ChartChest

Adversarial Market Intelligence

Institutions run scripts.
We read them.

Every chart is a story institutions are writing in real time — accumulation, shakeouts, squeezes, distribution. We decode which script is running, how far along it is, and the role retail is being cast in — before the exit. Validated against 10,000+ live setups with 0.83 walk-forward CV AUC across five independent market regimes.

No account required  ·  Free analysis  ·  Stocks & crypto

0.83 Signal Ranking Accuracy Walk-forward CV AUC · 5 independent regimes
10,335 Live Setups Trained On Sep 2024 → Apr 2026 · no future leakage
15 Detection Engines Each independently verifiable · fused by X-Brain
+27% Ranking Accuracy Gained v1 (0.65) → current (0.83) · 8 months of evolution

Dual-Brain Architecture

Two adversarial intelligences.
One unified signal.

When manipulation signals and quantitative signals conflict, most platforms average them into noise. Ours knows which brain to trust — and when their disagreement is the signal itself.

Institutional Playbook Decoder

Every institutional move is a screenplay.
We give you the page number.

ChartChest pattern-matches multi-bar OHLCV narratives against five canonical institutional playbooks — identifying which script is active and who is designed to lose.

PLAYBOOK 01

The Load

Wyckoff Accumulation

Silent absorption of supply. Institutions quietly buy while price appears dead — before a controlled markup phase begins.

VICTIM: uninformed sellers

PLAYBOOK 02

The Shakeout

Manufactured Weakness

Deliberate selling to trigger retail stop-losses below support — then reversing to take the cleared long position.

VICTIM: stopped-out longs

PLAYBOOK 03

The Squeeze

Engineered Cascade

Gamma or short-squeeze mechanics are deliberately triggered. Forced covering creates a momentum cascade with predictable fuel.

VICTIM: short sellers

PLAYBOOK 04

The Distribution

Bag Passing

Offloading inventory into retail FOMO at peak narrative. Price holds while institutions exit through carefully managed supply.

VICTIM: bag holders

PLAYBOOK 05

The Battlefield

Desk vs Desk Warfare

Multiple institutional desks in direct conflict. Elevated volatility, two-sided prints, and real directional pressure from the winning desk.

VICTIM: uninformed participants


Architecture

15 engines. One read.

Independent layers of intelligence — each verifiable in isolation — fused into a single arbitrated conviction output.

MARKET STRUCTURE
ICT / Smart Money Concepts
Order Blocks Fair Value Gaps Liquidity Pools BOS / CHoCH AMD Cycles Swing Structure
DARK POOL INTEL
Off-exchange forensics
Absorption scoring Block inference VWAP bounce detection Options sweep Spread compression
CHESSMASTER
Adversarial desk modeling
Desk correlation Spoof detection Game state mapping Next-move forecast Silent Assassin mode
ML CORE
Per-symbol LightGBM
182-feature engineering Walk-forward CV Isotonic calibration Conformal intervals IC-adaptive weighting
REGIME + MACRO
Market environment backbone
VIX term structure Yield curve Credit stress Market Energy Index 6 energy regimes
SIGNAL ARBITRATION
6-engine evidence vote
ELITE ≥ 0.72 STRONG ≥ 0.58 STANDARD ≥ 0.42 Self-updating weights Thesis Writer (LLM)

Strategy Taxonomy

The AI classifies every setup into
how institutions actually trade.

Not just bullish or bearish. ChartChest labels each signal with the strategic archetype — so you know the expected hold, the mechanism, and the risk profile before you decide anything.

🏦

FOLLOW_THE_MONEY

Ride institutional accumulation

Dark pool flow is visible. Institutional absorption detected. Align with the direction of large capital — hold while the smart money holds.

TYPICAL HOLD: 3–7 days

ACCUMULATION_PLAY

Wyckoff Phase B/C base breakout

Spring, last-point-of-support, or base under resistance. Pre-load scenario dominant. Wait for the markup trigger — the setup has runway.

TYPICAL HOLD: 5–21 days

SQUEEZE_CAPTURE

Forced covering cascade

Gamma or short-squeeze mechanics are active. Dealer hedging or short-covering will force price. Fast momentum — enter early, exit before exhaustion.

TYPICAL HOLD: 0.5–3 days

🔄

COUNTER_MANIPULATION

Fade the exhausted manipulation

A pump-and-dump, bear raid, or stop hunt has reached exhaustion phase. The manipulators are done. Fade the move at the reversal point.

TYPICAL HOLD: 1–5 days

🌍

REGIME_ROTATION

Follow the macro money rotation

ETF flow and cluster intelligence confirm sector rotation. Macro and sector forces dominate. Longer duration, lower churn, higher conviction hold.

TYPICAL HOLD: 10–30 days

🧠

BANDIT LEARNER

Thompson Sampling per context

A Bayesian bandit learns which archetype wins in each combination of regime × market energy × manipulation phase — and routes accordingly.

UPDATES: every resolved outcome


Live Scanner

The ranked market.
Updated every session.

See every covered symbol sorted by active playbook tier, conviction score, and expected move — without opening a single chart.

Conviction-tiered ELITE → STRONG → STANDARD — only signals where the evidence weight crosses threshold.
Archetype-labeled Every row shows which institutional strategy the AI has identified as active.
Stocks and crypto unified Both asset classes through the same playbook decoder and signal arbitration pipeline.
One click to full thesis Open any signal for the full AI-generated bull/bear debate and conviction breakdown.

Validated Model Evolution

The model doesn't just claim accuracy.
It earns it by regime.

Every version of the model is walk-forward cross-validated — trained on the past, tested strictly on the future. AUC measures the model's ability to rank the best institutional setups above the noise, across five independent time regimes from 2024 to 2026.

We publish the numbers raw. v1 started at 0.65, barely above random for some folds. The current model sits at 0.83 — consistent across all five regimes with no holdout leakage. The improvement came from new data channels (dark pool forensics, macro emotion signals, playbook sequencing) not from overfitting to recent data.

X-Brain CV AUC  ·  walk-forward · embargo-protected
v1 · Sep 2024 0.6459
v3 · Oct 2024 0.6811
+35 bps
v6 · Jan 2025 0.7679
+87 bps
v9 · Apr 2025 0.8150
+47 bps
v11 · Feb 2026 0.8184
+34 bps
CURRENT · May 2026 0.8254
+7 bps

Adaptive Intelligence

The model doesn't stay still.
It counter-evolves.

Institutions adapt their scripts when patterns get spotted. Our evolution loop catches that. When a feature class degrades in live performance, the system retrains — not on the old regime, but on the freshest 19 months of adversarial behavior.

v1 (Sep 2024) struggled with just 35 high-conviction signals at 0.65 AUC. Systematic retraining — new dark pool forensics, playbook sequencing, and cross-brain calibration — compounded that to 0.83 AUC at 120× the signal volume. Each retrain cycle adds a new intelligence layer rather than just re-tuning the old one.

core/evolve.py  ·  autonomous loop
01
SIGNAL_TRACKING
Log current signals, resolve past outcomes against real close prices
02
ERA_DECAY_AUDIT
Track per-feature AUC across 4 rolling eras — flag any degrading signal class
03
BRAIN_RETRAIN
Retrain manipulation + structure brains on aligned labels with the freshest data
04
REASONER_RECAL
Recalibrate 15 detection engines — normalize gates against live Q95 distributions
05
HARDENED_CV
5-fold walk-forward benchmark with 30-day embargo — reject if AUC regresses
06
CHAMPION_GATE
New model only promotes if it beats prior champion on out-of-sample fold primacy

See which playbook
is running right now.

Type any ticker. Get the full institutional read. Free, no account.

Open the App →

Stocks & crypto  ·  No account required  ·  3 free analyses daily

AI signals for every major market

ChartChest covers 100+ stocks and crypto with AI breakout scanning, manipulation detection, and X-Brain signals. Each symbol has a dedicated analysis page.

Trading & Crypto Proxies
ETFs & Macro
Browse all covered markets →
ChartChest is an informational research tool and does not constitute financial advice. Past signal performance is not a guarantee of future results. Always conduct your own due diligence before making any investment decision.