-

Frontlines and Frontiers of Financial Ledgers
Read more: Frontlines and Frontiers of Financial LedgersAs high-frequency decentralized finance and institutional-grade digital settlement systems mature, next-generation ledger architectures face an enduring engineering tension: achieving extreme transaction throughput without sacrificing cryptographic…
-

Combinatorial Purged Cross-Validation in Non-Stationary Crypto Markets
Read more: Combinatorial Purged Cross-Validation in Non-Stationary Crypto MarketsContinuous financial time series violate this baseline assumption through serial correlation, overlapping trade horizons, and clustering volatility. To deploy institutional capital safely, quantitative developers must…
-

Podcast Episode: Machine Learning For Market Makers
Read more: Podcast Episode: Machine Learning For Market MakersPip: Welcome to the Susaanoo Systems Podcast, where we ask the challenging question: what does a trading platform do when it has nothing useful to…
-

Explainability of the Ensembles
Read more: Explainability of the EnsemblesMost people picture a forecasting system as a single model: data goes in, a number comes out, someone acts on it. That mental image is…
Writing Pillars
Know the market you are actually in
Markets do not behave the same way in every state, so a model that responds identically to every cycle will overreact in one regime and underperform in the next.
Susaanoo classifies the live market state—bull, bear, high-volatility, low-volatility accumulation, sideways, BTC-led, or alt-led—and estimates the risk of transition between them.
Seven specialist models sit behind that classification, so the platform changes its own behavior when the market changes state, before a single recommendation reaches you.
A view, a number, and a reason.
Every asset in the coverage universe is scored across four horizons—12-hour, 1-day, 7-day, and 30-day—and returned as BUY, WATCH, or AVOID. Each call carries a directional probability, an expected return, a confidence level, and the regime it was formed in. Nothing arrives as a bare score.
The default state is cash
A forecast is not a position. Before any signal becomes actionable it must clear four independent gates: directional edge above threshold, expected return that survives costs, sufficient model confidence, and capital guardrails that hold in stressed conditions. Signals that fail are returned as WATCH or AVOID with the failure recorded. Most cycles deploy a small fraction of available capital—by design.
From conviction to position size.
A signal only becomes useful when it answers “how much.” The overlay converts each qualified call into a concrete exposure: volatility-fitted sizing, a stop-loss calibrated to the instrument rather than a fixed percentage, and an allocation stance for the book as a whole. Position size shrinks automatically as model uncertainty rises, and bear regimes cut total exposure before any individual trade is considered.
Evidence you can put in front of a risk committee.
Backtests that look impressive are easy to produce and easy to ignore. Susaanoo tests signals using walk-forward analysis with leakage safeguards, combinatorial purged cross-validation, Monte Carlo and multi-asset simulation, k-fold testing, and regime-switching scenarios. The output is a validation package — not a headline number.
Every decision leaves a record.
Institutional adoption depends on explaining decisions after the fact, including the ones that did not work. Susaanoo produces model cards, data lineage, threshold configuration logs, signal audit records with reason codes, and client-ready export packs. Risk, compliance, and investment committees can reconstruct why the platform said BUY, WATCH, or AVOID on any asset, on any date, at any horizon.
