Portfolio Management Appendix

How SurgeFlow maps portfolio-management theory into benchmark choice, active return, attribution, risk budgeting, simulation, and implementation discipline.

Category: Portfolio

Scope of this appendix

This appendix governs how the Portfolio, Backtesting, Factor, and AI factor-agent surfaces should be consumed. Backtesting is the evidence ledger, Portfolio is the display-only factor-sleeve risk surface, Factors is the construction and publication-gate source, and the AI agent must cite those surfaces before turning a factor view into a portfolio comment. - Backtesting: active return, tracking error, IR, attribution, resampling, and data-quality labels - Portfolio: equal-weight factor-sleeve risk, survivor factors, correlations, VaR/ES, and holdings drill-down - Factors: construction gates, publication state, factor direction, and point-in-time controls - AI factor agent: explanation layer only, never an unsupported recommendation engine

Benchmark and active-return contract

Portfolio-management evidence starts with a benchmark. A usable benchmark should represent the opportunity set, be investable or cheaply replicable, publish verifiable weights before the measurement window, and produce timely return data after the window. Active return only has meaning after that benchmark is named: active return = portfolio return - benchmark return. The position-level version is active weight = portfolio weight - benchmark weight, so value added comes from overweighting assets that outperform the benchmark or underweighting assets that lag it. - Market-cap-weighted index: default market comparator when the sleeve is broad and benchmark-like - Equal-weight universe: useful legacy diagnostic, but not always a realistic investable benchmark - Factor-neutral benchmark: preferred when asking whether a sleeve adds stock-selection value after common factor exposure - Unbenchmarked return: display-only performance, not active-management evidence

Active management ledger

The Backtesting tab should be read as an active-management ledger. Sharpe ratio measures excess return per unit of total risk. Information ratio measures active return per unit of active risk, where active risk is tracking error. Active share measures how far holdings differ from the benchmark. Batting average and hit/miss diagnostics show frequency, but they do not replace magnitude-aware IR. - Tracking error = standard deviation of portfolio return minus benchmark return - Information ratio = mean active return / tracking error, with horizon and annualization stated - Alpha and beta are regression diagnostics; alpha should not be treated as pure skill unless benchmark and factor controls are credible - A high hit rate can still lose money if losing periods are large; read it beside active return, drawdown, and IR

Fundamental Law interpretation

The active-management ledger uses the Fundamental Law as a discipline check: expected IR is driven by forecast skill, effective breadth, and transfer coefficient. Information coefficient is the relationship between forecast signal and realized active return. Breadth is the number of genuinely independent bets, not merely the number of tickers. Transfer coefficient measures how efficiently forecasts become actual weights after constraints, liquidity, cost, turnover, and risk gates. - Practical lens: FL_IR = IC x sqrt(effective breadth) x transfer coefficient - High correlation between bets reduces effective breadth even when the ticker count is high - Long-only, turnover, liquidity, and position limits reduce transfer coefficient - Aggressive active-risk targets can lower realized IR when constraints bind harder

Factor models in portfolio management

Multifactor models translate raw holdings into interpretable exposures. In SurgeFlow, factor models support three jobs: return attribution, risk attribution, and portfolio construction. A sleeve may look strong because of market beta, style exposure, security selection, or a residual component. Portfolio users should read these separately rather than treating total return as one undifferentiated score. - Return attribution: separate benchmark return, factor tilt, security selection, cost drag, and residual - Risk attribution: estimate which factors and correlations drive volatility, VaR, ES, and drawdown risk - Construction: use factor exposures, covariance, VIF, and gate state to avoid redundant bets - Residual: large unexplained residual is a warning that the model may be missing sector, country, liquidity, or event risk

Risk budgeting and downside diagnostics

Risk management is a system of limits, not a single number. VaR estimates a loss threshold at a stated confidence level. Expected shortfall or CVaR estimates average loss beyond that threshold. Drawdown captures path damage. Scenario and sensitivity analysis ask how the portfolio behaves under specified shocks, including events not well represented in recent history. - VaR: useful for a probability-linked loss threshold, but sensitive to history, volatility, and correlation assumptions - ES/CVaR: more informative for tail severity because it looks beyond the VaR cutoff - Maximum drawdown: path-dependent damage that daily volatility can hide - Scenario and sensitivity tests: final stress screens beside position limits, exposure limits, stop-loss limits, and active-risk limits

Backtesting and simulation discipline

Backtests are hypotheses under historical conditions. A credible process separates strategy design, simulation, and output review. It must label survivorship bias, look-ahead bias, reporting lag, data snooping, and p-hacking risk. Rolling-window backtesting preserves chronology; historical simulation resamples observed returns; Monte Carlo simulation draws from an assumed joint distribution and therefore depends heavily on distribution choice and covariance calibration. - Point-in-time data protects against look-ahead and survivorship contamination - Multiple tested variants require stronger evidence thresholds and out-of-sample confirmation - Bootstrap and historical simulation help test sampling luck, but remain tied to observed history - Monte Carlo adds flexibility, but normal assumptions can understate skew, fat tails, and tail dependence

Implementation and vehicle-aware costs

A portfolio edge is only consumable after implementation costs. For single-name sleeves this means turnover, spread, liquidity, market impact, borrow or short-sale limits if applicable, and tax assumptions. For ETF or fund vehicles it also means expense ratio, tracking error, premium/discount behavior, creation/redemption mechanics, and total cost of ownership. SurgeFlow should label which costs are modeled, which are placeholders, and which are outside the current surface. - Gross return: before modeled trading costs - Net return: after the stated cost model only, not after every real-world friction - Cost drag: must be read beside turnover and liquidity - ETF/fund translation: tracking error and vehicle costs can erase a paper factor edge

How to consume the live tools

Use the Portfolio tab to inspect factor sleeves, survivor gates, risk readiness, top holdings, factor correlation, and distribution diagnostics. Use Backtesting to ask whether a sleeve added active value against the correct comparator after cost and sample-quality labels. Use the Factor tab to confirm whether the underlying factor was publishable and current. Use the AI factor agent only after those source surfaces agree. - Do not promote a display-only Portfolio sleeve into a trade without Backtesting and factor-gate support - Prefer benchmark-relative metrics when the question is manager skill or active value - Prefer downside diagnostics when the question is capital preservation - Treat missing, stale, or draft-only data as a blocker for confident conclusions

Appendix

Operating summary for portfolio-management use. - Benchmark first: active return only exists relative to a named, investable, timely benchmark. - Active management: read IR, tracking error, active share, IC, breadth, transfer coefficient, and sample size together. - Attribution: separate factor tilt, security selection, risk weighting, cost drag, and residual. - Risk budget: VaR, ES/CVaR, drawdown, scenario, sensitivity, position limits, and stop-loss rules are complementary tools. - Backtesting: distinguish rolling backtests, bootstrap/historical simulation, and Monte Carlo; label survivorship, look-ahead, reporting lag, and data-snooping risk. - Implementation: turnover, liquidity, spreads, market impact, tracking error, fees, and taxes decide whether paper alpha can be consumed.