OPEN-SOURCE SCRIPT
Magnet Compass Phase 3
MAGNET COMPASS
An intraday scalping system for NQ, ES, and GC futures, designed for a 30-second chart. It identifies 15-minute "magnet" candles at key session times — moments when one extreme of the candle has a measurable empirical bias to get tagged within the next 15 minutes — and arms trades only when live conditions agree with that bias.
═══════════════════════════════
THE IDEA IN ONE LINE
At each 15-minute close in a known magnet bucket, the indicator checks whether live features (close location, VWAP side, HTF trend) agree with the historical first-hit bias for that bucket. If they agree and the close sits in the mid-band of its range, it arms a trade toward the bias-side extreme.
═══════════════════════════════
WHY TWO LAYERS (PRIOR + LIVE)
The historical bias alone gives a 5–19 percentage-point edge on which extreme gets tagged first — real, but too small to trade on. A live-feature filter that vetoes the days when current conditions argue the other way is what turns the prior into a tradable setup.
From the underlying feature analysis:
Close location only matters in the middle of the range. At the extremes it's geometric noise ("close at the high → high gets hit trivially").
VWAP side and the bucket prior carry the real signal. HTF EMA trend is near-noise but kept as a tiebreaker.
So an entry fires only when: armed bucket + prior agrees with live score + close in mid-band + |score| ≥ 2.
═══════════════════════════════
SYMBOL COVERAGE
Auto-detects from chart symbol — no input changes needed.
NQ1! — 16 buckets (London 03:45, RTH 09:30–14:00, Asia 20:00–21:15)
ES1! — 19 buckets (London 03:00–04:00, RTH 09:30–14:00, Asia 20:00–21:15)
GC1! — 12 buckets (RTH 09:30–15:00, no London/Asia)
All times ET.
═══════════════════════════════
HOW TO READ THE CHART
Every 15 minutes an info label appears on the magnet candle:
Bkt 11:30 AM ARM ↑HIGH
Prior HIGH +19.3 | clLoc 0.42 | vwap +1 | htf +1 | score +4
Bkt — the 15-minute close time
ARM / SKIP — armed direction, or skip reason
Prior — bias side and signed edge (percentage points)
clLoc — close location in the candle (0 = low, 1 = high). Tradable band is 0.30–0.70.
vwap / htf — feature scores (−1, 0, or +1, signed toward HIGH-first)
score — sum of features. Need |score| ≥ 2 and same sign as prior to arm.
Lines drawn at each armed window:
Aqua solid — magnet high and low (anchored to the actual bars that made the extremes)
Gray dashed — midline (50% of magnet range)
On entry: yellow entry, lime target, red stop
WIN / LOSS labels at exit
The execution window is background-shaded only when the system armed for that bucket.
═══════════════════════════════
ARM CONDITIONS
All must hold at the confirmed 15-minute close:
Bucket has a non-zero prior
Prior and live features agree (same sign)
clLoc is between 0.30 and 0.70
|live score| ≥ 2
If "NY only" mode is on, bucket falls between 09:30 and 15:45 ET
═══════════════════════════════
ENTRY AND EXIT
Entry fires on the first confirmed 30-second bar after the 15-minute close (~30 seconds after). Skipped if that bar already wicks the bias-side extreme.
Stop styles (5 options):
Opposite extreme (default) — widest stops, ≈1:1 R:R, highest win-rate
Midline — half the loss size, but more stop-outs
Fixed ticks — entry ± N ticks
ATR — k × 15-minute ATR
Nearest unswept swing — closest swing past entry that hasn't been violated since forming; falls back to opposite extreme if none available
Target: bias-side magnet extreme by default, or midline (toggle).
Window close: if the trade is still open at the next 15-minute close, it's classified by the last 30-second close vs entry. Past entry = WIN, otherwise LOSS. Ties count as LOSS.
═══════════════════════════════
SCALP MODE (SETUP A — SWEEP FADE)
Optional toggle that flips the model for fade-style scalps. Instead of trading toward the bias-side magnet, it fades it: bias HIGH → resting SHORT limit at magnet high + buffer, targeting the midline. Fixed point-based stops/targets (default 12/12). Optional half-off at 1R, breakeven shift, and runner scale.
Backed by an extended-stat finding that the midpoint mean-reverts ~79% of the time after the magnet candle.
═══════════════════════════════
NON-REPAINTING
Uses request.security with [1] and lookahead_on — the canonical non-repainting pattern for HTF data. All state changes are gated to barstate.isconfirmed. Signals are identical in historical and real-time/replay.
Trade-off: entries fire on the first confirmed 30-second bar after the 15-minute close (~30s later) rather than instantly.
═══════════════════════════════
CUSTOMIZATION
Every visible element has its own color picker (magnet lines, midline, entry/stop/target, ARM/SKIP backgrounds, label text, table colors).
Both tables — live stats and reference bucket list — have 8-position placement dropdowns.
Toggles for every visual: lines, info labels, WIN/LOSS labels, window shading, both tables.
═══════════════════════════════
HONEST CAVEATS
This is a win-rate-dependent system with ~1:1 R:R, not a payoff-dependent one. A small drop in win-rate kills the edge.
The leaderboard priors are snapshots from a specific data window. They need periodic refresh as more data accumulates.
Window-close reclassification slightly understates loss magnitude when the exit lands near entry (fees and slippage not modeled).
Forward-test on sim until you've logged 50+ resolved trades. Small samples don't tell you much.
═══════════════════════════════
DISCLAIMER
For educational and research purposes only. Not financial advice. Past performance of any backtested setup does not guarantee future results. Trade your own account.
An intraday scalping system for NQ, ES, and GC futures, designed for a 30-second chart. It identifies 15-minute "magnet" candles at key session times — moments when one extreme of the candle has a measurable empirical bias to get tagged within the next 15 minutes — and arms trades only when live conditions agree with that bias.
═══════════════════════════════
THE IDEA IN ONE LINE
At each 15-minute close in a known magnet bucket, the indicator checks whether live features (close location, VWAP side, HTF trend) agree with the historical first-hit bias for that bucket. If they agree and the close sits in the mid-band of its range, it arms a trade toward the bias-side extreme.
═══════════════════════════════
WHY TWO LAYERS (PRIOR + LIVE)
The historical bias alone gives a 5–19 percentage-point edge on which extreme gets tagged first — real, but too small to trade on. A live-feature filter that vetoes the days when current conditions argue the other way is what turns the prior into a tradable setup.
From the underlying feature analysis:
Close location only matters in the middle of the range. At the extremes it's geometric noise ("close at the high → high gets hit trivially").
VWAP side and the bucket prior carry the real signal. HTF EMA trend is near-noise but kept as a tiebreaker.
So an entry fires only when: armed bucket + prior agrees with live score + close in mid-band + |score| ≥ 2.
═══════════════════════════════
SYMBOL COVERAGE
Auto-detects from chart symbol — no input changes needed.
NQ1! — 16 buckets (London 03:45, RTH 09:30–14:00, Asia 20:00–21:15)
ES1! — 19 buckets (London 03:00–04:00, RTH 09:30–14:00, Asia 20:00–21:15)
GC1! — 12 buckets (RTH 09:30–15:00, no London/Asia)
All times ET.
═══════════════════════════════
HOW TO READ THE CHART
Every 15 minutes an info label appears on the magnet candle:
Bkt 11:30 AM ARM ↑HIGH
Prior HIGH +19.3 | clLoc 0.42 | vwap +1 | htf +1 | score +4
Bkt — the 15-minute close time
ARM / SKIP — armed direction, or skip reason
Prior — bias side and signed edge (percentage points)
clLoc — close location in the candle (0 = low, 1 = high). Tradable band is 0.30–0.70.
vwap / htf — feature scores (−1, 0, or +1, signed toward HIGH-first)
score — sum of features. Need |score| ≥ 2 and same sign as prior to arm.
Lines drawn at each armed window:
Aqua solid — magnet high and low (anchored to the actual bars that made the extremes)
Gray dashed — midline (50% of magnet range)
On entry: yellow entry, lime target, red stop
WIN / LOSS labels at exit
The execution window is background-shaded only when the system armed for that bucket.
═══════════════════════════════
ARM CONDITIONS
All must hold at the confirmed 15-minute close:
Bucket has a non-zero prior
Prior and live features agree (same sign)
clLoc is between 0.30 and 0.70
|live score| ≥ 2
If "NY only" mode is on, bucket falls between 09:30 and 15:45 ET
═══════════════════════════════
ENTRY AND EXIT
Entry fires on the first confirmed 30-second bar after the 15-minute close (~30 seconds after). Skipped if that bar already wicks the bias-side extreme.
Stop styles (5 options):
Opposite extreme (default) — widest stops, ≈1:1 R:R, highest win-rate
Midline — half the loss size, but more stop-outs
Fixed ticks — entry ± N ticks
ATR — k × 15-minute ATR
Nearest unswept swing — closest swing past entry that hasn't been violated since forming; falls back to opposite extreme if none available
Target: bias-side magnet extreme by default, or midline (toggle).
Window close: if the trade is still open at the next 15-minute close, it's classified by the last 30-second close vs entry. Past entry = WIN, otherwise LOSS. Ties count as LOSS.
═══════════════════════════════
SCALP MODE (SETUP A — SWEEP FADE)
Optional toggle that flips the model for fade-style scalps. Instead of trading toward the bias-side magnet, it fades it: bias HIGH → resting SHORT limit at magnet high + buffer, targeting the midline. Fixed point-based stops/targets (default 12/12). Optional half-off at 1R, breakeven shift, and runner scale.
Backed by an extended-stat finding that the midpoint mean-reverts ~79% of the time after the magnet candle.
═══════════════════════════════
NON-REPAINTING
Uses request.security with [1] and lookahead_on — the canonical non-repainting pattern for HTF data. All state changes are gated to barstate.isconfirmed. Signals are identical in historical and real-time/replay.
Trade-off: entries fire on the first confirmed 30-second bar after the 15-minute close (~30s later) rather than instantly.
═══════════════════════════════
CUSTOMIZATION
Every visible element has its own color picker (magnet lines, midline, entry/stop/target, ARM/SKIP backgrounds, label text, table colors).
Both tables — live stats and reference bucket list — have 8-position placement dropdowns.
Toggles for every visual: lines, info labels, WIN/LOSS labels, window shading, both tables.
═══════════════════════════════
HONEST CAVEATS
This is a win-rate-dependent system with ~1:1 R:R, not a payoff-dependent one. A small drop in win-rate kills the edge.
The leaderboard priors are snapshots from a specific data window. They need periodic refresh as more data accumulates.
Window-close reclassification slightly understates loss magnitude when the exit lands near entry (fees and slippage not modeled).
Forward-test on sim until you've logged 50+ resolved trades. Small samples don't tell you much.
═══════════════════════════════
DISCLAIMER
For educational and research purposes only. Not financial advice. Past performance of any backtested setup does not guarantee future results. Trade your own account.
Skrip open-source
Dengan semangat TradingView yang sesungguhnya, pembuat skrip ini telah menjadikannya sebagai sumber terbuka, sehingga para trader dapat meninjau dan memverifikasi fungsinya. Salut untuk penulisnya! Meskipun Anda dapat menggunakannya secara gratis, perlu diingat bahwa penerbitan ulang kode ini tunduk pada Tata Tertib kami.
Pernyataan Penyangkalan
Informasi dan publikasi ini tidak dimaksudkan, dan bukan merupakan, saran atau rekomendasi keuangan, investasi, trading, atau jenis lainnya yang diberikan atau didukung oleh TradingView. Baca selengkapnya di Ketentuan Penggunaan.
Skrip open-source
Dengan semangat TradingView yang sesungguhnya, pembuat skrip ini telah menjadikannya sebagai sumber terbuka, sehingga para trader dapat meninjau dan memverifikasi fungsinya. Salut untuk penulisnya! Meskipun Anda dapat menggunakannya secara gratis, perlu diingat bahwa penerbitan ulang kode ini tunduk pada Tata Tertib kami.
Pernyataan Penyangkalan
Informasi dan publikasi ini tidak dimaksudkan, dan bukan merupakan, saran atau rekomendasi keuangan, investasi, trading, atau jenis lainnya yang diberikan atau didukung oleh TradingView. Baca selengkapnya di Ketentuan Penggunaan.