Algorithmic trading · audit-honest · live demo
Quantitative trading, honestly audited.
Algorithms held to the standard of published science. Every hypothesis pre-registered. Every result audited, deflated, and made public — including the failures.
The Vrelum research log
A thousand tests.998 rejected.Two are live.
Every test pre-registered before it ran. Every verdict — wins and failures alike — published the moment it landed.
Top active strategies
Balanced · 1×GC Daily Breakout
Event-driven specialist
Silver Breakout v2
The graveyard
1,000+ tests. 2 shipped.
Across 17 ML paradigms, 50+ named sprints, and roughly a thousand hyperparameter configs since 2024, two strategies cleared every audit gate (R86b · R99b). The rest live here — categories, then 21 recent rejections.
By research direction
Each row is a paradigm we systematically tested — full post-mortems live in our internal experiments log.
ML overlay (meta-learning)
Add a model on top of existing strategies — pick winners, scale by confidence, skip losers. All five attempts hurt Sharpe via diversification loss.
Direct ML direction prediction
Predict next-bar direction with LightGBM / XGBoost on price + cross-asset features. AUC ceiling 0.50-0.51 — no extractable signal at hourly resolution.
Position sizing & filter schemes
Adjust position size or skip trades by volatility, regime, or time-of-day. Vol-target had 60 sub-configs; one passed but overfit a single hyperparameter.
Statistical arbitrage
Trade the spread between cointegrated assets reverting to long-run mean. XAU/XAG not cointegrated (p=0.26), half-life 142 days — untradeable.
Alternative data
Non-price signals (news, sentiment, social) for direction. GDELT raw + smart variants: no |IC| > 0.05 stable in both IS and OOS.
Volatility prediction
Predict next-period vol and scale exposure. Vol IS predictable (Spearman 0.17) — but unexploitable on retail XAU spot without options access.
HMM regime classifier
3-to-5-state Gaussian Hidden Markov Models to detect trend/chop/crisis. Train +3.8 → OOS +0.18 collapse. Overfits on 75 train trades.
COT sentiment filter
CFTC Commitment-of-Traders speculator positioning as entry gate. Pre-2025 pattern broke post-2025 — random null P=49.6%.
Sprint D LightGBM grids
Direction prediction on raw bars across 35 hyperparameter configs. Ceiling -0.5 Sharpe. Inversely-predictive in some folds.
Liquidity-grab / microstructure
Trade liquidity hunts at swing highs/lows. Sharpe -0.97 at hourly resolution. Needs Level-2 / tick data — retail-impossible.
Event-window search
Brute-force test every economic event × horizon combination. ~1000 tests with strict Bonferroni; only 42-Day Bill Auction survived (t=4.74).
21 recent sprint-level rejections
Detailed rejections from the 2026 sprint cycles, with the audit reason that killed each one.
Killed sprints stay published forever — both as a discipline forcing function and a public record of what we tried.
Live operations
We trade in public.
Real account · real money · real losses. Every position published.
Trust takes time.Start small.
Open a free account. Read every audit. Watch the bots trade in real time. Deploy capital only when you're convinced — not before.
For builders
Submit a strategy. Pass the same R83 audit our internal bots clear. List on the marketplace and earn revenue share when capital flows through.
Join builders waitlistFor institutions
Custom allocation, whole-fleet deployment, or audit-as-a-service for your own quant team. Direct line to the founding engineer.
Talk to usFor press
Logos, the R83 methodology paper, founder bios, audit framework whitepaper, and high-res brand assets — everything you'd need to write about us.
Download press kitCommon questions
Frequently asked questions
Things people ask about audit methodology, capital safety, risk caps, and the marketplace. Don't see yours here? Email support@vrelum.com.
Because we deflate. The R83 audit framework cuts in-sample Sharpe down by Bonferroni-corrected statistical bounds — typically removing 15–30% of headline performance. Competitors usually show in-sample backtests; we publish what your live account will actually see, post-deflation. The math is conservative on purpose.
No. Implementation stays internal. We publish methodology, audit results, the sprint registry (including all 998 rejections), and every live trade outcome. We don't publish exact entry/exit thresholds, signal logic, or parameter values — that protects strategy capacity and prevents reverse-engineering from competitors.
We've stress-tested this. The audit assumes a worst-case correlated drawdown of -28% per year. Each bot pauses at -7% drawdown (warn) or -10% (halt). A portfolio-wide circuit breaker triggers at -15%. Your capital stops trading well before total loss — the fleet self-limits.
It's the opposite. Anyone in quant knows most strategies fail out-of-sample — a 100% win rate signals overfitting or lying. Our 21:2 reject-to-ship ratio is the math working correctly. Publishing the graveyard lets you audit our audit. If you can't see what we rejected and why, you can't trust what we shipped.
Yes, but trade-only permission. ❌ Withdraw ❌ Transfer ✅ Trade ✅ Read. We never custody funds. If Vrelum disappears tomorrow, your account stays at your broker, untouched. We can place orders on your behalf — we cannot move your money.
$1,000 minimum (below this the broker's minimum lot sizes break position sizing). $5,000–$10,000 recommended for the full 6-bot allocation to work as designed. Larger accounts get proportional sizing — risk-per-trade percentage stays the same.
We're in 6-month walk-forward validation. The system needs to prove its live behavior matches deflated estimates before we recommend real capital. Real-account launch is gated on the audit confirming live performance matches forward expectations within tolerance — not on a calendar date.
BYOM = "Bring Your Own Model." Quant developers submit a strategy, pass the same R83 audit our internal bots clear, get listed on the marketplace, and earn revenue share on capital their bot manages. Phase 2 is invite-only beta. Public submissions open later, dependent on audit-pipeline automation.