD stats
- Market regime signal
- Every 15 days: live VOCX-10 picks with entry signals and confidence scores.
- VOCX-10 current picks with full signals
- Last period evaluation & alpha breakdown
- Full confidence scores per stock
- Performance dashboard
- Regime status updates
- Every 15 days: live VOCX-5 highest-conviction picks — the model's strongest signals.
- Everything in Core
- VOCX-5 concentrated picks
- Curated stock news per position
- Regime & sector signals
Get in touch.
For any information, technical issues, subscription enquiries, or any other matter — just send us an email. We read every message and aim to reply within 3 working days.
✉ info@pvocx.com8207 Schaffhausen
Switzerland