This diagram describes a complete event‑driven research framework built around ATR compression/expansion. Interpreting it step‑by‑step:

1. ATR Compression → Expansion → Event (T0)

- ATR Compression: Period where Average True Range (ATR) is significantly lower than its own recent history → volatility contraction (price “coiling”).

- Expansion Event: A sudden increase in ATR (and usually range and volume) breaks the compression → volatility expansion.

- Event = T0: The bar/day where expansion is detected is tagged as the “event day” (T0). All analysis is anchored to this point.

2. Persistence Metric (Continuous Value)

- A continuous measure that captures “how long” and “how strongly” the post‑event state persists. Example research metrics:

- Rolling ATR ratio (ATR(t)/ATR(T0)) over next N bars.

- Fraction of bars in next N days where volatility stays above a threshold.

- Trend persistence: % of closes above/below T0 close or T0 breakout level.

- Purpose: Quantify whether ATR expansion leads to a sustained regime (trending/volatile) or quickly fades.

3. Event Attributes (Features at T0)

These are the characteristics of the event candle and local context at T0, used later as explanatory variables:

- ATR magnitude

- Relative ATR: ATR(T0) / ATR(lookback) (e.g., 14D/100D).

- Absolute ATR (in % of price) to normalize across stocks.

- Direction

- Upward expansion (large green candle, close near high) vs downward expansion (large red candle, close near low).

- Event age

- Where T0 lies in a broader structure: e.g.,

- Age of prior trend (how long previous up/down move persisted),

- Days since last significant ATR expansion,

- Distance from recent swing high/low.

- Idea: You create a feature vector for each event:

- Example: `[ATR_ratio, direction, prior_trend_strength, volume_surge, gap_presence, etc.]`.

4. Two Main Future Paths: Continuation vs Retracement

A. Continuation Path

- You assume T0 expansion may kick off a trend/impulse move.

- Measure outcomes via:

- MFE (Maximum Favourable Excursion): Max gain from T0 over a forward window.

- Example: MFE_10D = max( (High_t - Close_T0) / Close_T0 ) for t in [T0+1 … T0+10].

- 10D / 20D returns: Simple forward returns; e.g.,

- R10 = (Close_T0+10 − Close_T0) / Close_T0,

- R20 similarly.

- Research goals:

- How often does price continue in the event direction?

- What is the distribution of MFE and 10D/20D returns by event attributes (ATR magnitude, direction, etc.)?

- Example filters: “Upward expansion + high ATR_ratio + near 52‑week high” → does this subset show strong positive median 20D returns?

B. Retracement Path

- Here, you focus on mean reversion / failure after expansion.

- Measure outcomes via:

- MAE (Maximum Adverse Excursion): Max loss from T0 over a forward window.

- Example: MAE_10D = max( (Close_T0 − Low_t) / Close_T0 ) for t in [T0+1 … T0+10].

- Drawdown: Depth and duration of pullback after T0 before any new high/low.

- Research goals:

- For which attribute combinations do events mostly revert instead of trend?

- How large and how fast are typical retracements after an expansion event?

5. Statistical Validation: Discovery → Validation → Confirmation

The pipeline implied by the diagram is:

1. Discovery Phase

- Use historical data (e.g., NIFTY 500 stocks, daily data) to:

- Identify all ATR compression → expansion events.

- Compute attributes at T0 and forward outcomes (MFE, MAE, 10D/20D returns, drawdowns).

- Explore cross‑sections:

- By ATR magnitude buckets, direction, trend context, sector, market regime, etc.

- Aim: generate candidate hypotheses, e.g.,

- “Upward expansion events after a 20‑day low ATR regime, breaking above recent range high, have high positive 20‑day expectancy.”

2. Validation Phase

- Test these candidate rules on out‑of‑sample or more recent data:

- Use different time period or stock universe.

- Check robustness to parameter changes (lookback variations, thresholds).

- Apply statistical tests:

- Sharpe/Sortino of event‑based returns,

- t‑tests on mean forward returns vs. 0,

- Bootstrapping or Monte Carlo to test if performance is statistically significant vs noise.

3. Confirmation Phase

- Only retain patterns that:

- Hold across multiple samples and parameter variations,

- Do not degrade meaningfully out‑of‑sample,

- Survive transaction cost and slippage assumptions (important in Indian equities, especially midcaps/smallcaps).

- A “confirmed” pattern can then be used as:

- A signal component in a systematic strategy, or

- An event filter for discretionary trade setups (e.g., only trade breakouts when event attributes and validated stats are favourable).

6. How This Becomes a Practical Research Workflow (Example, not advice)

A concise, implementable structure:

1. Define compression/expansion (e.g., ATR_14 < 20‑day ATR_14 percentile for compression; expansion when ATR_14 crosses above 80th percentile with a large range bar).

2. Mark event T0 for each occurrence in your historical Indian equity universe.

3. Compute attributes: ATR_ratio, direction, volume_surge, distance to 52‑week high/low, prior trend strength, event age, etc.

4. Compute outcomes: MFE_10D, MFE_20D, MAE_10D, MAE_20D, forward returns, max drawdown from T0, etc.

5. Segment events by attributes and study distributions of these outcomes.

6. Develop rules for continuation vs retracement expectations and test them in discovery, then validation, then confirmation as per the diagram.

If you have any further queries, please connect with us on 022-6290-10141 (Timings : 09.00 AM to 05.00 PM) or you can email us on info@cniinfoxchange.com