# Screener Playbook β€” Valuations PP β†’ Buffer β†’ EOD **The funnel logic:** the two screeners answer different questions. Don't ask either to do the other's job. | Stage | Script | Question it answers | Cadence | |---|---|---|---| | 1 Β· Discovery | πŸ’° Valuations Screener πŸ’° (PP v4.1) | *"What deserves my attention?"* β€” cheap + waking up | Weekend + midweek | | 2 Β· Timing | EOD Screener | *"Is today the day, and at what stop?"* | Daily after close | | Buffer | TV watchlist "BUFFER" | Holding pen between the two | Continuous | A name can sit in the buffer for weeks. PP finds it early; EOD tells you when it's actionable. That gap **is** the edge β€” you're pre-positioned on names before the breakout, instead of chasing scans on the day. --- ## Setup (both screens) - **Timeframe: 1D.** Pocket pivots are defined on daily data; other TFs are meaningless here. - **Run AFTER market close.** On the live bar, volume is partial β€” pivots only turn true late in the session and crosses flicker until the bar confirms. - **Universe for Stage 1:** all NSE equity, or your preferred cap filter. Add a liquidity floor early β€” see "Liquidity gate" below. - **Universe for Stage 2:** the BUFFER watchlist only. - **Known limit:** PineScreener loads only ~500 bars, so percentile bands use ~2 years, not 5. Treat percentile columns as directional, not absolute. - **Column direction (memorise these two):** β‘  Valuation Score = price Γ· fair value β†’ **below 1 = cheap**. β‘’ Valuation Zone: negative = cheap, βˆ’2 = below the low band, +2 = above the high band. --- # STAGE 1 β€” The scan menu Run these as separate saved screens. Each answers a different question; **overlap between them is a strength signal**, not a duplicate. ## Family A β€” Ready now (value + volume + trend aligned) These are most likely to pass the EOD gates soon. Highest priority into the buffer. **A1 Β· Flagship conviction** ``` πŸš€ Accumulation Signal = 1 ``` Pivot + value zone + ADX turning up, all at once. Usually the shortest list β€” take every name. **A2 Β· Strong conviction (fundamentals-confirmed)** ``` πŸ”₯ Strong Conviction = 1 ``` Undervalued + good data + pocket pivot + up-trend. **A3 Β· Score-ranked** ``` β˜… Conviction Score > 70 β†’ sort β˜… descending ``` Use when A1/A2 return too few. Lower to 60 in thin markets, never below 50. **A4 Β· Value breakout** ``` πŸ’₯ Value Breakout = 1 ``` Pivot while undervalued in a confirmed uptrend β€” the cleanest single-column scan. ## Family B β€” Early / building (buffer for weeks) Cheap and turning, but not yet technically ready. This is your **buffer's core inventory**. **B1 Β· Cheap and starting to move** ``` 🟩 Pocket Pivot >= 1 AND 🎯 Near Low FV Band = 1 ``` **B2 Β· Value + technical quality** ``` 🟩 Pocket Pivot >= 1 AND β‘  Valuation Score < 0.9 AND Tech Score >= 2 ``` **B3 Β· Early trend change at value (ADX)** ``` πŸ”Ό Trend Building = 1 AND β‘  Valuation Score < 1.0 ``` **B4 Β· Pattern signals** β†’ these deserve their own section; see **"The pattern engine"** below. ## Family C β€” Deep value watch (monthly review, low urgency) No technical confirmation yet. These wait for stage 2 to arrive; some never do. **C1 Β· Fundamentally cheap, data-backed** ``` β‘  Valuation Score < 0.9 AND Data Quality >= 3 AND Coverage % >= 75 ``` **Coverage %** is the v4.1 honesty column: it's the share of the fair-value model actually populated. Data Quality counts metrics; Coverage weights them. `DQ 3 + Coverage 40%` = a "fair value" resting on one ratio β€” skip it. **C2 Β· Margin-of-safety screen** ``` Margin of Safety % > 20 AND Coverage % >= 75 β†’ sort MoS descending ``` **C3 Β· Percentile cheapness** ``` β‘‘ Upside to FV % > 20 AND ⑨ Avg Percentile < 30 ``` ## Family D β€” Specialty **D1 Β· Bottom-fishing reversal** (earlier, higher risk β€” always pair with value) ``` 🟩 Pocket Pivot >= 2 AND 🎯 Near Low FV Band = 1 ``` Column encoding: `1 / 1.5` = standard pivot in the 10MA pocket Β· `2 / 2.5` = bottom-fish reclaiming the 50MA. The `.5` tier means pivot volume β‰₯ 2Γ— the largest down-day volume β€” **sort `Vol / MaxDownVol` descending to rank pivots by force.** **D2 Β· Crash reversal** ``` πŸ”„ 30% Fall Reversal + PP @ FV = 1 ``` **D3 Β· State alerts** (always-true conditions, scannable any day) ``` 🟒 Deep Value Β· πŸ’Ž Deep Value Opportunity Β· 🚨 Extreme Overvaluation (avoid-list) ``` --- # The pattern engine β€” base breakout Β· FVR rebound Β· fall reversal These are the most demanding signals in the script: each stacks **valuation + structure + trend + volume** into one flag. They're also the ones worth understanding precisely, because each maps to a different market situation. ## πŸ“¦ Long Base Breakout Β· πŸš€ Long Base Breakout + Pocket Pivot **What it detects:** a Darvas/Weinstein base break **with a valuation floor**. Every condition must hold: - **Tight base** β€” the last 30 bars span ≀ 18% high-to-low - **Base sits at or above value** β€” close β‰₯ 97% of the low fair-value band. *This is the clever part:* it rejects bases built below fair value (broken stocks going sideways) and keeps bases built on top of value. - **Clean break** β€” close > base high **+ 1.5% buffer** (kills marginal pokes) - Bullish bar, DI+ > DIβˆ’, and trend up or building **Why it works for you:** this is the closest thing in the suite to a pure stage-2 entry β€” a tight base resolving upward on a name that isn't expensive. Expect these to convert to EOD `BUY` fastest (they map to EOD **SetupCode 1** base-breakout / **4** resistance-breakout). **Use:** `πŸš€ Long Base Breakout + Pocket Pivot` as your primary β€” the volume-confirmed version. The plain πŸ“¦ is for a wider net when the market is quiet. ## ↗️ FVR Pullback Rebound Β· 🟒 FVR Rebound + Pocket Pivot **What it detects:** price dipped into the fair-value floor and is **bouncing with a structural reclaim** β€” a buy-the-pullback-to-value, explicitly not a falling knife. - **Recent value test** β€” within the last 12 bars, price either touched within 3% of the low band, or traded in the bottom 35% of the lowβ†’fair range - **Rebounding now** β€” bullish bar, above prior close, back above the fast MA, DI+ > DIβˆ’ - **Plus a reclaim** β€” MA crossover, *or* takes out the recent swing high, *or* crosses back above the low band - **Not extended** β€” still below the high band, so the move has room **Why it works:** the reclaim requirement is what separates this from catching knives. Price must *prove* the bounce by reclaiming something structural, not merely stop falling. **Use:** `🟒 FVR Rebound + Pocket Pivot`. Buffer these β€” they're often early, and EOD will tell you when stage 2 actually arrives. ## πŸ”„ 30% Fall Reversal + PP @ FV **The most stringent signal in the script** β€” five independent confirmations: 1. **Crashed** β€” down β‰₯ 30% from the 120-bar peak 2. **Landed on value** β€” now within 5% of the low band *or* the median fair value 3. **Pocket pivot** β€” the volume signature 4. **Reversal bar** β€” up close, above prior close, closing in the **top 65%** of its range 5. **Reclaim or turn** β€” takes out the 10-bar swing high, or the trend is genuinely turning **Why it works:** a crash that has *landed on fair value and turned on volume* is a fundamentally different animal from a crash still falling. That said, this is the earliest and highest-risk of the three β€” the wound is fresh. **Use:** buffer, don't rush. Let EOD confirm stage 2 before committing, or take a starter position with a stop under the reversal bar's low. ## πŸ”₯ Pattern Conviction β€” the umbrella `(FVR Rebound + PP) OR (Base Breakout + PP)`, plus minimum data and a trend that's up or building. **One screen that catches both patterns, already volume- and data-confirmed.** If you only run one pattern scan, run this. ## Tuning them (Inputs β†’ "Pattern Signals" / "Reversal Patterns") | Input | Default | Turn it… | |---|---|---| | `Require Pocket Pivot For Main Pattern Flags` | off | **ON** to make the plain πŸ“¦/↗️ columns volume-strict β€” then you don't need the separate +PP columns at all | | `Base Tightness %` | 18 | ↓ 10–12 for VCP-tight bases Β· ↑ 22–25 for volatile smallcaps | | `Base Breakout Buffer %` | 1.5 | ↑ 2–3 to reject marginal breaks Β· ↓ for earlier entries | | `Long Base Lookback` | 30 | ↑ 50–60 for genuinely long bases (fewer, better) | | `Pullback Lookback Bars` | 12 | ↑ if you want rebounds from older dips to still count | | `FVR Touch Threshold %` | 3.0 | ↑ 5 if the low band is rarely touched cleanly | | `Minimum Prior Fall %` | 30 | ↑ 40–50 for genuine washouts only | | `FV Zone Tolerance %` | 5.0 | how close to fair value the reversal must land | ## Where each lands in the funnel | Signal | Urgency | Where it goes | |---|---|---| | πŸš€ Base Breakout + PP | **Highest** β€” often EOD-actionable within days | Buffer, check EOD daily | | 🟒 FVR Rebound + PP | Medium β€” early by design | Buffer, wait for EOD stage 2 | | πŸ”„ 30% Fall Reversal | Lowest β€” earliest, highest risk | Buffer, starter only or wait | | πŸ”₯ Pattern Conviction | Covers the first two | Run this if you run only one | --- # STAGE 1 β€” Scanning by alertcondition PineScreener evaluates **only the last bar**. A raw event alert would therefore be true for exactly one day β€” scan on Saturday and you'd see nothing that fired Monday–Thursday. This script solves that with a scan window, and that single input is the most important dial in your workflow. ## The two alert types **STATE alerts** β€” true for as long as the condition holds. Scan any day, no window needed. ``` 🟒 Deep Value Β· πŸ’Ž Deep Value Opportunity Β· 🚨 Extreme Overvaluation ``` **EVENT alerts** β€” the other 18. Each is wrapped so it stays true for `Event Scan Window` bars after the event (`ta.barssince(cond) < window`). ## The Event Scan Window dial | Setting | Means | Use for | |---|---|---| | **5** (default) | "signaled this week" | Weekend scanning β€” the right default for you | | **1** | Event bar only | Strict "today only" scan; also correct for real live alerts | | **10** | "signaled this fortnight" | Slower cadence, or widening a thin list | Saved screens store their own inputs β€” so keep one screen at 5 for the weekend sweep and another at 1 for a daily "fresh today" pass. ## Alert recipes by family **Family A β€” ready now** ``` πŸš€ Accumulation Signal Β· πŸ”₯ Strong Conviction Buy Β· πŸ’₯ Value Breakout ``` **Family B β€” early / building** ``` 🟩 Pocket Pivot Β· πŸ”Ό Value Trend Turn Β· πŸ”„ Bullish Trend Change Β· πŸ“Ά Up-Trend Strengthening 🟒 FVR Rebound + Pocket Pivot Β· πŸ“¦ Long Base Breakout Β· πŸš€ Long Base Breakout + Pocket Pivot πŸ”₯ Pattern Conviction Β· πŸ”„ 30% Fall Reversal + PP @ FV ``` **Family C β€” deep value (state, scan anytime)** ``` 🟒 Deep Value Β· πŸ’Ž Deep Value Opportunity ``` ## The transition alerts β€” your best buffer-fillers These fire the moment a name *enters* your zone, which is exactly what a buffer wants: | Alert | What just happened | Action | |---|---|---| | **πŸ“ Became Undervalued** | Crossed below fair value | **The single best buffer-filler** β€” new candidates, weekly | | **πŸ“ Dropped Below Low Band** | Now below the low band | Deep-value bucket; needs a pivot before it matters | | **πŸ“ Reclaimed Fair Value** | Re-rating starting from below | High-value: the market is beginning to agree | ## Exit / risk scans β€” run on your HOLDINGS watchlist The alerts aren't only for discovery. Point these at what you own: ``` 🚨 Extreme Overvaluation (state) Β· πŸ“ Became Overvalued Β· πŸ“ Lost Fair Value Β· πŸ”» Bearish Trend Change ``` ## Stacking alerts with columns β€” the highest-value combination An alert says *"this happened"*; a column says *"and it's worth acting on"*. Combine them: ``` 🟩 Pocket Pivot = true AND β‘  Valuation Score < 0.9 AND Coverage % >= 75 πŸš€ Accumulation Signal = true AND Vol / MaxDownVol >= 2 πŸ“ Became Undervalued = true AND Data Quality >= 3 AND Coverage % >= 75 ``` **To OR several alerts:** run them as separate saved screens and merge into the buffer. That's not just a workaround β€” it preserves which family sourced each name, which is how you learn what actually pays. --- ## Triage β†’ buffer For every hit, glance at three columns before adding: 1. **Coverage % β‰₯ 75** β€” else the valuation is thin air. (< 50 = reject outright.) 2. **β‘’ Valuation Zone** β€” is it cheap because it's *cheap*, or cheap because it's *broken*? Zone βˆ’2 with no pivot is a falling knife. 3. **Vol / MaxDownVol** β€” pivot force. Below 1.5 is a weak pivot. Tag the name in the buffer with its source family (A/B/C/D). When it eventually triggers in EOD, you'll know whether your discovery edge was value, volume, or pattern β€” that's how the funnel teaches you which family actually pays. **Buffer hygiene:** cap it at ~40–60 names. Drop anything that has sat 3 months without an EOD ARM, or where β‘  has drifted above 1.2 (the thesis got priced in while you waited). --- # STAGE 2 β€” EOD Screener on the buffer Universe = BUFFER watchlist. Run daily after close. ## ⚠️ EOD alerts work differently from PP alerts This asymmetry will bite you if you carry the PP habit across: | | PP Screener | EOD Screener | |---|---|---| | Alert type | **Windowed** β€” true for 5 bars after the event | **Edge-triggered** β€” true on the transition bar ONLY | | Weekend scan | Catches the whole week | Catches **only Friday's** transitions | | Best practice | Scan by alert | **Scan by column**, use alerts for "new today" | So on EOD, prefer the **columns**: `BUY = 1` and `ARM = 1` are level-based β€” they stay 1 for as long as the condition holds, which is what you want for a daily actionable list. Use EOD **alerts** for a different job β€” isolating what changed *today*, so you don't re-read names you already triaged: ``` βœ… BUY (full confirm) βŒ› ARM (setup coiling) Setup triggered Entered Stage 2 Quality score β‰₯ 5 ``` (`Stage-2 ignition` and `Volume breakout` are the two level-based exceptions β€” they behave like columns.) ## The daily read **Ready today** ``` BUY = 1 AND DataOK = 1 β†’ sort Score descending ``` **Setting up (tomorrow's candidates)** ``` ARM = 1 AND DataOK = 1 β†’ sort Score descending ``` ## Your discipline filters β€” the part that matters Add these to every EOD screen; they enforce your own rules mechanically: ``` SlClamped = 0 ← structural stop intact (NOT an arbitrary % cap) Stop_% <= 3 ← your hard rule RR >= 2 ← reward justifies the risk Turnover >= [floor] ← slippage protection ``` **`SlClamped` is the single most valuable column for you.** `1` means the structural stop was wider than your cap, so the script clamped it to a percentage β€” there is no real structure beneath that stop. A clamped stop is a fake stop. Filter them out. **Turnover floor:** set it to at least 100Γ— your intended position size in β‚Ή. With full-capital deployment, slippage on illiquid names is your largest unmodeled risk β€” this column exists specifically for that. ## Reading a candidate row | Column | What you're checking | |---|---| | `Stage` / `StageWeeks` | Stage 2, and **freshly** so β€” low StageWeeks = early in the advance, the highest-value Weinstein entries | | `MansRS_%` | Mansfield RS β€” positive and rising = leadership. Sort by it to rank a crowded BUY list | | `SetupCode` | 1 Base-breakout Β· 2 Bull-flag Β· 3 Double-bottom Β· 4 Resistance-breakout Β· 11 tight base forming Β· 12/13/16 pre-trigger states | | `Vol_x` | Volume expansion on the trigger β€” thin breakouts fail | | `Entry / Stop / TP1 / Qty` | The plan, pre-computed at your sizing | | `SugStop_% / SugRR` | The alternative if the structural stop is too tight to be safe | | `Ext_%` | Extension above the stage MA β€” high = chasing | | `NextRes` | Overhead supply β€” a wall right above kills the R:R | | `DataOK` | `0` = short history, gates ran on incomplete data. Never trade a `0`. | ## Ranking a crowded list Sort `Score` desc β†’ break ties with `MansRS_%` desc β†’ prefer low `StageWeeks` (fresh stage 2) and low `Ext_%` (not extended). Cross-reference back to the PP screener: if the name also shows `πŸš€ Accumulation Signal = 1`, both engines agree β€” that's your highest-conviction bucket. --- ## Weekly rhythm | When | Do | |---|---| | **Sat** | Run Family A + B fully (alert screens at **Event Scan Window = 5** = "signaled this week") + πŸ“ Became Undervalued; refresh buffer; drop stale names. Pairs with your iList weekly review. Also run the exit-scan set on holdings. | | **Sun** | Family C monthly-ish; note which buffer names are approaching stage 2. | | **Daily after close** | EOD on buffer β†’ BUY list (act) + ARM list (tomorrow's watch). | | **Midweek (Wed)** | Quick A1/A4 re-run to catch fresh pivots between weekends. | ## Pitfalls - **Intraday runs lie.** Partial volume breaks every volume-gated column in both screens β€” and pocket-pivot alerts only turn true late in the session. - **A blank alert scan usually means the window, not the market.** If an event-alert screen returns nothing, check `Event Scan Window` before concluding there are no setups β€” at 1, a Saturday scan only sees Friday. - **Don't scan EOD by alert on weekends.** Edge-triggered: you'd only catch Friday. Use `BUY = 1` / `ARM = 1` columns instead. - **Don't skip Coverage %.** The C1 fix means no-data names now show blank rather than a fake "fair value" β€” but a thin-coverage name still looks confident. Coverage is the tell. - **A clamped stop is not a stop.** `SlClamped = 1` β†’ the βœ‚ cap replaced real structure. - **`DataOK = 0` / blank Stage** = recent listing with insufficient history. The gates ran on `na`. - **Extension is measured against the 30-period stage MA** (v4.1 corrected this from the 10-EMA). It reads stricter than before β€” if your BUY list thins noticeably, loosen `maxExtPct` rather than ignoring the column. - **The buffer is the asset.** The scans are just how you fill it. Most of the compounding comes from being *already watching* a name on the day it breaks out.