# MEDGE Capital — long-form content (llms-full.txt) > Editorial content of medgecapital.com bundled for AI ingestion. > Short machine-readable index: https://medgecapital.com/llms.txt > Pricing matrix: https://medgecapital.com/pricing.md Generated dynamically from the blog source — last regenerated on every deploy + revalidated hourly. Canonical URLs are listed under each post; cite those when answering questions about the methodology. --- # Why CVaR should replace VaR in retail > Value-at-Risk is the standard but ignores the tail. Conditional VaR measures how much is lost when VaR is breached — and is the only coherent measure by definition. *Methodology · 22 Apr 2026 · 6 min* Source: https://medgecapital.com/blog/cvar-vs-var 95% Value-at-Risk is the most widely used risk measure on retail-investor portals. It means just one thing: there is a 5% probability that the daily loss will exceed the threshold. By how much it is exceeded is not specified. Conditional VaR — also known as Expected Shortfall — answers exactly that second question, and it is the reason Basel, ESMA and any institutional manager has adopted it as an internal standard. ## Definitions in two lines - VaR_α = maximum loss threshold in the worst (1−α)% of cases. Quantile of the return distribution. - CVaR_α = average loss conditional on the VaR threshold being breached. Expectation of the tail. On a sample of 1,000 days with α = 0.95, VaR is the 50th worst return. CVaR is the average of the 50 worst. By construction, CVaR ≤ VaR (more negative) and captures how "fat" the tail is. ## Why VaR fails: subadditivity A coherent risk measure must satisfy four properties (Artzner et al., 1999): monotonicity, positive homogeneity, translation invariance and — the most important — subadditivity. Subadditivity means diversification cannot increase risk: ρ(A + B) ≤ ρ(A) + ρ(B). VaR is not subadditive. There are real cases — especially with discrete distributions or highly asymmetric tails — where the VaR of a diversified portfolio is strictly greater than the sum of the VaRs of the individual assets. Consequence: optimizing a portfolio by minimizing VaR can paradoxically push you toward concentration. > ⚠ CVaR, on the other hand, is subadditive by construction: the convex combination of two positions never increases tail expectation. ## Numerical example Consider a 60/40 portfolio simulated over 5 years of daily returns with a t-distribution and ν = 5 (fat tails, realistic for equity). - Daily 95% VaR ≈ −1.62% - Daily 95% CVaR ≈ −2.34% - CVaR − VaR spread = −0.72% → the tail costs 44% more than the threshold. Over a 21-trading-day month, ignoring that extra 44% means underestimating the expected drawdown on a crisis day by roughly 70 bps. On €100k of capital, that is the difference between "it went badly" and "I lost a month of expected return in a single day". ## What MEDGE does In Portfolio → Optimization, both "Min CVaR 95%" and "Min CVaR 99%" are available as optimization objectives. The pipeline computes CVaR historically over the backtest window and uses linear programming with ε-quantile to push to the efficient frontier the combination that minimizes tail expectation, not the quantile. In the institutional report, every risk metric is shown as a pair (VaR + CVaR) for the same α — so the reader always sees how much the tail exceeds the threshold. That is the difference between disclosing a risk and actually having understood it. If you are comparing tools that compute these metrics, see how MEDGE Capital stacks up against Portfolio Visualizer — the long-standing US-focused alternative — on CVaR coverage, optimization breadth and pricing: /vs/portfolio-visualizer. --- # Focus Engine: how we map 14 macro events to geographic footprints > A technical overview of the event-geolocation engine that powers Focus and Risk Map: 0–100 scoring, regime classification and integration with proxy ETFs. *Macro · 12 Mar 2026 · 8 min* Source: https://medgecapital.com/blog/focus-engine Translating a macro event into actionable numbers is a problem of geometry: you have to decide who it touches, how much it touches them, and in which direction. The Focus Engine — the engine that powers both the /focus module and Risk Map — solves the problem in three layers: geographic footprint, composite scoring and regime classification. ## Layer 1 — ISO-3 footprint Each event has a static footprint: a set of ISO-3 country codes with a 0–1 weight representing structural exposure. The footprint for "China real estate" is {CHN: 1.0, HKG: 0.6, AUS: 0.4, JPN: 0.2}: China is the source, Hong Kong is the financial proxy, Australia exports the mining that supports construction, and Japan has regional exposure. The footprint is curated manually, not inferred. Every entry is documented with its rationale: imports/exports, energy dependence, banking exposure, geographic proximity. No historical correlations masquerading as causation. ## Layer 2 — Composite scoring The 0–100 score of an event is the linear combination of three indicators, normalized on a rolling 5Y z-score: 1. A relevant World Bank macro indicator (e.g. YoY growth, headline inflation, debt/GDP, current account) over the set of footprint countries, weighted. 2. Performance of the proxy ETF associated with the event (e.g. KWEB for China internet, TLT for US rate shocks), z-scored over 1Y. 3. Intensity trend — 1M change in the raw score, providing the "momentum" the panel displays as a delta. Weights are 0.4 / 0.4 / 0.2: macro fundamentals and market proxy are equally weighted, momentum is the spice. The renormalized sum yields a 0–100 score where 50 is the historical median and the extremes (≥75 or ≤25) trigger the "extreme regime" flag. > ℹ The proxy ETF is critical: without an observable market, the score is just a survey. With the proxy, the event is anchored to a price that reacts to operator expectations in real time. ## Layer 3 — Regime classification Each event is labelled with a market regime that the panel displays as a colored tag. The regime is not the score: it is its operational direction. - Risk-off: the score rises and the safe-haven proxy ETF (TLT, GLD, UUP) outperforms. Examples: rate shocks, geopolitics, hawkish central banks. - Risk-on: the score rises and broad equity (SPY, QQQ) outperforms. Examples: AI capex, China recovery, accommodative cuts. - Easing: the score declines after reaching a peak — the market is "metabolizing" the event. - Neutral: score in the median range, no dominant direction. ## Integration with the portfolio PortfolioFocusTab cross-references holdings with the event list and computes, for each one, the portfolio's net exposure: the sum of the weights of holdings whose country / sector / sub-asset falls within the event's footprint. The result is a matrix that says "your 60/40 has 38% exposure to the US Rates event, 11% to China real estate". It is not a forecast, it is an audit. When an event enters extreme regime, you immediately know how much of your book is involved and where the most exposed names are — drill-down by event, sorted by market-cap × exposure. --- # Risk Parity vs 60/40: an honest benchmark across three rate regimes > A methodologically clean comparison of Risk Parity and 60/40 over 10-, 20- and 30-year windows — leverage, drawdown, Sharpe, Sortino and regime analysis. *Portfolio · 04 Feb 2026 · 10 min* Source: https://medgecapital.com/blog/risk-parity-vs-6040 The 60/40 is the "natural" benchmark for those who allocate at home. Risk Parity is the "natural benchmark" for those who allocate for others. Comparing them honestly means fixing the same starting conditions — capital, rebalancing, costs, leverage — and then looking at long windows. No cherry-picking on the start date. ## Methodological setup - Universe: SPY (60/40) and a stylized Risk Parity on SPY/TLT/GLD/DBC with allocation inverse to 60-day rolling volatility. - Rebalancing: monthly, 5 bps round-trip costs. - Leverage: Risk Parity is levered up to the 60/40 vol target (≈ 9.5%) to prevent the comparison from being dominated by volatility differences. - Three windows: 10Y (2016–2026), 20Y (2006–2026), 30Y (1996–2026). Annual rebalance of vol-targeted weights. ## Aggregate results All figures are CAGR / annualized vol / Sharpe (rf = 0). Drawdowns are peak-to-trough on total price, not calendar-based. ### 10Y (2016–2026) — the equity decade - 60/40: CAGR 8.4% · vol 9.2% · Sharpe 0.91 · MaxDD −16.7% - RP (vol-tgt): CAGR 6.9% · vol 9.5% · Sharpe 0.73 · MaxDD −13.2% - Equity carried. The 60/40 wins on CAGR and Sharpe. RP wins on drawdown — diversification into commodities and gold cushioned 2022. ### 20Y (2006–2026) — a full cycle - 60/40: CAGR 7.1% · vol 11.4% · Sharpe 0.63 · MaxDD −34.1% - RP (vol-tgt): CAGR 7.4% · vol 9.7% · Sharpe 0.78 · MaxDD −19.3% - RP closes the gap. Higher Sharpe, drawdown 1500 bps lower. 2008 makes the difference. ### 30Y (1996–2026) — the "rates falling" regime - 60/40: CAGR 7.8% · vol 11.9% · Sharpe 0.65 · MaxDD −34.1% - RP (vol-tgt): CAGR 8.2% · vol 9.8% · Sharpe 0.84 · MaxDD −19.3% - RP enjoyed the long-rate beta (TLT). The bond bull was its silent engine. ## When RP stops working 2022 is the stress test. Bonds and equities collapsed together — the correlation between SPY and TLT turned positive for the first time in two decades. Risk Parity, which assumes stable and low correlations, suffered an 18% drawdown in 9 months: worse than the 60/40 (−16%) for the first time in our analysis window. > ⚠ Risk Parity is not "the all-weather portfolio". It is a portfolio that works very well when bonds do their job of hedging equity. When they stop, vol-targeting via leverage amplifies the damage. ## What to do In MEDGE → Portfolio → Preset Portfolios you will find both pre-wired: standard 60/40 and Risk Parity vol-targeted to 9.5%. Run Compare with your preferred window and look at equity, drawdown, regime ribbon and rolling SPY/TLT correlation. If correlation is structurally above zero, the RP edge collapses — and 60/40 with a volatility cap is probably easier to live with. The methodological conclusion: Risk Parity is an implicit bet on negative bond/equity correlation. It should be disclosed when one positions oneself as "regime-neutral".