🌸

Aoi Dreams • Long-Term Investment Research

Aoi-chan Investment Empire

Build slowly. Own wisely. Grow steadily. Empire View handles the daily tending; Investment Compass and Market Lab wait behind it when something deserves a closer look.

Educational use only. This tool does not provide individualized investment advice or predict future returns. Daily data can be delayed. Signals are evidence labels, not probabilities. Backtests are hypothetical and use simplified execution assumptions.

Stock data: none

Benchmark: none

Latest date: —

Data mode: daily / end-of-day

Dataset role: Development

🔑 Twelve Data Real Data Bridge API

Enter your Twelve Data API key once, then the Lab can request real daily OHLCV data for the stock and SPY directly. CSV drag-and-drop remains available as a fallback.

History

Twelve Data API Key

Remember key in this browser only

🔋 Daily API usage

Waiting for usage check

Daily usage not read yet

Daily Basic-plan quota window. Refreshes from Twelve Data /api_usage.

📅 Account usage

Not checked yet

Minute usage not read yet

Click refresh for Twelve Data's current usage windows.

Auto-refresh after real-data load (uses 1 credit) No usage check yet.

Ready. Add your Twelve Data API key, enter a ticker, then click Load Real Data.

📥 Drop Stock CSV Hereor click Import Stock CSV above.
Expected: Date, Open, High, Low, Close, Volume.

🌎 Drop Benchmark CSV HereUse SPY or another broad-market daily CSV.
Benchmark improves market-regime confidence.

The API key is sent only to Twelve Data when the Lab requests market data. The credit meter updates from Twelve Data response headers when available.
The Lab requests daily OHLCV for your ticker and SPY, then calculates all indicators locally in your browser.
If Twelve Data returns an account, coverage, or quota error, the exact provider message appears here.
The API key is never included in Analysis JSON or Experiment CSV exports.

Evidence Score

—/100

Load data to begin

Entry Quality

—/100

—

Risk

—

ATR pending

Confidence

—/100

—

🎯 Educational Signal

Waiting for market data

Position context

Market environmentUnknown

Last close—

Daily change—

Volume vs 20-day avg.—

🧠 Why the model thinks this

Evidence in favor

No analysis yet.

Concerns & decision gates

No analysis yet.

📈 Price + Moving Averages

Load at least 200 daily observations for the full model.

CloseSMA 20SMA 50SMA 200

📐 Technical Readings

SMA 20—

SMA 50—

SMA 200—

RSI 14—

MACD—

Signal line—

Histogram—

ATR 14—

Distance from SMA 50—

👑 Watchlist0 / 5

💰 Opportunity Reserve—

🌱 Add / Opportunity Alerts0

🚪 Review / Reduce Alerts0

Lunchtime rule: Empire View records the latest completed U.S. daily close. If Twelve Data includes today's still-forming daily candle while the market is open, the Watchlist ignores it for decisions. Intraday movement can be interesting, but it does not get to boss the investment process around.

📜 Watch History

One record per stock per completed market date. Updating again on the same date replaces that day's snapshot instead of creating duplicates.

Date	Ticker	Owned?	Decision	Price	OWN	ADD	REDUCE	FIT	Evidence	Entry Q	Market
No watch history yet.

🔬 Research Department

Market Lab

The Empire lives upstairs. This is where we tear ideas apart, test them on untouched data, and keep the machinery honest.

📊 Current Analysis

🧪 Backtesting & Validation

🧭 Point-in-Time & Entry/Exit Research

⚗️ Research Arenas

⚙️ Backstage

Data & Settings

Single-stock loading, Twelve Data credentials, CSV imports, usage controls, and maintenance tools live here so the Empire dashboard stays clean.

☁️ Portable Empire Backup

Recommended workflow: host the Empire on one Netlify URL, then use this backup file to move your watchlist, Watch History, portfolio settings, and research records between phone, home PC, and work PC. The Twelve Data API key is intentionally not included in the backup.

Suggested routine: Export a backup to Dropbox after meaningful changes, then import that file on any other device that should match the current Empire state.

📡 Single-Stock Data Loader

Use this when you want to load one company into Investment Compass or the Market Lab. Empire View uses its own five-stock Update Empire workflow.

🧰 Utilities

Clearing the watchlist removes the five active slots but preserves accumulated Watch History.

🏠 OWN • Ownership Quality

—/100

Load data to begin

Company-quality fundamentals remain separate from price-history quality.

🌱 ADD • Opportunity

—/100

—

Timing score, not a forecast of return.

🚪 REDUCE • Risk

—/100

—

High score means more reasons to review or protect the position.

🧺 PORTFOLIO FIT

—/100

Set portfolio inputs

Measures concentration capacity, not whether the stock will rise.

🧭 Current Investment Compass

Load real or demo data to begin.

Market-history quality—

Evidence confirmation age—

Portfolio capacity ceiling—

The Compass deliberately separates ownership, timing, sell risk, and portfolio fit. ADD is gated: at least 4/6 fundamental categories researched, Evidence ≥75 for 2 completed sessions, attractive timing, and available portfolio capacity are all required before the Lab can display an ADD candidate.

📚 Fundamental Snapshot Manual bridge

Until automatic point-in-time financial statements are wired into the standalone Lab, record the company research here. Unknown items are not scored.

Revenue trend

Earnings trend

Free cash flow

Debt burden

Valuation

Capital allocation

Fundamental completeness0 / 6

Fundamental quality score—

💰 Position + Portfolio Sandbox

Educational planning only. “Capacity ceiling” is the most additional dollars that fit under your chosen concentration limit, not a recommendation to invest that amount.

Average cost / share

Purchase date

Current position value

Total portfolio value

Available cash pool

Max one-stock position

Sector exposure now (%)

Current position concentration—

Profit-protection status—

🔮 Prediction Journal

Save what the Compass believed before the future happened. When later daily data is loaded, the Lab can fill actual 3-, 6-, and 12-month price outcomes for that snapshot.

Date	Ticker	Price	Compass Decision	OWN	ADD	REDUCE	FIT	3M Actual	6M Actual	12M Actual
No prediction snapshots saved yet.

📈 Trend

—/100

20/50/200-day alignment, price location, and moving-average slope.

⚡ Momentum

—/100

RSI, MACD line/signal, zero line, and histogram direction.

📊 Participation

—/100

Recent price direction paired with volume versus its 20-day average.

v0.3 boundary: fundamentals are deliberately not included in the automatic score yet. The first build tests whether our technical and market-regime logic behaves sensibly before we add point-in-time financial statements.

🧪 Transparent Backtest

Single-position, unlevered test. A signal based on a completed day executes at the next available open. v0.3 separates development from validation so we can stop teaching the model the answers to its own exam.

Dataset name

Dataset role

Development dataset: tuning is permitted, but results should not be treated as independent validation.

Frozen research benchmark

Entry Evidence ≥ 75 Entry Quality ≥ 90 Exit Evidence < 40 Max Hold 90 days 🧪 Development

Entry evidence ≥

Entry quality ≥

Exit evidence <

Max holding days

Trading cost each side

Slippage each side

Start capital

Action

💵 Return basis Price return Dividend data not loaded. Stock and strategy returns exclude dividends.

Strategy Return

—

Price return

Buy & Hold

—

Price return

Max Drawdown

—

Exposure

—

Trades—

Win rate—

Average winner—

Average loser—

Ending value—

Backtest period—

Return basisPrice return

Strategy dividends received / reinvested—

Buy & Hold dividends received / reinvested—

Run the test after loading stock data. Treat results as hypothetical laboratory output, never as a forecast.

🔎 Backtest Diagnostics

Expectancy / trade—

Profit factor—

Average hold—

Best / worst trade—

📈 Growth of $10,000 Equivalent

Run the backtest to compare strategy, buy & hold, and benchmark.

StrategyBuy & HoldBenchmark

🌊 Strategy Drawdown

Run the backtest to see peak-to-trough pressure.

Drawdown from prior strategy peak

🎚️ Entry Evidence Buckets

Buckets adapt to the tested entry threshold.

No diagnostics yet.

🎯 Entry Quality Buckets

Buckets adapt to the tested entry-quality threshold.

No diagnostics yet.

🌎 Market Regime at Entry

No diagnostics yet.

🚪 Exit Reasons

No diagnostics yet.

🧭 Experimental Trend Suitability

Descriptive only. This does not change the frozen strategy or filter trades. It asks whether the stock historically offered persistent, directional trends rather than repeated whipsaws.

Trend Suitability

—

Run backtest

No stock-character diagnosis yet.

60-day efficiency—

Bullish alignment—

20/50 crosses / yr—

200-day rising—

High-quality signal days with positive 20-day follow-through—

Median 20-day return after high-quality signal days—

🧠 Diagnostic Interpretation

Run the backtest and the lab will look for patterns rather than merely declaring victory or defeat.

🛤️ Trade Path Diagnostics

Descriptive only. This examines what price actually did after each executed entry. It does not alter the frozen 75 / 90 / 40 / 90 strategy.

Trade Path Quality

—

Run backtest

Run a backtest and the Lab will compare how far trades traveled in your favor versus against you, how quickly winners developed, and how concentrated the profits were.

Median MFE—

Median MAE—

Median Peak Capture—

Trades Reaching +5%—

Trades Reaching +10%—

Median Days to +5%—

Median Failure Speed—

Best 3 Share of Gains—

🟢 Winners

Median MFE—

Median MAE—

Median hold—

Median peak capture—

🔴 Losers

Median MFE before loss—

Median MAE—

Median hold—

Median days to -2%—

#	Entry → Exit	Return	MFE	MAE	Peak capture	Days to +5%	Days to +10%	Days to -2%
Run the backtest to inspect executed trade paths.

🧭 Threshold Sensitivity

The lab reruns the same data behind the scenes to find the next stricter setting that would actually alter the trade sequence.

Executed Evidence Range—

Executed Entry-Q Range—

Next stricter Evidence—

Next stricter Entry-Q—

Next stricter Exit—

Next shorter Hold—

Current trade signature—

Validation state—

📜 Trade Ledger

#	Entry	Exit	Days	Evidence	Entry Q.	Regime	Entry Price	Exit Price	Return	Reason
Run the backtest to populate the ledger.

🧬 Validation Lab

Tiny Samurai is our development dataset. The frozen configuration below is the configuration we carry into new, unseen datasets. Clean validation runs are saved locally in this browser and aggregated here.

Frozen research configuration

Entry ≥ 75 Entry Quality ≥ 90 Exit < 40 Hold ≤ 90 days

Validation discipline: Import a new dataset, mark it Validation, load the frozen settings, and run it once. If you inspect that result and then tune parameters on the same dataset, change its role to Development. v0.3.8.5 marks non-frozen validation runs as contaminated.

📚 Clean Validation Aggregate

Datasets0

Profitable—

Beat Buy/Hold—

Median Strategy—

Median Buy/Hold—

Median Gap—

Median Max DD—

Median PF—

Median Expectancy—

Median Exposure—

Trend Diagnostics—

Median Trend Score—

Trade Path Diagnostics—

Median Path Score—

🟢 Profitable + PF > 1

No cohort data yet.

🔴 Losing or PF ≤ 1

No cohort data yet.

No clean validation datasets have been recorded yet.

🧾 Saved Experiments

Dataset	Role	Status	Basis	Trend	Path	Entry	Entry Q.	Exit	Hold	Return	Buy/Hold	Max DD	PF	Trades
No experiments saved yet.

🗺️ Trend × Trade Path Research Map

This map does not change the frozen strategy. It places each clean validation dataset by Trend Suitability and Trade Path Quality, then compares the original eight-stock research cohort with future unseen stocks.

Research discipline: MSFT, AAPL, JPM, XOM, JNJ, CAT, PG, and VZ are the Research Cohort because the diagnostics were developed while examining them. Any different clean-validation stock becomes Forward Validation. Forward results are the real test of whether this map generalizes.

Research Cohort0/8

Forward Validation0

Research Trend ↔ Return—

Research Path ↔ Return—

Forward Profitable—

Forward Median Return—

Forward Median PF—

Forward Beat Buy/Hold—

Descriptive research map

Guide lines use the diagnostics' existing descriptive cutoffs: Trend-friendly ≥ 58 and Constructive Path ≥ 55. They were not optimized for stock returns.

Research cohort Forward validation Losing strategy result

🧭 Map Region Summary

🌿 Trend-friendly + Constructive Path

No datasets yet.

🛤️ Below Trend-friendly + Constructive Path

No datasets yet.

🌬️ Trend-friendly + Below Constructive Path

No datasets yet.

🪨 Below Both Guides

No datasets yet.

📚 Mapped Datasets

Dataset	Cohort	Trend	Path	Strategy	Buy/Hold	PF	Max DD	Region
No mapped datasets yet.

The original eight-stock map is exploratory. Do not tune the frozen strategy or diagnostic cutoffs to improve it. Add new unseen stocks and judge whether the relationships persist.

🧪 Forward Validation Report

This report grades only stocks outside the original eight-stock research cohort. The goal is not to make the strategy look good. The goal is to see which relationships survived contact with unseen stocks.

Forward sample

clean forward-validation stocks

Profitable—

Beat Buy/Hold—

Median Return—

Median PF—

Trend ↔ Return—

Path ↔ Return—

Trend ↔ PF—

Path ↔ PF—

Complete clean forward validations to generate the report.

🗺️ Forward Results by Map Region

No forward data yet.

⚖️ Forward Winners vs Losers

No forward data yet.

🧭 What Survived Forward Testing?

No forward data yet.

🧨 What Did Not Survive Cleanly?

No forward data yet.

📐 Map Coverage

Forward stocks in Trend-friendly + Constructive region—

Forward stocks below both guides—

Forward stocks in mixed regions—

No coverage assessment yet.

📋 Forward Validation Ledger

Stock	Trend	Path	Region	Strategy	Buy/Hold	Gap	PF	Max DD	Verdict
No forward-validation stocks recorded yet.

💾 Experiment Continuity

Export from one Lab version and import into the next if your browser isolates local HTML storage. The API key is never included in experiment files.

⏳ Point-in-Time Suitability Lab

This lab asks the harder question: what could we have known before the trade? Each executed entry receives a pre-entry Trend Suitability score calculated only from data available through the completed signal day.

No-lookahead rule: the pre-entry score uses up to the prior 504 trading sessions, ending on the day before execution. Early trades use all available history once at least 200 sessions exist. Future prices, future exits, MFE, MAE, and the eventual trade result are never used to calculate the pre-entry score.

Current dataset

—

trades with valid pre-entry suitability

Median Pre-Entry Score—

Pre-Entry ↔ Trade Return—

Pre-Entry ↔ MFE—

Winner vs Loser Median—

Run a clean backtest in v0.3.8.5 to populate the point-in-time study.

🎯 Pre-Entry Suitability Buckets

These are descriptive buckets, not trading gates. We are looking for monotonic behavior before inventing any rule.

Pre-Entry Score	Trades	Win Rate	Avg Return	PF	Median MFE	Median MAE
Run a backtest first.

🧠 What Was Knowable Before Entry?

No point-in-time data yet.

🚪 Exit Timing Audit

This does not change the sell rule. It measures what the existing exit retained or surrendered before the position was closed.

No exit audit yet.

🔎 Buy / Sell Research Clues

No clues yet.

📋 Trade-by-Trade Point-in-Time Ledger

#	Entry	Pre-Entry Trend	History Used	Return	MFE	MAE	Exit Evidence	Peak Giveback	Exit Reason
Run a v0.3.8.5 backtest to populate this ledger.

Important: even if high pre-entry scores look better, we will not turn them into a buy filter until the relationship is tested on trades/stocks that were not used to discover the cutoff. Likewise, an exit pattern is not a sell rule until it survives a separate untouched test.

🧺 Pooled Point-in-Time Trade Study

Instead of judging one stock at a time, this study pools individual executed trades across clean frozen runs. Research-cohort trades and forward-validation trades remain separate so we can see whether a pattern survives outside the stocks used to develop the idea.

Research boundary: the score buckets remain the existing descriptive bands <40, 40–49, 50–59, and 60+. v0.4.4 does not search for an optimal cutoff. A stock must be rerun in v0.4.4 (or later) once so its individual PIT trades can be stored for pooling.

Pooled clean PIT sample

0 trades

0 stocks with stored trade-level data

Research Trades—

Forward Trades—

PIT ↔ Return—

PIT ↔ MFE—

Winner vs Loser PIT—

Rerun clean frozen stocks in v0.4.4 to begin the pooled study.

🌐 All Pooled Trades

PIT	Trades	Win Rate	Avg Return	Median Return	PF	Median MFE	Median MAE	Avg Hold
No pooled trades yet.

🧪 Research Cohort Trades

PIT	Trades	Win Rate	Avg Return	PF	MFE	MAE
No research PIT trades stored yet.

🌱 Forward-Validation Trades

PIT	Trades	Win Rate	Avg Return	PF	MFE	MAE
No forward PIT trades stored yet.

🚧 Eligibility-Floor Stress Test

These comparisons use pre-existing descriptive boundaries only. They are not optimized filters and do not alter the strategy.

No pooled data yet.

🚪 Pooled Exit / Giveback Audit

No pooled data yet.

🧠 What the Pool Is Saying

No pooled data yet.

📚 PIT Source Coverage

Stock	Cohort	PIT Trades Stored	Median PIT	Strategy Return	PF	Last Captured
No v0.4.4 PIT trade data stored yet.

Next scientific step: if an eligibility pattern appears, freeze the interpretation first. Do not immediately modify 75 / 90 / 40 / 90. The candidate rule must then face untouched stocks or a later time period.

🧭 Entry Context + Exit Timing Study

This layer asks two different questions without changing the frozen strategy: what did we know before entry? and what observable deterioration appeared after a trade peaked but before the existing exit?

Entry boundary: every entry feature is calculated from the completed signal day or earlier. PIT trajectory compares the current point-in-time environment with its value 20 sessions earlier; Evidence, Entry Quality, RSI, and MACD trajectory use prior completed sessions only. No eventual trade outcome participates in these entry measurements.

Exit boundary: the sell-side study is retrospective research. It looks for standard deterioration states after the intratrade peak (Evidence <60 / <50, close below SMA20, MACD histogram below zero, RSI below 50) and measures how long they appeared before the actual exit. These are audits, not new sell rules.

Full-context pool

0 trades

0 stocks captured in v0.4.1.2

Forward Trades—

Forward PIT ↔ Return—

Forward PIT Δ20 ↔ Return—

Evidence Exits—

Winner Peak Capture—

Rerun clean frozen datasets in v0.4.4 to store full entry and exit context.

🎯 Forward Entry Context Fingerprint

Forward-validation trades only. Correlations describe association, not probability or causation.

Knowable before entry	N	Return r	MFE r	Winner median	Loser median
No full-context forward trades yet.

📈 Direction Matters? Forward Trajectory Buckets

Existing descriptive ±5-point bands. No cutoff optimization is performed.

Trajectory	Trades	Avg Return	PF	Median MFE	Median MAE
No trajectory data yet.

🧩 PIT Level × PIT Direction

Forward trades. Each cell shows trade count, average return, and PF.

PIT Level	Falling ≤-5	Flat -4…+4	Rising ≥+5
No matrix data yet.

🚪 Human-Readable Exit Audit

No full-context exits yet.

⚠️ Post-Peak Warning Lead Time

Forward Evidence exits only. “Lead” is completed trading sessions between the first post-peak warning and the actual exit signal.

Warning	Seen	Median lead	Return at warning	Final return	Further giveback
No warning audit yet.

🧠 What This Layer Is Saying

No full-context evidence yet.

📚 Full-Context Source Coverage

Stock	Cohort	Context Trades	Median PIT	Median PIT Δ20	Strategy Return	Captured
No v0.4.4 context data stored yet.

Research discipline: if this study suggests a candidate entry or exit condition, freeze the interpretation first. The next step is untouched validation, not immediate optimization of 75 / 90 / 40 / 90.

⛰️ Strict Trailing Deterioration + Gate Age Study

This correction asks two point-in-time questions: how long had each individual entry gate already been qualified? and what deterioration could the Lab actually have observed from running maxima known on that completed day? The frozen 75 / 90 / 40 / 90 strategy remains unchanged.

Gate age: Evidence age and Entry-Quality age are counted separately. This avoids pretending the combined gate can normally remain qualified for days before entry. Bands remain descriptive: 1 day, 2–4, 5–9, 10+.

Strict point-in-time trailing audit: from entry onward, every completed day updates only maxima already observed so far. Evidence / RSI / MACD warnings use running maxima, ATR warnings use the running highest close, and profit-protection warnings use the running highest completed close. Profit giveback does not arm until the position has first closed at least +5% or +10%. No eventual peak date is used to trigger a warning.

Eligible-record rule: this tab uses only experiments saved by v0.4.4. Older records remain available elsewhere in the Lab but are excluded from every denominator here.

Strict-study pool

0 trades

0 stocks captured in v0.4.1.2

Forward Trades—

Evidence Age ↔ Return—

Entry-Q Age ↔ Return—

Evidence Exits—

Best Early Warning—

Rerun clean frozen datasets in v0.4.4 to store strict gate-age and trailing-warning data.

🌱 Evidence Age + Entry-Q Age Buckets

Forward trades only. Each gate is aged separately through the completed signal day before execution.

Gate	Age	Trades	Win Rate	Avg Return	PF	Median MFE	Median MAE
No gate-age data yet.

🧭 Which Gate Arrived Last?

The first day both frozen score gates are satisfied is normally the entry signal. This view separates which component was newly arriving versus already mature.

No gate-age data yet.

📉 Strict Running-Max + Armed Profit Warning Audit

Eligible forward Evidence exits only. Profit-protection warnings arm only after a completed close first reaches +5% or +10%, then measure giveback from the highest completed close seen so far. “Still +” shows whether the position was profitable when the warning fired.

Point-in-time trailing warning	Seen	Median lead	Still +	Return at warning	Final return	Further giveback
No peak-relative warnings yet.

🏁 Earlier Than Actionable Absolute Warnings?

No warning comparison yet.

💰 Profit Preservation Cases

No full-context exits yet.

🧠 What This Study Is Saying

No peak/maturity evidence yet.

📚 Peak/Maturity Source Coverage

Stock	Cohort	Context Trades	Median Evidence Age	Median Entry-Q Age	Strategy Return	PF	Captured
No v0.4.1.2 strict-study data stored yet.

Research discipline: attractive maturity bands or peak-relative warnings remain research candidates only until frozen and tested on untouched data.

⚔️ Counterfactual Exit Arena

Five independent versions of the frozen system enter with the same 75 / 90 rules. The four candidates add armed profit protection to the existing Evidence <40, Entry-Q deterioration, and 90-day exits. Signals use completed closes and execute at the next session open.

Fixed policies, no tuning: Baseline Evidence <40; +5% close arm with 25% or 50% giveback; +10% close arm with 25% or 50% giveback. The arm is based on the highest completed close known so far. Once armed, giveback is measured from that running highest close.

Untouched exit-validation cohort was pre-registered before seeing Arena results: PEP, MCD, WMT, ORCL, CSCO, ABBV, UPS, RTX. These eight are the exam. Previously studied stocks are automatically labeled Exploratory and are never counted as untouched validation.

Current stock

No stock loaded

Load real or CSV data first.

Untouched Captured0 / 8

Current Baseline—

Best Candidate—

Best Δ vs Baseline—

Load a stock, then run the Arena. It uses the frozen entry settings automatically; you do not need to run the ordinary backtest first.

🥊 Current-Stock Policy Comparison

Each policy is a complete independent backtest. Earlier exits may allow different later re-entries, which is intentional because we are testing the whole system, not merely cosmetically editing the same trade ledger.

Policy	Return	Δ vs Base	Max DD	PF	Trades	Win Rate	Avg Winner	Avg Loser	Best Trade	Top-3 Gross Wins	Exposure	Avg Hold	Protective Exits
Run the Arena to compare exit policies.

🦖 Monster-Winner Preservation

No Arena result yet.

🧠 Current-Stock Interpretation

No Arena result yet.

🧪 Untouched Exit-Validation Aggregate

The aggregate below uses only the pre-registered eight untouched stocks. Each stock contributes one result per policy. “Beat baseline” means that policy produced a higher total strategy return on that stock. Median statistics are reported across stocks to reduce domination by one spectacular ticker.

Policy	Stocks	Beat Baseline	Median Return	Median Δ	Median PF	Median Max DD	Median Avg Winner	Median Best Trade	Median Top-3 Retention	Median Exposure
No untouched Arena results captured yet.

📚 Untouched Source Coverage

No untouched stocks captured yet.

⚖️ Arena Verdict

No verdict until untouched validation begins.

Decision discipline: do not invent a sixth exit after seeing these results. The four candidates are frozen for this Arena. A candidate earns promotion only by surviving the pre-registered untouched cohort without destroying the large winners that fund a trend-following system.

🚦 Entry Sequence Arena

The promoted +5% completed-close arm / 50% giveback exit is now frozen for every policy. This Arena changes only which qualified setups are allowed to enter. Signals use completed-day information and execute at the next session open.

Five frozen entry policies: Baseline immediate qualification; Entry Quality arrives last after Evidence is already established; Evidence confirmed first (at least two Evidence-qualified days); both gates fresh together; and the mirror control where Evidence arrives last after Entry Quality is already established.

Untouched entry-validation cohort pre-registered before results: COST, LOW, QCOM, HON, TMO, LIN, SBUX, SO. Previously studied tickers are exploratory and cannot enter the untouched aggregate.

Current stock

No stock loaded

Click a cohort pip or load data above.

Untouched Captured0 / 8

Current Baseline—

Best Candidate—

Best Δ vs Baseline—

Click a cohort pip to load that ticker. If it was already fetched during this browser session, the Lab reuses the cached stock and SPY data instead of spending another Twelve Data request.

🥊 Current-Stock Entry Policy Comparison

Every policy uses the same risk gate, SPY market gate, costs, slippage, +5/50 promoted exit, Evidence <40 fallback, Entry-Q deterioration fallback, and 90-day maximum hold. A candidate may intentionally skip setups, so trade retention and winner retention are shown alongside return.

Entry Policy	Return	Δ vs Base	Max DD	PF	Trades	Trade Retention	Win Rate	Avg Winner	Avg Loser	Best Trade	Top-3 Retention	Exposure	Avg Hold
Run the Entry Arena to compare sequencing policies.

🧬 Entry-Sequence Fingerprint

No Entry Arena result yet.

🦖 Winner Preservation

No Entry Arena result yet.

🧪 Untouched Entry-Validation Aggregate

Only the eight pre-registered untouched entry stocks count here. A candidate's median trade retention is measured against baseline so an almost-empty strategy cannot “win” merely by refusing to participate.

Entry Policy	Stocks	Beat Baseline	Median Return	Median Δ	Median PF	Median Max DD	Median Trade Retention	Median Top-3 Retention	Median Exposure	Median Avg Winner
No untouched Entry Arena results captured yet.

📚 Untouched Source Coverage

No untouched entry stocks captured yet.

⚖️ Entry Arena Verdict

No verdict until untouched validation begins.

Pre-registered survival check: a candidate must beat baseline on at least 5/8 stocks, have positive median return improvement, retain at least 40% of baseline trades, and retain at least 60% of baseline top-three gross winning returns. We do not change these criteria after seeing results.

🗺️ Strategy Compatibility Arena

This phase asks the swamp question directly: should this stock be eligible for this strategy at all? The entry rule is frozen as Evidence confirmed first. The exit is frozen as the promoted +5% completed-close arm / 50% giveback.

Two-period design: the first 60% of each stock's loaded history is the Compatibility Observation Period. The last 40% is the Evaluation Period. A candidate's eligibility decision is made from the observation period only and is then frozen before evaluation trading begins.

Five frozen compatibility policies: No filter; PIT Trend Suitability ≥40; PIT Trend Suitability ≥50; PIT Trend Suitability ≥60; and Prior Signal Follow-Through Majority, which requires at least 8 fully resolved historical qualifying episodes and more than 50% positive 20-session follow-through. The 40/50/60 bands are the descriptive score bands already used earlier in the Lab, not thresholds searched on this cohort.

Untouched compatibility cohort pre-registered before results: ADP, LMT, AMGN, MU, BKNG, DUK, MDT, NKE. This Arena does not count any previously studied ticker toward its verdict.

Selected stock

ADP

Selected only. Press Run when ready.

Untouched Captured0 / 8

Observation Suitability—

Historical Follow-Through—

Best Current Policy—

Clicking a ticker only selects it. Market data is loaded or restored from session cache only after you press Run.

🧭 Observation-Period Compatibility Profile

No compatibility profile yet.

🥾 Evaluation-Period Policy Comparison

If a candidate rejects the stock, it stays in cash for the evaluation period. That is intentional, because avoidance is the behavior under study.

Compatibility Policy	Eligible?	Eval Return	Δ vs Base	Max DD	PF	Trades	Win Rate	Best Trade	Top-3 Retention	Exposure
Run the Compatibility Arena to evaluate this stock.

🏠 Habitat Interpretation

No result yet.

🛣️ What Was Avoided?

No result yet.

🧪 Untouched Compatibility Aggregate

Each policy is evaluated across the eight pre-registered stocks. Rejected stocks contribute a 0% cash return to the equal-weight portfolio. This prevents a swamp-avoidance policy from hiding what “do nothing” actually means.

Compatibility Policy	Stocks	Eligible Stocks	Beat Baseline	Equal-Wt Portfolio	Median Stock Return	Median Δ	Median PF	Median Max DD	Median Top-3 Retention
No untouched compatibility results yet.

📚 Untouched Source Coverage

No untouched compatibility stocks captured yet.

⚖️ Compatibility Verdict

No verdict until untouched validation begins.

Pre-registered survival check: a compatibility policy must beat baseline on at least 5/8 stocks, improve the equal-weight portfolio return, keep at least 4/8 stocks eligible, and retain at least 60% median top-three winner return. We do not change these requirements after seeing the cohort.

🌦️ Dynamic Entry Regime Arena

The Compatibility Arena showed that a stock's historical personality can change. This experiment therefore asks a different question at each potential entry: is the stock's current point-in-time regime favorable enough to admit the trade?

Frozen strategy underneath: Evidence must already have been qualified for at least two completed sessions, Entry Quality must be ≥90, the risk gate and non-hostile SPY gate remain active, and the promoted +5% completed-close arm / 50% giveback exit remains frozen.

Five pre-registered regime policies: Baseline with no additional dynamic filter; Aligned Trend Structure; Not Unusually Extended; Normal Volatility; and Balanced Dynamic Regime, which requires all three. “Unusually” means above the stock's own prior 60-session 75th percentile, calculated without the current signal day. No numerical threshold is optimized on this cohort.

Untouched dynamic-regime cohort pre-registered before results: AVGO, GILD, CME, WM, GE, FDX, ELV, COP. Previously studied stocks remain exploratory and cannot enter this verdict.

Selected stock

AVGO

Selected only. Press Run when ready.

Untouched Captured0 / 8

Current Baseline—

Best Candidate—

Best Δ vs Baseline—

Clicking a ticker only selects it. Press Run to restore cached data or load it from Twelve Data.

🚦 Current-Stock Dynamic Policy Comparison

Every policy is a complete independent backtest. Dynamic filters are evaluated only when the frozen baseline entry is otherwise eligible. Filtering one entry can change later exits and re-entries, which is intentionally part of the counterfactual system.

Regime Policy	Return	Δ vs Base	Max DD	PF	Trades	Trade Retention	Win Rate	Avg Winner	Best Trade	Top-3 Retention	Exposure	Blocked Setups
Run the Dynamic Regime Arena to compare point-in-time filters.

🧬 Dynamic Filter Fingerprint

No Dynamic Regime result yet.

🦖 Winner Preservation

No Dynamic Regime result yet.

🧪 Untouched Dynamic-Regime Aggregate

Only the eight pre-registered untouched stocks count below. The survival test penalizes filters that improve results merely by refusing most trades or chopping the large winners.

Regime Policy	Stocks	Beat Baseline	Median Return	Median Δ	Median PF	Median Max DD	Median Trade Retention	Median Top-3 Retention	Median Exposure	Median Blocked Setups
No untouched Dynamic Regime results yet.

📚 Untouched Source Coverage

No untouched dynamic-regime stocks captured yet.

⚖️ Dynamic Regime Verdict

No verdict until untouched validation begins.

Pre-registered survival check: a dynamic filter must beat baseline on at least 5/8 stocks, have positive median return improvement, retain at least 50% of baseline trades, and retain at least 70% median top-three gross winning returns. These requirements stay fixed after results arrive.

🧪 Full-System Replication Arena

No new trading rule is allowed in this phase. We are testing whether the research stack that survived earlier untouched exams actually replicates as a complete system on a larger fresh cohort.

Original system: immediate Evidence ≥75 + Entry Quality ≥90 entry; Very High risk and Hostile SPY still block entry; exits use Evidence <40, Entry Quality <25, or the 90-day maximum hold.

Research system: the same gates, except Evidence must already have been qualified for at least two completed sessions; once a completed close reaches +5%, the promoted 50% giveback profit-protection exit is armed. Existing fallback exits remain active.

Reality checks: both systems are compared with Buy & Hold and matched-period SPY. Twelve Data stock and benchmark returns remain price returns unless dividend data has separately been loaded.

Untouched replication cohort pre-registered before results: MDLZ, CRM, NFLX, TXN, PLD, EOG, CL, MO, MAR, PNC, ETN, SYK. These 12 tickers have not participated in the earlier research cohorts.

Selected stock

MDLZ

Selected only. Press Run when ready.

Untouched Captured0 / 12

Original—

Research—

Research Δ—

Clicking a ticker only selects it. Press Run to restore cached data or load it from Twelve Data.

🥊 Current-Stock Replication Comparison

The Original and Research systems are independent full backtests. Buy & Hold and SPY use the same matched test window beginning after the 200-session warm-up.

System / Benchmark	Return	Δ vs Original	Max DD	PF	Trades	Win Rate	Avg Winner	Avg Loser	Best Trade	Top-3 Gross Wins	Exposure
Run the Replication Arena to compare the finished systems.

🦖 Research vs Original Winner Economics

No replication result yet.

🌍 Reality Check

No replication result yet.

📚 Twelve-Stock Replication Aggregate

Every captured stock contributes one independent result. Equal-weight portfolio return is the arithmetic mean of the 12 stock-level percentage returns. It is a comparison device, not a modeled rebalanced portfolio.

System	Stocks	Beat Original	Equal-Wt Return	Median Return	Median Δ	Median PF	Median Max DD	Median Top-3 Retention	Median Exposure	Beat B&H	Beat SPY
No untouched replication results yet.

📍 Untouched Source Coverage

No untouched replication stocks captured yet.

⚖️ Replication Verdict

No verdict until the 12-stock exam begins.

Pre-registered primary pass: the Research system must beat Original on at least 8/12 stocks, have positive median return improvement, produce a higher equal-weight return, have median max drawdown no worse than Original, and retain at least 75% of Original median top-three gross winning returns. All five requirements must pass. Buy & Hold and SPY are reported as external reality checks, not tuned pass criteria.

🗂 Recent Daily Data

Date	Open	High	Low	Close	Volume	RSI	MACD

🔬 Method & Data Notes

v0.7.1 portability infrastructure: Adds a top-of-screen Export Empire Backup / Import Empire Backup workflow designed for a Netlify + Dropbox routine. The portable backup includes watchlist slots, Watch History, portfolio sandbox settings, fundamental snapshots, prediction journal, and research logs, but intentionally excludes the Twelve Data API key so credentials remain device-local.

v0.7.1 device workflow: Recommended pattern is one Netlify URL for the app itself plus periodic JSON backups stored in Dropbox. Export from one device after meaningful updates, then import that same file on phone, home PC, or work PC to synchronize the Empire state across devices.

v0.7.0 Investment Empire: Renames the product to Aoi-chan Investment Empire and makes Empire View the clean default home. Primary navigation is reduced to Empire View, Investment Compass, Market Lab, and Data & Settings. The old single-stock loader, Twelve Data bridge, API controls, and CSV tools move backstage to Data & Settings. The full research suite remains intact inside Market Lab.

v0.7.0 Watchlist workflow: Watchlist ticker and owned/not-owned changes now auto-save. The main daily action is renamed Update Empire. Clear-all maintenance is demoted to Data & Settings, while each card uses the explicit Remove from Watchlist label so removing a watch candidate cannot be confused with selling a real position.

v0.7.0 product shell: APP_NAME and APP_VERSION are master values for the current product identity and version. Browser title, header, footer, and current export metadata derive from them at runtime.

v0.6.0 Empire Watchlist: Adds five persistent watch slots, owned/watch context, credit-efficient Update All, latest-completed-close decision snapshots, one record per ticker per completed market date, Watch History, Opportunity Reserve display, and compact ADD / REDUCE alert counts. Watchlist updates deliberately ignore a same-day still-forming daily candle before 4:10 p.m. New York time. The Watchlist requests SPY once per session when needed and then requests each watch stock directly to reduce Twelve Data credit usage.

Version-control cleanup: APP_VERSION is now the single master current-version value. Browser title, main header, footer, and new export metadata read from that value at runtime so future releases do not require hand-updating the same number in several places.

v0.5.1 hard-gate correction: Attractive timing alone can never produce an ADD label. A new-position ADD candidate requires at least 4 of 6 fundamental categories researched, Ownership Quality ≥60, Evidence ≥75 for at least 2 completed sessions, Add Opportunity ≥75, and usable portfolio capacity with Portfolio Fit ≥55. Missing fundamentals produce RESEARCH CANDIDATE; missing confirmation produces WAIT FOR CONFIRMATION; missing portfolio inputs produce OPPORTUNITY IDENTIFIED • SET PORTFOLIO LIMITS.

v0.5.1 Investment Compass: The Lab now separates four different decisions: Ownership Quality, Add Opportunity, Reduce/Sell Risk, and Portfolio Fit. Ownership uses a manual fundamental snapshot as a bridge until automatic point-in-time financial statements are wired; with fewer than four researched fundamental categories, the ownership score is explicitly provisional and based primarily on market-history quality. Add Opportunity is a soft timing summary informed by Entry Quality, Evidence confirmation, market environment, and extension. Reduce Risk incorporates the promoted +5% completed-close / 50% giveback logic for owned positions when cost and purchase date are available. Portfolio Fit measures concentration capacity only.

v0.5.0 Prediction Journal: Point-in-time Compass snapshots can be saved locally before future outcomes are known. When later daily data for the same ticker is loaded, the Lab calculates actual 63-, 126-, and 252-session price returns (approximately 3, 6, and 12 months) where enough future sessions exist. This creates a forward-validation dataset for later predictive research.

v0.5.0 Visual refresh: The default interface uses a lighter slate-gray workspace with darker chart lines and text for lower eye strain. The underlying research arenas and stored browser data remain intact.

v0.4.6.1 Replication cohort correction: ADBE was viewed in another research tab before its intended replication run and is therefore excluded from the pristine replication verdict. MDLZ replaces ADBE as the twelfth replication stock. Existing ADBE replication data may remain stored locally but is ignored by the untouched aggregate because it is no longer a member of REPLICATION_UNTOUCHED.

v0.4.6 Full-System Replication Arena: No new strategy rule is introduced. Original = immediate Evidence ≥75 / Entry Quality ≥90 with Evidence <40, Entry Quality <25, and 90-day exits. Research = Evidence confirmed for at least two completed sessions before entry plus the promoted +5% completed-close / 50% giveback exit, with the same fallback exits. Both retain the Very High risk and Hostile SPY entry blocks. Buy & Hold and matched-period SPY are external reality checks. Untouched cohort: MDLZ, CRM, NFLX, TXN, PLD, EOG, CL, MO, MAR, PNC, ETN, SYK. Primary replication requires all five pre-registered conditions: Research beats Original on ≥8/12, positive median Δ, higher equal-weight return, median max drawdown no worse than Original, and ≥75% median top-three winner retention.

v0.4.5 Dynamic Entry Regime Arena: Entry is frozen as Evidence confirmed first and exit is frozen as +5% completed-close arm / 50% giveback. Four extra entry filters are compared against baseline: aligned trend structure, current SMA50 extension not above its own prior-60-session 75th percentile, current ATR% not above its own prior-60-session 75th percentile, and a balanced combination requiring all three. The current signal day is excluded from percentile estimation. Untouched cohort: AVGO, GILD, CME, WM, GE, FDX, ELV, COP.

v0.4.5 Twelve Data usage behavior: Auto-refresh Usage now defaults OFF. Manual Refresh Usage remains available and still consumes one Twelve Data credit when used.

v0.4.4 Strategy Compatibility Arena: The first 60% of each stock history is used only to characterize compatibility; the last 40% is held out for evaluation. Entry is frozen as Evidence confirmed first, and exit is frozen as +5% completed-close arm / 50% giveback. Compatibility policies are fixed before the untouched cohort: no filter, PIT ≥40, PIT ≥50, PIT ≥60, and prior qualifying-signal follow-through majority with at least 8 resolved episodes. Untouched cohort: ADP, LMT, AMGN, MU, BKNG, DUK, MDT, NKE.

v0.4.4 Pip Selection UX: Cohort pips now select only. A selected pip is disabled so a second click cannot trigger anything. Market data is loaded or restored from session cache only when the Arena Run button is pressed. Completed saved results can be inspected without re-fetching the ticker.

v0.4.3 Entry Sequence Arena: The promoted +5% completed-close / 50% giveback exit is frozen for every entry policy. Five fixed entry policies are compared as complete independent systems. The untouched entry-validation cohort was pre-registered as COST, LOW, QCOM, HON, TMO, LIN, SBUX, and SO before Arena results were observed. The survival screen is fixed at ≥5/8 baseline wins, positive median return improvement, ≥40% median trade retention, and ≥60% median top-three gross-winner retention.

v0.4.3 Session Market Cache: Live stock histories loaded during the current browser session are cached in memory, and SPY is reused after its first successful fetch. Clicking cohort pips selects the ticker and restores cached data when available; otherwise it performs the normal Twelve Data load.

v0.4.2 Counterfactual Exit Arena: Five fixed exit policies run as independent complete systems with the same frozen 75 / 90 entry logic, next-open execution, costs, slippage, market gate, risk gate, Entry-Q deterioration fallback, Evidence <40 fallback, and 90-day maximum hold. Armed policies use only completed closes known at the time. The untouched exit-validation cohort was pre-registered as PEP, MCD, WMT, ORCL, CSCO, ABBV, UPS, and RTX before Arena results were observed.

v0.4.1.2 Armed Profit Protection patch: Evidence age and Entry-Quality age remain separate. Profit giveback warnings now arm only after a completed close has reached +5% or +10%, then measure 25% / 50% giveback from the highest completed close seen so far. Tiny intraday gains can no longer trigger the profit-protection rows. The strict-study tab counts only v0.4.1.2-compatible records.

v0.4.4 Entry + Exit Study: Full pre-entry context is captured only from completed information available before next-session execution. The exit warning audit is retrospective and looks only for when standard deterioration states became visible after a trade peak. No new entry or exit gate is applied.

v0.4.4 Path timing correction: For next-open exits, MFE/MAE now stop at the actual exit price rather than using the rest of that day's high/low after the position was already closed. Rerun a stock in v0.4.4 to refresh its stored path statistics under the corrected timing logic.

v0.3.8.5 Credit meter: The usage parser now explicitly reads Twelve Data daily_usage / daily_limit and current_usage / plan_limit counters, showing both daily and minute usage numerically with percentages.

v0.3.8.5 Credit meter: The Lab still reads Twelve Data credit headers opportunistically, but browser CORS can hide custom headers. By default it now follows a successful real-data load with the official /api_usage endpoint so the numerical meter can populate reliably. That endpoint costs 1 API credit and can be disabled with the checkbox beside the refresh button.

v0.3.8.5 Credit meter: Twelve Data credit balance is read from documented API response headers when available. The optional Refresh Usage button calls /api_usage, which Twelve Data documents as costing 1 API credit.

v0.3.8.5 Point-in-Time boundary: Pre-entry suitability is calculated only from information available through the completed signal day before next-session execution. Exit timing diagnostics are retrospective audits and do not alter the frozen sell logic. Neither layer becomes a decision gate until it survives separate untouched validation.

v0.3.8.5 Forward Report: Forward-only correlations, map-region results, and winner/loser comparisons are calculated only from stocks outside the original eight-stock research cohort. These are still descriptive and sample-size sensitive. The report explicitly flags untested regions and failed hypotheses.

v0.3.8.5 Research Map: The map separates the original eight-stock Research Cohort from all different Forward Validation stocks. Correlations and region summaries are descriptive. Do not change frozen strategy settings or diagnostic cutoffs to improve the map.

v0.3.8.5 Trade Path layer: MFE/MAE, peak capture, time-to-threshold, and winner concentration are descriptive diagnostics calculated from the same executed trades. They do not change signal generation or the frozen 75 / 90 / 40 / 90 rules.

v0.3.8.5 research boundary: Trend Suitability is a descriptive diagnostic only and does not alter entries, exits, or the frozen 75 / 90 / 40 / 90 rules. Twelve Data daily/weekly/monthly prices are split-adjusted; dividends require separate data. The Lab remains in price-return mode unless dividend data is explicitly loaded.

Real-data loading: v0.3.8.5 uses Twelve Data’s /time_series endpoint for daily OHLCV. Your API key is used only for requests to Twelve Data and can optionally be remembered in browser local storage. It is not included in Lab exports. CSV import remains available if an API request is unavailable.

For manual imports, expected CSV fields: Date, Open, High, Low, Close, Volume. Column names are matched case-insensitively; Adj Close is accepted as a close substitute. The demo dataset is synthetic and is explicitly treated as a Development dataset. Use different, unseen datasets for Validation.

Evidence = 45% trend + 35% momentum + 20% participation. Entry quality is scored separately, risk is based primarily on ATR% and price extension, and the benchmark acts as a market-regime/confidence modifier. Decision gates can cap a bullish label when entry quality is poor, volatility is very high, or the market regime is hostile.

The backtest waits until enough history exists for the 200-day average, calculates each signal from completed day t, executes at day t+1 open, and marks the portfolio at that day's close. v0.3 retains the daily equity path and entry-state metadata, adds saved experiment history, threshold sensitivity, Development/Validation roles, and aggregate clean-validation results across datasets. Trading cost and slippage assumptions are configurable. v0.3 stores experiment summaries in browser local storage when available; no brokerage credentials or portfolio accounts are used.

Aoi-chan Market Lab stores no account credentials and sends no portfolio information anywhere.