Term StructureVIX/VIX3M ratio (vol curve; backwardation = panic)81
VIX/VIX3M ratio: 0.83
Methodology: Equal-weighted composite of 7 normalised signals (0–100). Each component uses rolling historical percentile vs full available history. · 0–24 = Extreme Fear · 25–44 = Fear · 45–55 = Neutral · 56–74 = Greed · 75–100 = Extreme Greed · Updated daily
REGIME ALLOCATION BACKTEST — 5 YEAR
BK ALLOCATION
SPY B&H
60/40
Total Return
+54.9%
+93.3%
+44.7%
CAGR
+9.1%
+14.1%
+7.7%
Sharpe
0.58
0.76
0.34
Max Drawdown
-8.2%
-18.8%
-12.7%
Regime-aware monthly rebalancing · rf=4.5% · Full detail in Edge tab · Past performance not indicative of future results
TODAY'S HEADLINES
Selection by rule · Headlines verbatim from Yahoo Finance · No commentary by BKIQ
▶LISTED ALTERNATIVES(7)Avg score: 58 | Highest: Commodities Index (80)Click to expand
Commodities Index
BCI
80
CRISIS
Drawdown
+71.3
+12.0
+8.8
+0.0
+3.4
+21.6
Diversified Commodity
PDBC
73
CRISIS
Drawdown
+56.1
+21.0
-0.5
+0.0
-12.1
+16.9
Listed Private Equity
PSP
65
STRESSED
Drawdown
+37.4
+3.9
+18.4
+32.9
-34.6
+5.1
Energy Infrastructure
AMLP
60
STRESSED
Tail Risk
-7.3
+19.0
+36.2
+0.0
-18.1
+14.0
US Infrastructure
IFRA
49
MODERATE
Drawdown
+16.8
+3.0
+10.3
+0.0
-27.5
-3.8
Global REITs
REET
42
MODERATE
Tail Risk
-11.0
+4.9
+13.5
+0.0
-46.1
+4.1
High Yield REIT
KBWY
38
MODERATE
Trend
-10.9
-4.5
-9.1
+0.0
-26.0
-10.2
What fragility measures: Fragility tracks price stress characteristics — not the safety or quality of an instrument. A safe-haven asset can show elevated fragility when its price is declining. T-Bills (BIL): Fragility reflects rate sensitivity and reinvestment risk, not credit or liquidity risk. FX pairs: Excluded from this tab — currency pair pillars (Contagion, Trend, Vol Stress) are not meaningful for FX. FX instruments appear on Performance, Risk, and Analysis tabs.
BK Fragility Framework · Drawdown 22% + CVaR 20% + Contagion 18% + Volatility 15% + Trend 15% + Vol Stress 10% · CRISIS ≥70 · STRESSED 55–69 · MODERATE <55 Pillar scores are standardised z-scores relative to history (positive = above average stress) · Top Driver = highest contributing pillar · Negative scores = below historical stress average (healthy signal)
CORRELATION WINDOW
60D
Rolling daily returns
INSTRUMENTS
20
Key representatives
AVG CORRELATION
0.12
Ex-diagonal (high = contagion risk)
CROSS-ASSET CORRELATION MATRIX — 60D
-1.0+1.060-day rolling correlation · as of 10 Jul 2026
STRONGEST CORRELATIONS
QQQ vs Tech
+0.96
SPY vs ACWI
+0.96
QQQ vs ACWI
+0.93
SPY vs QQQ
+0.92
ACWI vs EEM
+0.91
TLT vs LQD
+0.91
QQQ vs EEM
+0.91
EEM vs Tech
+0.90
HOW TO READ
■ Red = move together (+1.0) ■ White = no relationship (0.0) ■ Blue = move opposite (−1.0)
High average correlation = contagion risk Diversification works when colours are mixed 60-day window captures current market regime
Correlation = 60-day rolling Pearson correlation of daily returns · Key 20 instruments selected as representatives of each asset class
Advanced Factor Analysis — Principal Component Analysis (PCA)
PCA reveals hidden risk factors driving cross-asset moves — identifying which latent factors explain the majority of portfolio variance. Replacing RSR in Q2 2026.
Tier 1 — State Machine (headline): Deterministic classifier on ACWI. Stressed: vol ≥ 70th pct OR dd ≤ 30th pct. Crisis: vol ≥ 90th pct OR dd ≤ 10th pct. Ex-ante expanding quantiles, shifted t−1. Auditable, no ML dependencies. Tier 2 — Hidden Markov Model (conviction): 3-state Gaussian HMM on [returns, vol, dd]. Walk-forward retrain every 21 days. Posterior probabilities = conviction gauge. High entropy = regime uncertainty. Transition risk fires when HMM diverges from SM (5–15 day lead time). Tier 3 — Gaussian Mixture (cross-validation): 3-component GMM on [returns, vol, dd]. Walk-forward retrain every 21 days. Captures non-linear clusters that HMM may miss. Consensus: Union-of-risk — most severe model call wins. Conservative by design: false positives preferred over missed crises. Model Agreement: Count of models on same call. 3/3 = full model consensus. 1/3 = models diverge.
REGIME-WEIGHTED FRAMEWORK OUTPUT — ILLUSTRATIVE
Worked example of how the Risk Appetite Score translates regime state into asset-class weights
RISK APPETITE SCORE
55
→ MILD RISK-ON
⚠ ILLUSTRATIVE METHODOLOGY OUTPUT — NOT A PORTFOLIO RECOMMENDATION
The weights and instruments below illustrate how the Risk Appetite Score formula translates a regime state into bucket weights and representative instruments. This is a worked example of the framework's output, not a recommended portfolio, not a backtested strategy, and not a set of positions the reader should hold.
Specific instruments named are the highest-scoring instrument in each bucket on the composite metric as of today. They are not recommendations. Weights are a deterministic function of the regime state; they do not account for transaction costs, liquidity, correlation, or individual circumstances.
No reader should interpret this card as investment advice.
EQ Growth
35%
EWS(Score 78)
EQ Defensive
15%
XLV(Score 79)
Fixed Income
20%
HYD(Score 71)
Real Assets
15%
KBWY(Score 84)
Cash
10%
BIL(Score 57)
Alts
5%
ITA(Score 66)
RAS = Regime(35%) + Fragility Inv(30%) + Fear & Greed(20%) + Vol Inv(15%) · Highest composite score per bucket = highest BK Composite Score
This methodology example is one of several research directions documented on the Research tab. It is not the output of a live or recommended strategy. BKIQ is a personal research project — see About tab for full disclosure.
CURRENT REGIME CONTEXT
Factual state summary — no model interpretation
CALM REGIME
Framework state: fragility score 66.5, volatility at the 50th percentile, drawdown 1.2% from peak. Signal distribution: 23 RED, 19 AMBER, 71 GREEN. Rising-volatility instruments: 30 of 113. Cross-asset correlation (30-day): 0.30. The regime classification weights the fragility and volatility components more heavily than the drawdown component under the regime-conditional weighting rules documented on the Regime tab.
HIGHEST MONTH-TO-DATE PERFORMER
Solana
+12.0%
Rolling 1M Return (21 trading days)
HIGHEST FRAGILITY SCORE
Treasuries 0-3Y
94/100
BK Fragility Score · CRISIS
INSTRUMENTS WITH RISING VOL
30
of 113 showing elevated vol vs 1M ago
Elevated = current 20D vol > vol 1M ago
REGIME ALLOCATION BACKTEST — 5 YEAR (SIMPLIFIED 6-INSTRUMENT MODEL)
ⓘ Note: BK cumulative return is lower by design — the strategy trades raw upside for drawdown protection. Risk-adjusted performance (Sharpe 0.58 vs 0.76) and maximum drawdown (-8.2% vs -18.8%) both favour BK. See table below.
BK ALLOCATION
SPY B&H
60/40
Total Return
+54.9%
+93.3%
+44.7%
CAGR
+9.1%
+14.1%
+7.7%
Sharpe
0.58
0.76
0.34
Max DD
-8.2%
-18.8%
-12.7%
⚠ Backtest model note: This 5-year simulation uses a simplified 3-regime, 6-instrument allocation (SPY · TLT · GLD · BIL · HYG · EEM) with hardcoded regime weights — separate from the live BK Dynamic Allocation model shown above, which uses the full 97-instrument universe and the RAS formula. Assumptions: monthly rebalancing · zero transaction costs · rf = 4.5% · gross of fees · one-day execution lag applied (regime signal from day i executes on day i+1). Chart shows cumulative total return; Sharpe measures risk-adjusted return — a strategy with lower absolute return but significantly lower volatility can achieve a higher Sharpe. The chart and Sharpe column can therefore appear to disagree while both being correct.
⚠ BACKTEST METHODOLOGY — READ BEFORE ACTING
WHAT WAS TESTED
Model: 3-regime allocation (Calm/Stressed/Crisis) Universe: 6 instruments only SPY · TLT · GLD · BIL · HYG · EEM Period: Apr 2021 – Apr 2026 (5 years) Rebalancing: Monthly · First trading day Regime signal: Prior month-end classification
ASSUMPTIONS & LIMITATIONS
Transaction costs: 0 bps (gross of all fees) Risk-free rate: 4.5% annualised Execution: One-day lag applies ⚠ Does NOT test the live 97-instrument RAS model ⚠ Regime weights are hardcoded, not dynamic ⚠ Past performance ≠ future results
WHAT HAS NOT YET BEEN BACKTESTED
· BK Fragility Framework predictive validity (hit rate, false positive rate, avg drawdown after CRISIS signal) · BK Composite Score forward returns (top quintile vs bottom quintile, 21-day holding period) · Full 97-instrument RAS model backtest
REGIME ALLOCATION WEIGHTS
Instrument
MODERATE
STRESSED
CRISIS
SPY (US Equities)
45%
25%
10%
TLT (Long Treasuries)
10%
20%
20%
GLD (Gold)
10%
15%
25%
BIL (Cash / T-Bills)
5%
20%
35%
HYG (High Yield Credit)
15%
15%
5%
EEM (Emerging Markets)
15%
5%
5%
Monthly Returns — BK Allocation Strategy
Year
Jan
Feb
Mar
Apr
May
Jun
Jul
Aug
Sep
Oct
Nov
Dec
Full Year
2023
-0.2%
-3.6%
+3.8%
+0.8%
-0.7%
+3.6%
+2.5%
-2.1%
-4.1%
-1.4%
+7.3%
+4.1%
+9.9%
2024
-0.3%
+2.9%
+3.0%
-2.4%
+3.3%
+2.3%
+2.0%
+1.9%
+2.8%
-1.1%
+2.4%
-2.2%
+15.3%
2025
+2.5%
+0.5%
-1.7%
+0.6%
+1.5%
+4.0%
+1.0%
+2.0%
+4.3%
+2.2%
+0.5%
+0.4%
+19.3%
YTD
+3.2%
+1.9%
-5.4%
+1.0%
+3.4%
-1.6%
+0.1%
—
—
—
—
—
+2.5%
Monthly returns based on simplified 6-instrument backtest model. Gross of fees · Zero transaction costs assumed · rf = 4.5%
MODELS ON THIS PAGE
1. Regime-Weighted Framework Output (Illustrative) — LIVE. Risk Appetite Score = Regime(35%) + Fragility Inv(30%) + Fear & Greed(20%) + Vol Inv(15%). Drives bucket weights. Highest composite score per bucket selected by BK Composite Score (5-factor composite on Intel tab). 2. Regime Allocation Backtest — VALIDATED. Simplified 6-instrument model with monthly rebalancing. 5-year evidence. Demonstrates the regime-switching concept; not a replica of the live RAS model. Status: RAS allocation is live but not yet backtested as a full 97-instrument model. The backtest above validates the regime-switching principle only.
Edge = regime-aware portfolio intelligence · Allocation shifts automatically as market regime changes · For informational purposes only · Not investment advice · Past performance not indicative of future results
BK
Bhavesh Kamdar
FRM · CQF · Risk Manager
INVESTMENT PHILOSOPHY
"Risk is not something to be avoided — it is something to be understood, measured and navigated. After 25 years managing risk across global asset management firms, I have seen every market cycle, every crisis and every recovery. The pattern is always the same: fragility builds slowly, then breaks suddenly."
Bhavesh Kamdar is a senior risk professional with 25 years of experience in global asset management, spent building risk frameworks across equities, fixed income, commodities, and alternatives.
Holding both the Financial Risk Manager (FRM) designation and the Certificate in Quantitative Finance (CQF), Bhavesh combines deep quantitative expertise with practical investment risk management experience across equities, fixed income, commodities and alternatives.
The BK Fragility Framework was born from a simple observation: traditional risk models measure volatility after it has arrived. Bhavesh built the framework to detect structural vulnerability before it crystallises into loss.
1. Nature of This Tool
This dashboard is a personal, non-commercial research project. It is built and maintained by Bhavesh Kamdar for private analytical use. It is not a financial product, not a regulated service, and is not offered commercially to any third party.
2. Not Investment Advice
Nothing on this dashboard constitutes investment advice, a solicitation to buy or sell any security, or a recommendation of any investment strategy. All content is observational and descriptive — it describes what the models output, not what any person should do with their capital.
3. Model Outputs Are Not Predictions
Framework scores (fragility, regime, fear & greed, composite) are quantitative model outputs derived from historical price and volume data. They describe current statistical conditions — they do not predict future prices, returns, or market behaviour. Past model performance is not indicative of future results.
4. Data Limitations
Price and volume data is sourced from Yahoo Finance via yfinance. Data may contain errors, gaps, stale prices, or corporate-action anomalies (see the DATA REVIEW flag on the Performance tab for known issues). No warranty is made as to data accuracy or completeness.
5. No Commercial Relationship
This tool is not affiliated with, endorsed by, or sponsored by any employer or institution. It is developed independently in a personal capacity. No subscription, payment, or commercial arrangement exists or is offered.
6. Personal Use Only
This dashboard is hosted publicly solely for technical convenience (GitHub Pages). It is not distributed, marketed, or promoted to any audience. Any person accessing it does so for their own information and takes sole responsibility for any use they make of the content.
Open Questions
Research Directions
Analytical questions this framework has not yet answered. Each section below identifies a research gap, the methodology being explored, and what evidence would be needed to validate it.
⚖
Research Area 1
Portfolio Construction Under Regime Constraints
Does regime-conditioned mean-variance outperform unconditional?
Research question: does applying Markowitz within each regime state produce better risk-adjusted outcomes than a single static frontier?
Does risk parity hold up in Stressed regimes?
Equal risk contribution weighting tends to concentrate in low-vol assets. Exploring whether this is protective or deceptive in crisis.
What is the rebalancing frequency that maximises Sharpe net of costs?
Monthly vs quarterly vs threshold-based rebalancing. Transaction cost sensitivity across different regime states.
How much does correlation structure change between Calm and Stressed?
Rolling 60/120-day correlation matrix. Hypothesis: diversification degrades precisely when it is needed most.
Does beta to benchmark vary predictably with fragility score?
Exploring time-varying beta using 63/126-day windows against SPY and AGG. Is rising beta a leading fragility indicator?
What is the marginal value of adding a new bucket to the RAS model?
Current model has 6 buckets. Research question: does adding Alts or Crypto change regime-adjusted Sharpe materially?
▶
Research Area 2
Cross-Asset Signal Persistence
Does 12-1 momentum persist within the BK universe?
Cross-sectional momentum ranking across 113 instruments. Research question: does the signal decay faster in high-fragility regimes?
Do cointegrated pairs offer regime-independent return?
Exploring EWJ/EFA, GLD/SLV and similar pairs. Hypothesis: spread mean-reversion weakens during Crisis regime.
Does RSI divergence from price predict reversal at the asset-class level?
Mean-reversion screening. Quantifying false positive rate across regime states.
How much of the BK GREEN signal decays within 5 trading days?
Signal half-life analysis. Measuring whether composite scores lead or lag price by regime.
Is COT commercial positioning a leading indicator for commodity fragility?
Exploring CFTC data for crude, gold, wheat. Hypothesis: extreme commercial short correlates with fragility spikes.
Do earnings surprise magnitudes vary with the fragility score?
Exploring whether HIGH fragility periods coincide with larger post-earnings moves. Potential volatility timing signal.
💡
Research Area 3
Regime Detection Accuracy & Transition Lead Time
How early does HMM detect regime transitions vs GMM?
Comparing HMM and GMM lead times on the 2020 and 2022 episodes. Research question: which model minimises false positives?
Does a 3-state model outperform a 2-state model in real time?
Current model uses 3 states. Exploring whether adding a 4th (Recovery) state improves out-of-sample transition accuracy.
What is the false positive rate for transition risk flags?
Currently flagged as Elevated when models disagree. Measuring how often this precedes actual regime change vs mean-reverts.
Does analyst consensus data lead or lag the regime signal?
Exploring yfinance consensus ratios as a sentiment crosscheck on HMM/GMM regime classification.
Can macro event timing improve regime change probability estimates?
FOMC, CPI, NFP dates as covariates in regime transition probabilities. Do they add explanatory power?
Is the current 3-model consensus robust to instrument universe changes?
Testing whether adding or removing asset classes materially shifts regime classification on historical episodes.
▶
Research Area 4
Fragility Score Validation & Factor Decomposition
Does a high fragility score predict subsequent drawdown within 30 days?
Core validation question. Measuring hit rate on score > 70 → max drawdown > 10% within 21 trading days.
Which fragility pillars are most predictive of drawdown vs volatility?
Factor attribution: decomposing which of the 5 pillar components explains most of subsequent loss.
Is the equal-weighting of fragility pillars optimal?
Research question: do value/momentum/quality/low-vol weights derived from factor regressions outperform equal weights?
Does fragility dispersion across asset classes predict regime transition?
Hypothesis: rising cross-asset fragility dispersion (not just mean) is a leading regime indicator.
How does the fragility score behave for FX instruments excluded from display?
FX is excluded from the fragility tab. Research question: does including it change the system-level fragility reading?
Can the fragility score be extended to individual equities?
Current model is ETF/index-based. Exploring whether pillar methodology transfers to single-stock screening.
Research questions only · No conclusions implied · All analysis uses Yahoo Finance data · Scope subject to revision
Personal research · Not investment advice · No commercial offering