Skip to content

F. Microstructure Strategies (F.23--F.26)

Exploit predictable intraday/intraweek patterns and event-driven price dislocations in precious metals. These strategies operate on shorter time horizons and require higher-frequency data and execution capabilities.

F.23 Event Drift

Status: Planned

Economic Rationale

Precious metals exhibit statistically significant post-announcement drift following key macro events: FOMC rate decisions, non-farm payrolls (NFP), CPI releases, and ECB meetings. The drift occurs because the market underreacts to the initial surprise and then reprices over 1--3 trading days. Gold's sensitivity to real-rate expectations makes FOMC events the highest-impact catalyst.

Signal Formula (Proposed)

Step 1 -- Surprise computation:

\[ \text{surprise}_t = \frac{\text{actual}_t - \text{consensus}_t}{\sigma_{\text{surprise}, 5Y}} \]

Step 2 -- Directional mapping (event-specific):

Event Gold-bullish surprise Gold-bearish surprise
FOMC Dovish (cut, pause) Hawkish (hike, taper)
NFP Miss (weak labor) Beat (strong labor)
CPI Beat (inflation up) Miss (inflation down)

Step 3 -- Drift signal:

\[ \text{signal}_t = \text{direction} \times |\text{surprise}_t| \times \text{event\_weight} \]

Enter at T+15min (after initial volatility spike subsides). Hold for 1--3 days.

Step 4 -- Triple-barrier exit:

  • Take profit: 1.5 ATR
  • Stop loss: 1 ATR
  • Max hold: 3 days

Expected Properties

  • Max capacity: $15M
  • Alpha half-life: 3 days (very short)
  • Expected Sharpe: 0.5--0.9

References

  • Lucca & Moench (2015) "The Pre-FOMC Announcement Drift"
  • Kurov & Gu (2016) "Monetary Policy and Stock Prices: Does the 'Fed Put' Work?"
  • Rosa (2014) "The High-Frequency Response of Energy Prices to U.S. Monetary Policy"

F.24 Fix Dislocation

Status: Planned

Economic Rationale

The LBMA gold fix (10:30 AM and 3:00 PM London time) creates a structural microstructure pattern. Large institutional orders (e.g., central bank purchases, ETF rebalances) that converge on the fix create predictable price pressure. The AM fix tends to set a directional bias; the PM fix often sees reversion. Dislocations between spot and fix price are tradeable.

Signal Formula (Proposed)

Pre-fix drift:

\[ \text{drift}_t = \frac{P_{t, \text{10:00}} - P_{t, \text{09:00}}}{P_{t, \text{09:00}}} \]

Fix dislocation:

\[ \text{dislocation}_t = \frac{P_{\text{fix}} - P_{\text{pre-fix}}}{P_{\text{pre-fix}}} \]
\[ z_{\text{disl}} = \text{zscore}\!\bigl(\text{dislocation}_t,\; 252\bigr) \]

Post-fix reversion: When the fix pushes price away from pre-fix level by > 1.5 sigma, fade it. Enter at fix + 5 minutes. Target: 50% reversion within 2 hours.

Expected Properties

  • Max capacity: $10M
  • Alpha half-life: Intraday
  • Expected Sharpe: 0.4--0.7
  • Key requirement: Sub-second execution capability for fix window

References

  • Caminschi & Heaney (2014) "Fixing a Leaky Fixing: Short-Term Market Reactions to the London PM Gold Price Fixing"
  • Abrantes-Metz et al. (2012) "Libor Manipulation?"

F.25 Overnight

Status: Planned

Economic Rationale

Gold exhibits a documented overnight premium: returns from COMEX close to next open tend to be positive and statistically significant (the "overnight effect"). This is driven by Asian physical demand during London/Hong Kong trading hours, when COMEX is closed. The strategy captures this by going long at COMEX close and exiting at the open.

Signal Formula (Proposed)

Unconditional:

\[ r_{\text{overnight}} = \frac{P_{\text{open}, t+1} - P_{\text{close}, t}}{P_{\text{close}, t}} \]

Historical average: ~2--4 bps/night for gold.

Conditional overlay (improve Sharpe):

\[ \text{signal}_t = \begin{cases} \text{LONG} & \text{if Shanghai premium} > 0 \wedge \text{DXY} \text{ weakening intraday} \\ \text{FLAT} & \text{otherwise} \end{cases} \]

Expected Properties

  • Max capacity: $20M
  • Alpha half-life: 1 day
  • Expected Sharpe: 0.3--0.6 (unconditional), 0.5--0.8 (conditional)
  • Turnover: Very high (daily round-trip)

References

  • Lou et al. (2019) "A Tug of War: Overnight Versus Intraday Expected Returns"
  • Barclay & Hendershott (2003) "Price Discovery and Trading After Hours"

F.26 Seasonality

Status: Planned

Economic Rationale

Gold has documented seasonal patterns driven by physical demand cycles:

  • January effect: Rebalancing flows and Chinese New Year demand (pre-buying in Dec/Jan)
  • August-September: Indian wedding/festival season (Diwali, Dhanteras) drives buying
  • Tax-loss selling: Year-end selling pressure in December

These patterns are well-known, so the strategy uses them as a secondary overlay (10--20% weight) rather than a standalone signal.

Signal Formula (Proposed)

\[ \text{seasonal\_score}_{m} = \frac{\bar{r}_{m} - \bar{r}_{\text{all}}}{\sigma_{m}} \]

where \(\bar{r}_m\) is the average monthly return for calendar month \(m\) over the trailing 20 years.

Regime gate: Only activate when the seasonal signal aligns with at least one other macro signal (A.1--A.5).

Seasonal Calendar

Month Historical Bias Driver
Jan Bullish (+) CNY front-running, rebalancing
Feb Neutral Post-CNY digestion
Mar Slight bear (-) Tax-related selling
Apr Neutral
May Slight bear (-) "Sell in May" cross-asset
Jun Neutral
Jul Slight bull (+) Early Indian wedding buying
Aug Bullish (+) Indian demand ramp
Sep Bullish (++) Peak Indian demand (Dhanteras/Diwali)
Oct Neutral
Nov Slight bull (+) Chinese/Indian demand
Dec Mixed Tax-loss selling vs holiday demand

Expected Properties

  • Max capacity: Overlay only (no standalone capacity)
  • Alpha half-life: 30 days per episode
  • Expected Sharpe: 0.3--0.5 (standalone)

References

  • Baur (2013) "The Autumn Effect of Gold"
  • Lucey & Tully (2006) "Seasonality, Risk and Return in Daily COMEX Gold and Silver Data 1982-2002"