Turn independent strategy systems into one portfolio decision process
PDS sits above individual strategy engines. It evaluates providers, separates research credibility from portfolio usefulness, manages integration, forms portfolio-level decisions, and monitors whether each role remains justified. The current Core combines ADAA and Forecast-to-Rank Allocation (F2R) as complementary providers, while PDS handles portfolio integration, adaptive risk control, and monitoring at the system level.
PDS Core, Adaptive Risk Control, and Dynamic FX
PDS separates the source of portfolio opportunity from the controls applied around it. PDS Core is the main multi-strategy portfolio. PDS Adaptive is a bounded hybrid risk-control layer that can reduce total Core exposure when its defensive controller is active. Dynamic FX manages USD/KRW hedge exposure on a separate monthly decision clock.
PDS Core
Combines admitted strategy providers under a defined portfolio policy while preserving each provider's timing, identity, and research record.
PDS Adaptive
PDS Adaptive uses an RL signal inside a bounded hybrid controller with independent point-in-time volatility confirmation. It does not select securities or replace the Core allocation engines.
Dynamic FX Overlay
Applies a monthly USD/KRW hedge decision and cost schedule while keeping the underlying USD portfolio decision separate.
From independent providers to one portfolio decision
Independent strategy systems enter through defined interfaces and keep their own research and operating records.
Provider outputs are translated into a consistent portfolio representation before cross-strategy comparison.
Data timing, research quality, implementation assumptions, and operational readiness are reviewed before portfolio use.
A sound research result does not automatically enter the portfolio; incremental portfolio usefulness is assessed separately.
Approved providers are combined under a defined strategic allocation policy while each provider's identity and contribution remain traceable.
A bounded hybrid Adaptive layer can selectively reduce Core risk without replacing the underlying strategy engines.
Core, Adaptive, Preview, Dynamic FX, and provider states are refreshed and monitored under a common operating clock.
Research quality and portfolio usefulness are evaluated separately
PDS keeps research review separate from portfolio admission. A strategy can be well specified as research yet add little incremental portfolio value, while a useful portfolio role does not change the status of a failed or inconclusive research result. Admission therefore depends on both research quality and the role the strategy adds to the portfolio.
Is the strategy result credible?
Data timing, reproducibility, model specification, robustness, and result quality are assessed on their own terms.
Does it add a useful portfolio role?
Incremental diversification, role fit, operational burden, and interaction with existing providers determine whether the strategy belongs in the portfolio.
Complementary strategy systems inside the PDS portfolio process
ADAA and F2R are the strategy systems currently admitted to the Active Core. Each remains an independent Portfolio Strategy System with its own models, research record, and operating history; PDS is responsible for portfolio-level admission, integration, the current combined decision, and monitoring.
ADAA
A multi-asset strategy built around Decision Diversification, combining complementary decision horizons, opportunity views, defensive responses, and persistence.
View ADAAForecast-to-Rank Allocation (F2R)
A cross-asset forecasting system combining conventional supervised machine learning with Chronos-2 pretrained time-series forecasting before rank-based portfolio formation.
View F2RLive Completed Performance
Updated with each canonical PDS release. The cumulative chart compares PDS Core + Dynamic FX with PDS Core over their common history beginning in May 2017. The longer current-definition Core / Dynamic FX reconstruction begins in December 2005 and remains available in the full dashboard. The second chart shows the latest 12 completed holding months for Core + Dynamic FX and Adaptive + Dynamic FX. Adaptive begins in May 2021. Current MTD is excluded throughout.
Cumulative Performance
Common comparable history from May 2017 · completed through 2026-09-30 · current MTD excluded
Recent Completed Monthly Returns
Latest 12 completed holding months · 2025-10 → 2026-09 · current MTD excluded
| Portfolio | Completed period | Cumulative | CAGR | Vol. | Sharpe | MDD | Calmar |
|---|---|---|---|---|---|---|---|
| PDS Core + Dynamic FX | 2017-05-01 → 2026-09-30 | 261.0% | 14.6% | 11.2% | 1.28 | -12.3% | 1.18 |
| PDS Core | 2017-05-01 → 2026-09-30 | 208.7% | 12.7% | 11.0% | 1.15 | -13.6% | 0.94 |
| PDS Adaptive + Dynamic FX | 2021-05-03 → 2026-09-30 | 123.4% | 16.0% | 11.7% | 1.33 | -11.0% | 1.45 |
| PDS Adaptive | 2021-05-03 → 2026-09-30 | 91.8% | 12.8% | 11.7% | 1.09 | -12.0% | 1.07 |
One operating clock, with private infrastructure kept out of the public view
The public dashboard uses the same decision state, clocks, and portfolio data as the operating system. It omits machine-specific infrastructure and private controls rather than maintaining a separate investment view. Current decision state, Adaptive state, Preview, portfolio weights, Dynamic FX monitoring, and performance follow the same data and clock conventions.
Public operating views, private implementation
Public
- System architecture, provider relationships, and governance
- Current PDS Core, Adaptive, Preview, Portfolio, Performance, and Dynamic FX operating views
- Current decision clock and detailed portfolio targets through the PDS dashboard
- Historical performance views with support periods and evidence boundaries stated explicitly
Internal implementation
- Local filesystem paths, runtime/cache details, repository controls, credentials, and debug tooling
- Low-level controller training settings, seeds, and implementation-specific internal identifiers
- Brokerage-account holdings, realized account P&L, and order-routing infrastructure
- Unpromoted research branches and internal experimental work
Methods used in this system
These companion articles explain methods used directly in the system or needed to interpret its validation and outputs.
Portfolio Backtesting and Rebalancing
Defines portfolio accounting and rebalance timing used to maintain a continuous PDS Core performance path across provider decisions.
Drawdown, Underwater Duration, and Recovery
Defines drawdown, underwater duration, and recovery measures used to interpret path-dependent PDS Core downside behavior.
Turnover, Transaction Costs, and Net Performance
Defines turnover and transaction-cost conventions used when implementation effects are evaluated across portfolio decisions.
Strategy Similarity and Decision Diversification
Provides strategy-similarity and decision-diversification concepts used when PDS evaluates whether a provider adds useful diversification before admission.
Benchmark-Relative Portfolio Evaluation
Defines benchmark-relative portfolio measures used to separate absolute PDS Core performance from relative performance evidence.
Information Timing, Data Vintages, and Real-Time Macro Data
Defines information-timing and data-availability discipline for provider inputs that depend on macroeconomic information.
Reinforcement Learning for Adaptive Portfolio Risk Control
Explains RL-assisted adaptive risk control as a bounded defense layer, including state/action/reward design, offline evaluation limits, policy freezing, and Core-versus-Adaptive attribution.
Dynamic Currency Hedging and FX Overlay Accounting
Explains Dynamic FX overlay accounting, hedge-ratio timing, raw-versus-admissible FX observations, cost treatment, and base-currency attribution.