Jesse C. Stine "100x Superstock" A/H Selection Model v0.1

Date: 2026-06-17

Owner: JimSimons@breeze

Scope: A shares + Hong Kong shares

Status: research/spec only; no production orders

One-Line Definition

This model searches for rare, early-to-mid stage superstocks: stocks with extreme relative strength, orderly institutional accumulation, a large theme/business runway, and low-risk pullback or base-breakout entry points. It is an observation/candidate engine until separately backtested and approved.

Non-Negotiable Boundary

- This is not a replacement for Pinecorn P1 v1.5.

- It does not generate planned_orders.

- It does not change M-filter, CIO sign-off, exits, sizing, or SWS L2 cap rules.

- It may become a sidecar candidate/ranking model after A/H historical validation.

Data Priority

Per Kaite 2026-06-17 instruction:

- A shares: use RQData/RQSDK first.

- Hong Kong shares: use local RQData if coverage is available; otherwise use Wind for HK price, turnover, market cap, and financial fields.

- Wind is fallback for fields RQData cannot cover, not the default for A-share fields already available in RQData.

Universe

A Share Universe

Hard filters:

- Non-ST and name does not contain ST.

- Listed at least 252 trading days.

- 20d average traded value >= 80m RMB.

- Float market cap >= 3bn RMB.

- Close >= 3 RMB.

- Exclude stocks with long suspension gaps or unreliable price history.

Hong Kong Universe

Hard filters:

- HKEX main board preferred; GEM/very illiquid names excluded unless explicitly enabled.

- Listed at least 252 trading days.

- 20d average traded value >= 30m HKD for observation, >= 80m HKD for production-grade candidate.

- Market cap >= 3bn HKD; stricter >= 5bn HKD for production-grade candidate.

- Close >= 1 HKD.

- Exclude shell-like, prolonged-suspension, and extreme spread names.

Model Layers

Layer 1: Superstock DNA

Purpose: avoid ordinary momentum names that lack 100x-type optionality.

Score: 25 points.

Signals:

- Theme runway: belongs to an expanding high-TAM industry or structural cycle.

- Revenue acceleration: latest TTM or annual revenue growth >= 20%, with acceleration vs prior period preferred.

- Profitability inflection: net profit, gross margin, or operating cash flow improving from low base.

- Market cap runway: small/mid-cap enough to plausibly compound, but not too illiquid.

- Scarcity/leadership: leading name in a narrow theme, not the 10th follower.

Implementation:

- A shares: RQData PIT fundamentals + SWS/Wind industry where available.

- HK: Wind or RQData fundamentals; if PIT availability is weak, tag as lower-confidence rather than force a clean pass.

Layer 2: Relative Strength Leadership

Purpose: Stine-style superstocks must already be proving themselves in price before entry.

Score: 25 points.

Signals:

- 250d relative strength percentile >= 90%.

- 120d return percentile >= 90%.

- Close >= 85% of 52w high.

- Stock outperforms its local benchmark and industry peer basket.

- "Canary" check: if the stock loses RS while the market or peer leaders hold up, downgrade.

Benchmark:

- A shares: CSI 300/CSI 800 plus industry peer basket.

- HK: Hang Seng Composite or Hang Seng Tech when appropriate, plus HK peer basket.

Layer 3: Ordered Advance Quality

Purpose: distinguish controlled accumulation from chaotic speculative spikes.

Score: 20 points.

Signals:

- Weekly close above 10w EMA and 30w MA.

- 30w MA rising.

- Pullbacks are contained: latest 60d max drawdown preferably <= 25%.

- Up weeks show stronger volume than down weeks.

- Volatility contracts during bases/pullbacks.

- Avoid parabolic exhaustion: price > 25% above 10w EMA gets overheat penalty.

Interpretation:

- Good: stair-step advance, tight bases, repeated support at 10w/30w.

- Bad: vertical spike, huge gap sequence, or repeated high-volume downside breaks.

Layer 4: Magic Support Line / Low-Risk Zone

Purpose: entries should be near controllable risk points, not after euphoric extension.

Score: 20 points.

Primary support lines:

- Fast superstock support: 10-week EMA.

- Structural support: 30-week MA.

Low-risk entry conditions:

- Close is within -3% to +8% of 10w EMA after a pullback, and price has reclaimed the line; or

- Close is within -3% to +6% of 30w MA in a mature base, and weekly trend remains intact; or

- Breakout from a tight base with close > pivot * 1.005, but entry is not more than 15% above 10w EMA.

Positioning:

- First entry: 1/3 to 1/2 risk budget.

- Add-on: only after profitable confirmation or second tight-base breakout.

- No full-size entry on extended breakout.

Layer 5: BCD / Danger-State Exclusion

Purpose: avoid names where the superstock structure is already breaking.

Score: -30 to 0 penalty.

Operational danger checklist:

- Weekly close below 30w MA.

- Two weekly closes below 10w EMA with expanding volume.

- 20d or 60d RS drops below 70th percentile after prior leadership.

- Three distribution weeks in the latest five weeks.

- Failed breakout: close back below pivot within 5 sessions on heavy volume.

- Large earnings/announcement red flag not resolved.

- Sector canary: peer leaders break support first.

Candidate is blocked when danger penalty <= -20.

Scorecard

Candidate tiers:

- Tier S: score >= 85, no BCD block, entry in low-risk zone.

- Tier A: score 75-84, no BCD block, watch for low-risk entry.

- Tier B: score 65-74, observation only.

- Reject: score < 65 or BCD blocked.

Buy Logic

Setup Buy

Trigger:

- Tier S/A candidate.

- Price is near 10w EMA or 30w MA low-risk zone.

- Weekly trend intact.

- No BCD block.

Action:

- Observation model marks stine_setup_buy_candidate.

- Production action requires separate CIO sign-off and backtest approval.

Breakout Buy

Trigger:

- Tier S/A candidate.

- Tight base >= 4 weeks.

- Close > pivot * 1.005.

- Volume >= 1.5x 20d average.

- Close not more than 15% above 10w EMA.

Action:

- Mark stine_breakout_candidate.

- If >15% above 10w EMA, mark extended_breakout_watch instead.

Sell / De-Risk Logic

Hard exit signals for a future paper model:

- Weekly close below 30w MA.

- Close breaks entry stop or -10% from entry, whichever is tighter after slippage buffer.

- BCD block appears after entry.

- Failed breakout with heavy-volume reversal.

Trim / risk-reduction signals:

- Price > 25% above 10w EMA.

- Three consecutive wide-range up weeks after a long run.

- Volume climax plus next-week reversal.

- RS remains high but absolute trend starts breaking.

A/H Adaptation

Shared rules:

- Weekly trend and support-line logic.

- Relative strength percentile.

- Base breakout and low-risk zone.

- BCD danger-state checklist.

A-share-specific:

- ST filter.

- 10%/20% limit-up behavior must be tracked; avoid buying locked limit-up candidates.

- Use SWS L2 for industry/peer context.

- Existing P1 liquidity gate is reusable.

- EPS acceleration is the preferred fundamental engine where quarterly data supports it.

- Short-interest squeeze potential is not treated as available in A shares; do not use turnover, northbound, or margin-financing fields as an "equivalent" proxy.

HK-specific:

- No daily limit-up structure; use spread/turnover/liquidity controls instead.

- Exclude very low-priced and shell-like names more aggressively.

- Use HKD traded value and market cap.

- Southbound flow can be optional sponsorship proxy if available; otherwise use turnover expansion and institutional ownership proxy.

- Financial reporting frequency is lower; missing PIT fundamentals should lower confidence rather than create false precision.

- EPS QoQ acceleration is generally not available with the same frequency as A shares; use revenue acceleration, OCF acceleration, and semiannual YoY acceleration instead.

- HK short-interest / short-selling data can be a valid separate signal when available.

CIO Addendum: Phase 0 Constraints

Added after RayDalio CIO review on 2026-06-17. These are framework constraints for Phase 0 and backtest design; they are not production thresholds until validated.

1. EPS Acceleration Sequence

Stine-style fundamentals should prefer an acceleration sequence, not a single-point growth pass.

Mechanical A-share target:

- Latest 4 quarters EPS YoY or QoQ-style growth sequence when PIT quarterly data supports it.

- Score the slope and persistence of acceleration, e.g. improving from negative/low growth to multiple sequential high-growth quarters.

- Treat this as the first concrete expression of Superstock DNA, ahead of generic "growth is good" rules.

Important distinction:

- This is more aggressive than ordinary CANSLIM-style single latest-quarter YoY screening.

- If quarterly PIT fields are incomplete or stale, tag the signal eps_acceleration_unavailable rather than backfilling with a weaker proxy.

2. Hong Kong Fundamental Gap

HK stocks should not pretend to have the same EPS QoQ acceleration signal.

HK replacement anchors:

- Revenue acceleration across interim / annual reporting periods.

- Operating cash flow acceleration.

- Gross margin and operating margin inflection.

- Semiannual YoY acceleration where available.

Data quality must be explicit:

- fundamental_anchor=eps_acceleration for A-share names with valid quarterly data.

- fundamental_anchor=semiannual_revenue_ocf_acceleration for HK names.

- fundamental_anchor=low_confidence when HK reporting cadence or missing fields prevents clean measurement.

3. Short Interest Difference

Do not force one signal across both markets.

- A shares: short-interest squeeze is unavailable / not reliable enough for this model; no proxy substitution.

- HK: short-selling / short-interest data can be used when available and liquid enough.

Allowed A-share substitutes:

- None for short interest itself.

- Turnover, northbound, and capital flow may be separate sponsorship / demand signals, but must not be labelled as short-interest equivalents.

4. A/H Alpha Source Split

The shared scorecard is acceptable, but alpha drivers should be interpreted separately.

Expected A-share drivers:

- EPS acceleration sequence.

- Breakout / RS leadership.

- Controlled pullback to 10w/30w support.

Expected HK drivers:

- Base breakout.

- Insider buying / ownership signal when available.

- Short-selling / short-interest squeeze signal when available.

- Liquidity improvement from a small base.

Backtests must report A and H separately before any combined conclusion.

5. Holding Period Calibration

The book's 6-18 month holding-period framing may be too long for A-share momentum cycles.

Backtest variants:

- 3-6 month max-hold.

- 6-9 month max-hold.

- 9-18 month max-hold as an explicit Stine-original sensitivity, not the default.

Exit design should measure alpha decay after breakout / support-line entry rather than assume the original US mid-cap holding period transfers unchanged.

6. "100x" Narrative Boundary

Do not use "100x" as a mechanical field name or LP expectation.

Production / LP-safe naming:

- stine_superstock

- superstock_ah

- stine_ah_score

Avoid:

- hundred_bagger_score

- 100x_candidate

- any wording implying expected 100x return.

Narrative boundary:

- Jesse Stine's extreme historical return should be treated as an outlier case from a specific post-crisis catch-up regime and concentrated small-base compounding.

- A realistic research target, if validated, is improved drawdown-adjusted growth-stock selection, not a repeatable 100x promise.

Relationship To Existing P1/O'Neil Model

P1/O'Neil v1.5:

- Stronger fundamental hard gates.

- Weekly production scan can generate planned_orders after CIO sign-off.

- Optimized around A-share leader momentum with strict execution controls.

Stine Superstock v0.1:

- More tolerant of early profitability inflection and low-base stories.

- More focused on 10w/30w support-line entry quality and avoiding extension.

- Explicitly A/H cross-market.

- Research-only until backtested.

Implementation Plan

Phase 0: Feasibility

- Confirm A/H historical coverage for daily OHLCV, market cap, turnover, and basic financials.

- Confirm HK RQData coverage; if incomplete, document Wind fallback fields.

- Produce current A/H candidate snapshot.

Phase 1: Research-Only Scanner

Outputs:

- stine_superstock_ah_candidates.csv

- stine_superstock_ah_scorecard.csv

- stine_superstock_ah_summary.md

Columns:

- market, order_book_id / windcode, name

- date, close, traded_value_20d, market_cap

- rs_250d_pct, ret_120d_pct, dist_to_52w_high

- ma10w_distance, ma30w_distance

- ordered_advance_score, low_risk_zone_flag

- breakout_flag, extended_flag

- bcd_penalty, bcd_reasons

- stine_score, stine_tier

- data_source, data_quality_flag

Phase 2: Historical A/B

Compare:

- Existing P1 v1.5 baseline.

- Stine-only A-share model.

- Stine overlay on P1 candidate universe.

- A/H combined observation model.

Ship gate:

- No production use unless A/B improves Sharpe or drawdown-adjusted return without materially increasing tail risk.

- HK model must pass separate liquidity/slippage stress.

Initial Recommendation

Build as a research sidecar first, not as a production replacement. The model is most valuable as:

1. a cross-market A/H superstock watchlist,

2. an entry-quality overlay to avoid buying extended breakouts,

3. a danger-state canary layer for existing growth holdings.