Several updates have been made to the core calculation engine of Performance Watcher. These centre around volatility calculations for indices and the introduction of automated composites for index construction.
1. Portfolio and Composite Calculations
Performance (unchanged from PW2 → PW3):
Both use true time-weighted daily returns that account for inflows/outflows, avoiding distortions.Formula: Rt=(NAVt+OFt–IFt)/NAVt−1–1R_t = (NAV_t + OF_t – IF_t) / NAV_{t-1} – 1
Cumulative performance = chained daily returns.
Volatility
PW2: 7-day moving average of log-returns, annualised, with smoothing coefficient (φ = 0.9). Adjusted weekend effects but introduced scaling bias.
PW3: Realised standard deviation of trading-day returns. Scaling ensures composite volatility matches constituent average, allowing robust index construction.
2. Index Construction
PW2:
Equal-weight averages of discretionary portfolios per currency/risk. Minimum 3 portfolios + NAV thresholds. Capped weighting prevented dominance. Hybrid models mixed portfolios with strategy funds.
Issues: sensitive to outliers, overweight on funds.PW3:
Outlier cleaning: inter-decile filter if ≥15 portfolios per risk/currency.
Portfolios grouped into ecosystem composites, equal-weighted within ecosystems.
Fund composite fallback: 7–10 accumulation funds per risk/currency when ecosystem participation is too low.
Weighted blending: ecosystem composites dominate; funds fade as community depth grows.
Volatility: annualised, with scaling factor correcting diversification effects.
Ensures that 3-month volatility reflects the average of constituents.
3. Perfometer (Weather Score)
PW2:
Based on Sharpe ratio difference:
ϕ=(Rp–Rf)/σp–(RB–Rf)/σB\phi = (R^p – R_f)/σ^p – (R^B – R_f)/σ^B.
Issues:Creates unsatisfactory results in adverse market conditions, as shown in the chart below.
Scale mismatch between returns and volatility.
Long-term comparisons distorted.
Behavioural loss aversion not considered.
PW3:
Introduces an asymmetric framework:Losses penalised more than gains (behavioural, mathematical, accountability rationale).
High-volatility portfolios penalised even with strong returns.
Capital preservation rewarded in down markets.
Non-linear scoring replaces Sharpe difference.
Horizon-normalised for comparability.
Six refined weather levels: Radiant, Sunny, Sunny Spells, Cloudy, Overcast, Rainy.
4. Summary Table
Aspect | PW2 | Issues | PW3 Fix |
|---|---|---|---|
Portfolio Performance | Time-weighted daily returns | – | Same method retained |
Volatility | 7-day MA, φ=0.9 | Biased, inconsistent scaling | Realised 3M log-return volatility, |
Index Construction | Equal-weight avg. + fund mix, Volatility method | Outliers, fund-heavy, proxy volatility can be off target | Outlier cleaning + ecosystem composites + fund fallback + accurate average component vol calculation |
Perfometer | Sharpe ratio diff, symmetric | Scale mismatch, ignores asymmetry | Asymmetric, non-linear, horizon-normalised |
5. Key Takeaways
Consistency: Performance calculation remains robust and unchanged.
Precision: Volatility and index methodologies now reflect reality better.
Resilience: Outlier cleaning and ecosystem composites improve comparability.
Fairness: Perfometer now aligns with investor psychology and compounding math.
Transparency: Weather scores give a more intuitive and trustworthy summary.