Expand model tag support: add GLM-5.1, simplify Anthropic IDs, scan tags anywhere in message

- Flink update_bars debouncing
- update_bars subscription idempotency bugfix
- Price decimal correction bugfix of previous commit
- Add GLM-5.1 model tag alongside renamed GLM-5
- Use short Anthropic model IDs (sonnet/haiku/opus) instead of full version strings
- Allow @tags anywhere in message content, not just at start
- Return hasOtherContent flag instead of trimmed rest string
- Only trigger greeting stream when tag has no other content
- Update workspace knowledge base references to platform/workspace and platform/shapes
- Hierarchical knowledge base catalog
- 151 Trading Strategies knowledge base articles
- Shapes knowledge base article
- MutateShapes tool instead of workspace patch

gateway/knowledge/trading/strategies/etfs/alpha-rotation.md

@@ -0,0 +1,46 @@
+---
+description: "Rotates into sector ETFs with the highest Jensen's alpha estimated from a Fama-French factor regression, replacing raw cumulative return momentum with factor-adjusted alpha as the ranking signal."
+tags: [etfs, alpha, momentum, fama-french, rotation]
+---
+
+# Alpha Rotation
+
+**Section**: 4.2 | **Asset Class**: ETFs | **Type**: Momentum / Factor-Adjusted Rotation
+
+## Overview
+Alpha rotation is structurally the same as the sector momentum rotation strategy (Section 4.1), but replaces cumulative ETF returns `R_i^cum` with ETF alphas `alpha_i`. These alphas are the Jensen's alpha regression coefficients from a serial regression of each ETF's returns on the Fama-French factors, representing the ETF's return unexplained by common risk factors.
+
+## Construction / Signal
+Run a serial regression of ETF returns `R_i(t)` on the 3 Fama-French factors (MKT, SMB, HML):
+
+```
+R_i(t) = alpha_i + beta_{1,i} MKT(t) + beta_{2,i} SMB(t) + beta_{3,i} HML(t) + epsilon_i(t)   (364)
+```
+
+The regression coefficient `alpha_i` (Jensen's alpha) corresponds to the intercept and measures the ETF's risk-adjusted excess return relative to the Fama-French model. This alpha replaces `R_i^cum` as the ranking criterion.
+
+ETFs are ranked by `alpha_i` (descending). Buy top-decile ETFs (highest alpha) and optionally short bottom-decile ETFs (lowest/most-negative alpha).
+
+## Entry / Exit Rules
+- **Entry**: At rebalance, estimate alpha for each ETF over the estimation period; rank and enter positions in top-decile (long) and optionally bottom-decile (short).
+- **Exit**: Hold for the standard holding period; rebalance at next scheduled interval.
+- **Estimation period**: Typically 1 year; returns `R_i(t)` are daily or weekly.
+
+## Key Parameters
+- **Factor model**: 3 Fama-French factors (MKT, SMB, HML); note alpha here is Jensen's alpha for ETF returns, not mutual fund alpha
+- **Estimation period**: Typically 1 year
+- **Return frequency for regression**: Daily or weekly `R_i(t)`
+- **Holding period**: Same as sector momentum rotation (1–3 months)
+- **Ranking criterion**: `alpha_i` (intercept of Fama-French regression)
+
+## Variations
+- **4-factor model**: Add Carhart momentum factor MOM(t) to regression for a 4-factor alpha
+- **R-squared augmentation**: Combine alpha ranking with R-squared selectivity measure (see Section 4.3)
+- **Long-only**: Buy only top-decile ETFs by alpha
+
+## Notes
+- Estimation period is typically 1 year with daily or weekly return data.
+- Jensen's alpha here is defined for ETF returns (not mutual fund returns as in Jensen, 1968).
+- Alpha rotation is analytically cleaner than raw momentum rotation as it removes systematic factor exposures from the ranking signal.
+- The MA filter and dual-momentum variations from Section 4.1.1 and 4.1.2 can also be applied here.
+- Can be combined with R-squared (Section 4.3) to further refine ETF selection.