market-regime-detection

Name: market-regime-detection
Author: SKE-Labs

15stars2forks

How to install market-regime-detection?

ske-labs/agent-trading-skills/market-regime-detection

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

Installs into the current project.

Summary

Identify current market regime (trending, ranging, squeeze, or volatile) to select appropriate trading strategies. Uses only no-cost indicators (EMA slope, BB Width, ATR Ratio, Donchian) — avoids API-billed indicators like ADX/DMI/Supertrend.

Files · 1

View on GitHub

SKILL.md

1	# Market Regime Detection
2
3	Regime labels summarize observable trend and volatility states. They are model outputs with uncertainty, not natural market categories; measure whether they add held-out value to the downstream strategy.
4
5	## Regime Types
6
7	\| Regime \| EMA Slope (50, 5-bar) \| BB Width \| Price Action \|
8	\| ------------ \| --------------------- \| --------- \| ----------------------------------------- \|
9	\| Trending Up \| Positive normalized slope \| Expanding/steady \| Objective HH/HL or directional-return state \|
10	\| Trending Down \| Negative normalized slope \| Expanding/steady \| Objective LH/LL or directional-return state \|
11	\| Ranging \| Near-zero normalized slope \| Middle distribution \| Oscillation under predeclared rule \|
12	\| Squeeze \| Near-zero normalized slope \| Low percentile, optional BB inside KC \| Relative compression; direction unknown \|
13	\| Volatile / Chop \| Slope flipping run-to-run \| Very wide \| Whipsaws both directions \|
14
15	## Detection Inputs (all in the free indicator whitelist)
16
17	EMA slope candidate — compute over a predeclared horizon and normalize by price or ATR:
18	```
19	ema50_slope_pct = (ema50[-1] - ema50[-5]) / ema50[-5] * 100
20	```
21	Derive positive, negative, and near-zero bands from training data or rolling historical percentiles for the instrument/timeframe; freeze them before evaluation.
22
23	BB Width = `(bb_upper - bb_lower) / bb_middle * 100` — rolling-100-period percentile:
24	- `> 80th pctl` → expansion, supports a trend read
25	- `< 20th pctl` → squeeze candidate; no direction or imminent break is implied
26	- mid-band → ranging
27
28	ATR Ratio = `atr_14 / rolling_median(atr_14, 20)`:
29	Use rolling percentiles or training-set thresholds; ATR ratio measures volatility, not whether price is trending or choppy.
30
31	BB Inside KC — if `bb_lower > kc_lower AND bb_upper < kc_upper`, the Bollinger band is inside the Keltner channel → compression. Pair with low BB Width to confirm squeeze.
32
33	Donchian Position — `(close - donchian_lower) / (donchian_upper - donchian_lower)`:
34	- near 0 or 1 with EMA slope agreeing → confirmed trend continuation
35	- mid-band (0.4–0.6) with flat slope → ranging
36
37	## Composite Classification
38
39	\| EMA Slope \| BB Width \| ATR Ratio \| BB-in-KC \| Regime \|
40	\| ---------------- \| -------- \| --------- \| -------- \| -------------------- \|
41	\| Outside trained near-zero band \| Expanding \| Elevated percentile \| No \| Trending candidate \|
42	\| Inside trained near-zero band \| Low percentile \| Low percentile \| Yes \| Squeeze candidate \|
43	\| Inside trained near-zero band \| Middle percentile \| Middle percentile \| No \| Range candidate \|
44	\| Unstable sign \| Very high percentile \| Very high percentile \| No \| Volatile candidate \|
45	\| Mixed \| Mixed \| Mixed \| Mixed \| Transitional/uncertain \|
46
47	The table is a seed specification only. Do not count correlated inputs as independent agreement or map them directly to confidence/size. Calibrate a classifier, report posterior/calibration if available, otherwise report features plus `classified`, `uncertain`, or `no trade`.
48
49	## Regime Change Signals
50
51	\| Signal \| Meaning \|
52	\| ----------------------------------------------- \| -------------------------------------- \|
53	\| EMA50 5-bar slope crosses through ±0.15% \| Range → trend or trend → range pivot \|
54	\| BB Width expansion exceeds trained percentile/rate \| Volatility expansion; direction unknown \|
55	\| BB Width contracts to <20th pctl + BB inside KC \| Squeeze forming \|
56	\| ATR Ratio jumps above 1.5x off a low base \| Vol expansion, expect direction reveal \|
57	\| Price closes outside Donchian-20 boundary \| Breakout from prior range \|
58
59	## Strategy Selection Matrix
60
61	\| Regime \| Use These Skills \| Avoid \|
62	\| --------------- \| --------------------------------------------------- \| --------------------------------------- \|
63	\| Trending \| moving-average-crossover, pullback-trading, fibonacci-trading, trailing-stop \| mean-reversion, range-trading \|
64	\| Ranging \| mean-reversion, bollinger-bands, supply-demand-zones \| momentum-trading, MA crossover \|
65	\| Squeeze \| breakout-trading (arm BOTH directions, wait for break) \| range-trading (about to break) \|
66	\| Volatile / Chop \| wider stops, smaller size, prefer to sit out \| all trend-following, all mean-reversion \|
67	\| Transitional \| use only strategies validated in this state or wait \| untested regime substitution \|
68
69	## Workflo

People who liked this also viewed

6 picks

Type

market-regime-detection

15stars2forks

View on GitHub

How to install market-regime-detection?

ske-labs/agent-trading-skills/market-regime-detection

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

Installs into the current project.

Summary

Files · 1

View on GitHub

SKILL.md

1	# Market Regime Detection
2
3	Regime labels summarize observable trend and volatility states. They are model outputs with uncertainty, not natural market categories; measure whether they add held-out value to the downstream strategy.
4
5	## Regime Types
6
7	\| Regime \| EMA Slope (50, 5-bar) \| BB Width \| Price Action \|
8	\| ------------ \| --------------------- \| --------- \| ----------------------------------------- \|
9	\| Trending Up \| Positive normalized slope \| Expanding/steady \| Objective HH/HL or directional-return state \|
10	\| Trending Down \| Negative normalized slope \| Expanding/steady \| Objective LH/LL or directional-return state \|
11	\| Ranging \| Near-zero normalized slope \| Middle distribution \| Oscillation under predeclared rule \|
12	\| Squeeze \| Near-zero normalized slope \| Low percentile, optional BB inside KC \| Relative compression; direction unknown \|
13	\| Volatile / Chop \| Slope flipping run-to-run \| Very wide \| Whipsaws both directions \|
14
15	## Detection Inputs (all in the free indicator whitelist)
16
17	EMA slope candidate — compute over a predeclared horizon and normalize by price or ATR:
18	```
19	ema50_slope_pct = (ema50[-1] - ema50[-5]) / ema50[-5] * 100
20	```
21	Derive positive, negative, and near-zero bands from training data or rolling historical percentiles for the instrument/timeframe; freeze them before evaluation.
22
23	BB Width = `(bb_upper - bb_lower) / bb_middle * 100` — rolling-100-period percentile:
24	- `> 80th pctl` → expansion, supports a trend read
25	- `< 20th pctl` → squeeze candidate; no direction or imminent break is implied
26	- mid-band → ranging
27
28	ATR Ratio = `atr_14 / rolling_median(atr_14, 20)`:
29	Use rolling percentiles or training-set thresholds; ATR ratio measures volatility, not whether price is trending or choppy.
30
31	BB Inside KC — if `bb_lower > kc_lower AND bb_upper < kc_upper`, the Bollinger band is inside the Keltner channel → compression. Pair with low BB Width to confirm squeeze.
32
33	Donchian Position — `(close - donchian_lower) / (donchian_upper - donchian_lower)`:
34	- near 0 or 1 with EMA slope agreeing → confirmed trend continuation
35	- mid-band (0.4–0.6) with flat slope → ranging
36
37	## Composite Classification
38
39	\| EMA Slope \| BB Width \| ATR Ratio \| BB-in-KC \| Regime \|
40	\| ---------------- \| -------- \| --------- \| -------- \| -------------------- \|
41	\| Outside trained near-zero band \| Expanding \| Elevated percentile \| No \| Trending candidate \|
42	\| Inside trained near-zero band \| Low percentile \| Low percentile \| Yes \| Squeeze candidate \|
43	\| Inside trained near-zero band \| Middle percentile \| Middle percentile \| No \| Range candidate \|
44	\| Unstable sign \| Very high percentile \| Very high percentile \| No \| Volatile candidate \|
45	\| Mixed \| Mixed \| Mixed \| Mixed \| Transitional/uncertain \|
46
47	The table is a seed specification only. Do not count correlated inputs as independent agreement or map them directly to confidence/size. Calibrate a classifier, report posterior/calibration if available, otherwise report features plus `classified`, `uncertain`, or `no trade`.
48
49	## Regime Change Signals
50
51	\| Signal \| Meaning \|
52	\| ----------------------------------------------- \| -------------------------------------- \|
53	\| EMA50 5-bar slope crosses through ±0.15% \| Range → trend or trend → range pivot \|
54	\| BB Width expansion exceeds trained percentile/rate \| Volatility expansion; direction unknown \|
55	\| BB Width contracts to <20th pctl + BB inside KC \| Squeeze forming \|
56	\| ATR Ratio jumps above 1.5x off a low base \| Vol expansion, expect direction reveal \|
57	\| Price closes outside Donchian-20 boundary \| Breakout from prior range \|
58
59	## Strategy Selection Matrix
60
61	\| Regime \| Use These Skills \| Avoid \|
62	\| --------------- \| --------------------------------------------------- \| --------------------------------------- \|
63	\| Trending \| moving-average-crossover, pullback-trading, fibonacci-trading, trailing-stop \| mean-reversion, range-trading \|
64	\| Ranging \| mean-reversion, bollinger-bands, supply-demand-zones \| momentum-trading, MA crossover \|
65	\| Squeeze \| breakout-trading (arm BOTH directions, wait for break) \| range-trading (about to break) \|
66	\| Volatile / Chop \| wider stops, smaller size, prefer to sit out \| all trend-following, all mean-reversion \|
67	\| Transitional \| use only strategies validated in this state or wait \| untested regime substitution \|
68
69	## Workflo

People who liked this also viewed

6 picks

Type