market-correlation-trading

Name: market-correlation-trading
Author: SKE-Labs

15stars2forks

How to install market-correlation-trading?

ske-labs/agent-trading-skills/market-correlation-trading

$npx -y skills add SKE-Labs/agent-trading-skills --skill market-correlation-trading

Installs into the current project.

Summary

Measure time-varying cross-asset correlation, beta, lead-lag, and spread stationarity. Use when evaluating hedges, common factors, regime breaks, divergences, or convergence hypotheses.

Files · 1

View on GitHub

SKILL.md

1	# Market Correlation Trading
2
3	Measure time-varying co-movement and test whether a hedge, factor exposure, or stationary spread exists. Divergence can reflect a regime change; correlation alone does not imply convergence, causation, or lead-lag.
4
5	## Measurement
6
7	- Align exchange calendars, timestamps, currencies, and return intervals; do not correlate price levels.
8	- Compute rolling Pearson and rank correlation over predeclared short and long windows, with confidence intervals and enough observations.
9	- Estimate rolling beta or a multi-factor model when the objective is hedging.
10	- Require an economically linked, stationary spread and out-of-sample half-life before proposing convergence.
11	- Test lead-lag with lagged returns while controlling common factors and multiple comparisons.
12
13	## Divergence Signals
14
15	\| Scenario \| Signal \|
16	\| --- \| --- \|
17	\| BTC rallies but Nasdaq flat/down \| BTC overextended, watch for pullback \|
18	\| Gold falls but real yields also fall \| Gold undervalued, potential buy \|
19	\| Oil rises but energy stocks lag \| Energy stocks may catch up \|
20	\| VIX rises but S&P holds \| Hedging without selling, watch for resolution \|
21
22	Use 20/60 bars only as example windows. Calibrate the windows and breakdown threshold in training data, then report estimates, uncertainty, and observation count.
23
24	## Lead-Lag Hypotheses
25
26	For every proposed leader/follower pair, state the economic mechanism, data clock, lags tried, common-factor controls, and held-out result. Never copy a fixed lag table into a trade.
27
28	## Regime-Dependent Correlations
29
30	\| Regime \| Correlation Behavior \|
31	\| --- \| --- \|
32	\| Risk-on (normal) \| Traditional correlations hold \|
33	\| Risk-off (crisis) \| Everything correlates -- stocks, crypto, commodities drop; only USD/treasuries/gold rise \|
34	\| Inflation \| Stocks and bonds fall together (atypical positive correlation) \|
35	\| Deflation \| Stocks fall, bonds rise (traditional inverse) \|
36
37	## Workflow
38
39	### 1. Check Related Assets
40
41	```
42	get_candles(symbol="SPY", exchange=<exchange>, interval="1D", count=250)
43	get_candles(symbol="QQQ", exchange=<exchange>, interval="1D", count=250)
44	get_candles(symbol="TLT", exchange=<exchange>, interval="1D", count=250)
45	get_candles(symbol="GLD", exchange=<exchange>, interval="1D", count=250)
46	```
47
48	Compare recent performance (1W, 1M, 3M) across correlated pairs.
49
50	### 2. Research Macro Context
51
52	```
53	get_financial_news(topic="correlation stocks bonds regime shift <current year>", max_results=10)
54	get_economics_calendar(from_date=<start>, to_date=<end>, impact="high")
55	```
56
57	Determine whether current correlations are driven by Fed policy, inflation, or event-specific factors. High-impact macro events can trigger regime shifts.
58
59	### 3. Identify and Report Divergences
60
61	Report return definition, windows, estimates/uncertainty, beta/factor exposures, divergence z-score, stationarity result, costs, and `hedge`, `watch`, or `no trade`. Do not call an asset mispriced from correlation alone.
62
63	## Evidence and Validation
64
65	- Treat the setup as a testable hypothesis, not a prediction. Define thresholds, entry, invalidation, and exit before evaluating outcomes.
66	- Calibrate on the same instrument, venue, session, and timeframe. Use closed candles and a held-out or walk-forward sample; record every variant tried.
67	- Include spread, fees, slippage, borrow or funding, partial fills, and latency. Reject the setup when net expectancy is not positive or depends on one narrow parameter.
68	- Return observed inputs, missing data, cost assumptions, entry, invalidation, exit, and a valid, watch, or no-trade status.
69	- Research basis: [Ang & Bekaert](https://www.nber.org/papers/w7056) finds correlations and volatilities vary by regime and often rise in bad times; a static correlation table is not a trading signal.
70
71	## Key Rules
72
73	- Never assume correlations are permanent; show rolling estimates and adverse stresses toward +1 or -1.
74	- Include macro/factor regime and stress correlations; diversification can weaken without every asset moving together.
75	- Aggregate positions with shared beta/factor/scenario risk rather than relying on one pair correlation.
76	- Correlation is not causation or convergence; control for common drivers where possible.
77	- When a historical relationship changes, test for a regime break before proposing convergence.
78
79	## Related Skills
80
81	- sector-rotation -- Sector correlations reveal rotation opportunities
82	- economic-calendar-trading -- Macro events drive correlation regime shifts

People who liked this also viewed

6 picks

Type

market-correlation-trading

15stars2forks

View on GitHub

How to install market-correlation-trading?

ske-labs/agent-trading-skills/market-correlation-trading

$npx -y skills add SKE-Labs/agent-trading-skills --skill market-correlation-trading

Installs into the current project.

Summary

Measure time-varying cross-asset correlation, beta, lead-lag, and spread stationarity. Use when evaluating hedges, common factors, regime breaks, divergences, or convergence hypotheses.

Files · 1

View on GitHub

SKILL.md

1	# Market Correlation Trading
2
3	Measure time-varying co-movement and test whether a hedge, factor exposure, or stationary spread exists. Divergence can reflect a regime change; correlation alone does not imply convergence, causation, or lead-lag.
4
5	## Measurement
6
7	- Align exchange calendars, timestamps, currencies, and return intervals; do not correlate price levels.
8	- Compute rolling Pearson and rank correlation over predeclared short and long windows, with confidence intervals and enough observations.
9	- Estimate rolling beta or a multi-factor model when the objective is hedging.
10	- Require an economically linked, stationary spread and out-of-sample half-life before proposing convergence.
11	- Test lead-lag with lagged returns while controlling common factors and multiple comparisons.
12
13	## Divergence Signals
14
15	\| Scenario \| Signal \|
16	\| --- \| --- \|
17	\| BTC rallies but Nasdaq flat/down \| BTC overextended, watch for pullback \|
18	\| Gold falls but real yields also fall \| Gold undervalued, potential buy \|
19	\| Oil rises but energy stocks lag \| Energy stocks may catch up \|
20	\| VIX rises but S&P holds \| Hedging without selling, watch for resolution \|
21
22	Use 20/60 bars only as example windows. Calibrate the windows and breakdown threshold in training data, then report estimates, uncertainty, and observation count.
23
24	## Lead-Lag Hypotheses
25
26	For every proposed leader/follower pair, state the economic mechanism, data clock, lags tried, common-factor controls, and held-out result. Never copy a fixed lag table into a trade.
27
28	## Regime-Dependent Correlations
29
30	\| Regime \| Correlation Behavior \|
31	\| --- \| --- \|
32	\| Risk-on (normal) \| Traditional correlations hold \|
33	\| Risk-off (crisis) \| Everything correlates -- stocks, crypto, commodities drop; only USD/treasuries/gold rise \|
34	\| Inflation \| Stocks and bonds fall together (atypical positive correlation) \|
35	\| Deflation \| Stocks fall, bonds rise (traditional inverse) \|
36
37	## Workflow
38
39	### 1. Check Related Assets
40
41	```
42	get_candles(symbol="SPY", exchange=<exchange>, interval="1D", count=250)
43	get_candles(symbol="QQQ", exchange=<exchange>, interval="1D", count=250)
44	get_candles(symbol="TLT", exchange=<exchange>, interval="1D", count=250)
45	get_candles(symbol="GLD", exchange=<exchange>, interval="1D", count=250)
46	```
47
48	Compare recent performance (1W, 1M, 3M) across correlated pairs.
49
50	### 2. Research Macro Context
51
52	```
53	get_financial_news(topic="correlation stocks bonds regime shift <current year>", max_results=10)
54	get_economics_calendar(from_date=<start>, to_date=<end>, impact="high")
55	```
56
57	Determine whether current correlations are driven by Fed policy, inflation, or event-specific factors. High-impact macro events can trigger regime shifts.
58
59	### 3. Identify and Report Divergences
60
61	Report return definition, windows, estimates/uncertainty, beta/factor exposures, divergence z-score, stationarity result, costs, and `hedge`, `watch`, or `no trade`. Do not call an asset mispriced from correlation alone.
62
63	## Evidence and Validation
64
65	- Treat the setup as a testable hypothesis, not a prediction. Define thresholds, entry, invalidation, and exit before evaluating outcomes.
66	- Calibrate on the same instrument, venue, session, and timeframe. Use closed candles and a held-out or walk-forward sample; record every variant tried.
67	- Include spread, fees, slippage, borrow or funding, partial fills, and latency. Reject the setup when net expectancy is not positive or depends on one narrow parameter.
68	- Return observed inputs, missing data, cost assumptions, entry, invalidation, exit, and a valid, watch, or no-trade status.
69	- Research basis: [Ang & Bekaert](https://www.nber.org/papers/w7056) finds correlations and volatilities vary by regime and often rise in bad times; a static correlation table is not a trading signal.
70
71	## Key Rules
72
73	- Never assume correlations are permanent; show rolling estimates and adverse stresses toward +1 or -1.
74	- Include macro/factor regime and stress correlations; diversification can weaken without every asset moving together.
75	- Aggregate positions with shared beta/factor/scenario risk rather than relying on one pair correlation.
76	- Correlation is not causation or convergence; control for common drivers where possible.
77	- When a historical relationship changes, test for a regime break before proposing convergence.
78
79	## Related Skills
80
81	- sector-rotation -- Sector correlations reveal rotation opportunities
82	- economic-calendar-trading -- Macro events drive correlation regime shifts

People who liked this also viewed

6 picks

Type