BLUEPRINT IMPLEMENTATION · RESEARCH BUILD
Evidence before
exposure.
A local, persistent research system for market data, strategy experiments, cost-aware validation, deterministic risk, and paper operations.
LIVE OBSERVATION
Price history
LIVE BINANCE USDT SPOT TICKERS
Live opportunity scan
Live prices and the coin scanner follow the selected exchange: Binance scans active USDT spot pairs; Delta International scans its currently operational USDT perpetual contracts. Closed-candle strategy scans refresh on a rate-limited cycle while this page stays open; a full-universe pass takes tens of seconds or longer depending on network conditions. S1 is calibrated for 4H bars and S2 for daily bars. S3 ranks daily cross-sectional momentum but has no portfolio simulator. S4 is a closed-bar intraday breakout (1m–1h) with a separate provisional view. Provisional candles never count as confirmed signals. A setup is research evidence only: per-symbol validation, calibrated probability, measured costs, and the after-tax EV gate are not yet satisfied.
| Pair | Live price | 24h change | 24h quote volume · USDT | Setup / trade eligibility | Last closed bar | Why |
|---|---|---|---|---|---|---|
| Waiting for the all-market stream and candle scan. | ||||||
MANUAL RESEARCH ASSIST
Signal review
Load at least 60 valid closed bars; synthetic data never qualifies as evidence.
The monitor evaluates implemented strategies on closed candles. A passing screen only enables a paper plan. You still approve each paper entry; no real orders are sent.
CONTINUOUS STRATEGY MONITOR
Research and signal status
| Strategy | Closed-bar state · gates · movement · recent history | Evidence status | Next step |
|---|---|---|---|
| Connect market data to start scanning the implemented strategies. | |||
Movement values compare closed-bar trigger distance over recent bars. They are measurements, not forecasts. S1, S2, and S4 have single-asset closed-bar simulators. S3 is a screen, not a portfolio backtest; S5–S8 remain unavailable pending required data and execution models. “Screen passed” is preliminary, not a performance guarantee.
RESEARCH PIPELINE
Evidence status
RECENT AUDIT EVENTS
Activity
01 / DATA PROVENANCE
Market data
Public venue stream or timestamped OHLCV import. Every dataset has a visible source and quality state.
GLOBAL PRICE & MARKET AGGREGATION
Broad crypto market snapshot
CoinGecko returns market-cap-ranked coin snapshots (price, market cap, 24h volume and change), up to 250 coins per request. This is a cached REST snapshot, not a tick stream and not Binance execution data. Enter your own Demo API key; it stays in this browser session only.
| Rank | Asset | Price USD | Market cap | 24h volume | 24h change | Provider update |
|---|---|---|---|---|---|---|
| No snapshot loaded | ||||||
Page through the provider’s ranked catalog to explore additional listed assets. CoinGecko’s free Demo plan currently allows 10,000 calls per month and 100 calls per minute; each refresh uses one request. API key is kept in session storage and is not saved to the repo or local backup.
DEFI & ON-CHAIN CONTEXT
Protocol TVL
DeFiLlama public protocol snapshot: total value locked and reported 1-day / 7-day changes. TVL is protocol-level context, not a direct token price signal.
| Protocol | Category | Chains | TVL USD | 1d change | 7d change |
|---|---|---|---|---|---|
| No protocol snapshot loaded | |||||
CHAIN AGGREGATES
Chain TVL
Current chain-level TVL from DeFiLlama. Refreshes together with protocol data.
| Chain | TVL USD | Gecko ID |
|---|---|---|
| No chain snapshot loaded | ||
PROVIDER BOUNDARIES
Which data does what?
Fast market & strategy feed
- Binance WebSocket continues to drive live selected-pair prices, candles and scanner updates.
- CoinGecko adds a broader, multi-exchange market-cap catalog through paginated snapshots.
- CoinGecko snapshots do not feed candle-based strategy signals or replace venue-specific bid/ask data.
On-chain research & future providers
- DeFiLlama adds protocol and chain TVL context. It is not a full wallet-flow or order-book feed.
- CoinMarketCap requests require a protected server-side proxy; the key must not be embedded in this static browser app.
- Continuous cross-user collection, persistent archives and scheduled coverage need an always-on backend/VPS.
MARKET SCANNER
Live opportunity board moved
The 173-coin ticker and closed-bar strategy scan now sit on Overview above Signal Review. Market data remains the place for candle history, support/resistance, and indicator tools.
LIVE MARKET OVERVIEW · PUBLIC SPOT
Multi-timeframe trend scanner
A quick trend context map across reference spot pairs from the selected exchange. It is separate from the opportunity scan. Each cell compares the last closed candle with EMA 20 and EMA 50. It is context for research, not a prediction or trade signal.
| Asset | 15 min | 1 hour | 4 hours | 1 day | 1 week |
|---|
CLOSED BAR SERIES
BTCUSDT · 4H
INGESTION CONTRACT
Data checks
RECENT CLOSED BARS
Dataset preview
| Timestamp UTC | Open | High | Low | Close | Volume | Available at |
|---|---|---|---|---|---|---|
| No bars loaded | ||||||
02 / HYPOTHESIS REGISTRY
Strategy library
Save hypotheses, parameter versions, eligibility, and preregistration records. Every edit is a new trial.
SELECTED HYPOTHESIS
S1 · Trend / breakout
03 / REPRODUCIBLE EXPERIMENTS
Research lab
Closed-bar, long-only spot screening with explicit costs and a chronological out-of-sample split.
FINAL EVIDENCE · ONE SHOT
Untouched holdout vault
Reserve the newest 6 or 12 months from the currently loaded history. Ordinary research runs for this symbol and interval will exclude the locked period. Each preregistered strategy version can be evaluated once; results are saved whether favorable or not. Use only a dataset you have not inspected or tuned against.
| Strategy version | Status | OOS trades | Net return | Max drawdown | Recorded |
|---|---|---|---|---|---|
| Lock a suitable untouched dataset to begin. | |||||
SELECTED-COIN DATA HARVEST
Live order-book recorder
While this page is open and the selected Binance public stream is fresh, the browser samples top-5 depth, best quote, spread, microprice, and taker-flow features once per second. Records are local to this browser, capped at 5,000 samples, and are not historical exchange data or a shared training corpus.
SAVED MULTI-ASSET RESEARCH · AUDIT SUMMARY
50-coin historical archive
This reads the saved archive audit and null-baseline outputs into the site. It does not pretend the 4+ GB raw CSV is in the static bundle or already running through the browser backtester.
Gross null controls · costs not deducted
Label sensitivity · exploratory 20-minute trials
| Trial | Threshold | Train ambiguous | Test ambiguous | Selection allowed |
|---|---|---|---|---|
| Loading saved label trials… | ||||
Data coverage & mask impact by coin
| Coin | Raw rows | Removed rows | Removed windows | 1h non-overlap | Triple-barrier ambiguity | Zero-volume | Missing candles |
|---|---|---|---|---|---|---|---|
| Loading saved per-symbol audit… | |||||||
Inspect monthly mask concentration
| Coin | Month | Primary masked windows | Month-excluded windows | Total window loss | Rows removed | Fatal exclusion |
|---|---|---|---|---|---|---|
| Loading symbol/month audit… | ||||||
DATA COVERAGE · CURRENT LIMITS
Research limitations
Saved historical feature archive: 50 currently high-turnover Binance USDT spot pairs, 12,873,524 five-minute rows from November 2023 through October 2026, with one-hour close-to-close ±0.3% gross labels and a separate one-hour triple-barrier label. This is not a trained model; inference remains disabled. The Research Lab loads the saved audit and baseline summaries.
Backtest scope: experiments use the currently loaded, selected-symbol OHLCV history. The public scanner can list active Binance USDT spot coins, but that does not provide historical constituents, delisted assets, or point-in-time universe membership.
Still unavailable: survivorship-safe 500+ coin history, historical liquidity/spread filters, historical L2 depth and microprice, and an S3 multi-asset portfolio simulator. Those checks remain pending until supported data and simulators exist.
EXPERIMENT CONFIGURATION
Register a run
LATEST RESULT
No experiment run
REQUIRED LIVE-FAILURE CHECKS
Research stress suite
These checks target backtests that fail live: trade-order risk, repeated out-of-sample validation, and fragile parameter peaks. A pass is a research screen, never a live-trading approval.
| Fast / lookback | Slow / exit lookback | Stop (%) | Buffer (ATR) | Gate (%) | Max hold (bars) | Profit factor | OOS net return | Cell |
|---|---|---|---|---|---|---|---|---|
| Run a preregistered sensitivity grid | ||||||||
OUT-OF-SAMPLE REGIME BREAKDOWN
How the strategy behaved by market state
Run a backtest to see bull, bear, chop, and high-volatility event coverage. The classifier uses only closed-bar price and volatility context.
| Regime | OOS bars | Entry events | Win rate | Mean net / trade | Sum of trade returns |
|---|---|---|---|---|---|
| Run an experiment to populate regime results | |||||
Entry outcomes by UTC hour
| UTC hour | OOS entries | Win rate | Mean net / trade |
|---|---|---|---|
| Run an experiment to populate time-of-day results | |||
CROSS-TIMEFRAME SCORECARD
Saved strategy × horizon results
Uses fresh public Binance candles for the selected coin. Unsupported or short histories show “insufficient data”; results are not predictions. Matrix is saved in this browser and included in an explicit Supabase workspace backup.
| Strategy | Style / horizon | Bars | OOS return | OOS max DD | OOS trades | Data |
|---|---|---|---|---|---|---|
| Run the strategy × timeframe matrix | ||||||
EXPERIMENT REGISTER
Saved research runs
| Run ID / time | Strategy snapshot | Data source | In-sample | Out-of-sample | Net expectancy | Placebo p95 | Decision |
|---|---|---|---|---|---|---|---|
| No experiments yet | |||||||
04 / CHALLENGE APPARENT EDGE
Validation gates
Calendar time alone is not enough. Each gate needs data and operational evidence.
LATEST SAVED BACKTEST
No saved run selected
Monte Carlo shuffles the saved OOS trade sequence. With very few trades it is mechanically computable but not decision-grade. Walk-forward windows are research screens, not a live-trading approval.
MULTIPLE-TESTING AUDIT
Sharpe deflation & data sufficiency
PRE-REGISTERED TEST BATTERY
Interpretation limits
Implemented screening checks
- Timestamp order, duplicate and OHLC consistency checks.
- Chronological 70/30 train/test split; strategy parameters do not fit on the test segment.
- After-cost and tax-stress trade accounting.
- Matched random entry placebo with sampled original holding periods.
- Block-resampled 90% lower confidence bound and 1.5x cost sensitivity.
- Additional 0.5% and 1.0% adverse slippage on stop-loss and gap-through-stop exits.
- Approximate Deflated Sharpe Ratio screen that penalizes the number of synchronized saved parameter trials.
- Trade count, equity curve, drawdown and comparison with buy-and-hold or cash.
- Five rolling walk-forward windows with at least a 60-minute purge/embargo (rounded up to whole bars); selected intervals above 60 minutes use at least one complete bar.
Still required before canary
- Independent point-in-time data audit and immutable raw archive.
- Complete T×N matrix of all parameter trials, including failed and negative-result configurations, with aligned returns for valid CSCV/PBO and Deflated Sharpe Ratio review.
- Performance evaluation across bull trend, bear trend, and high-volatility chop; strategies that work only in one regime must be labeled as regime-specific. Executable capacity and measured live fill slippage also remain required.
- Survivorship-safe active and delisted coin universe; historical liquidity screens of at least $5M daily quote volume and at most 0.1% spread; report per-asset median and 25th percentile. Current selected-symbol tests do not satisfy this.
- S1/S4 false-breakout shadow-path rate, S2 mean-shift maximum adverse excursion and time-in-trade / gate thresholds, and S4 forming-vs-closed comparison with TFI, time-of-day and max-hold tests.
- S3 weekly/monthly portfolio simulator with 3/5/10 holdings, turnover and costs, absolute and relative momentum. S3 is scanner-only until implemented.
- Paper operations with reconciliation, fault injection, replay and outage recovery.
- Independent review of current tax/legal interpretation and venue eligibility.
- Signed human approval manifest and rollback / incident plan.
05 / ELIGIBILITY IS A PERMISSION SYSTEM
Horizon engine
Transparent first version: compare candidates, apply a switching penalty, and allow abstention.
DECISION OUTPUT
Why this horizon won
No validated strategy has demonstrated an after-cost edge under current data and policy assumptions.
| Candidate | Eligibility | Evidence | Cost / tax | Risk result | Why alternative lost |
|---|---|---|---|---|---|
| Evaluate a strategy with loaded data to compare research candidates. | |||||
RECOMMENDATION CONTRACT
Required disclosures
Scalping: Disabled for live use. Short moves can be consumed by fee, spread, latency, turnover, and tax liquidity drag. Depth and fill evidence required.
Intraday: Research only; frequent turnover magnifies costs and model noise. Requires independent after-cost validation and capacity evidence.
Swing: Longer holds still face gaps, reversals, outages, and tax realization. Stops are not guaranteed loss limits.
NO TRADE: Valid output whenever evidence, cost, confidence, liquidity, data, or risk gates fail.
06 / DETERMINISTIC RISK AUTHORITY
Risk & paper operations
Paper-only order state machine, sizing, portfolio exposure, loss limits, and incident controls.
POLICY CONFIGURATION
Provisional limits
Levels above paper are locked. There is no live order route or verified forward-fill record to justify capital promotion. A future level change requires the pre-registered trade-count, net EV margin, slippage, drawdown, data-health, and reconciliation gates. Demotion must be immediate; promotion must be fixed-step, logged, and delayed.
This is a governance boundary, not automated live position sizing. Current risk controls size simulated paper positions only.ORDER STATE MACHINE
Paper simulation
A paper fill is a local simulation, not a venue acknowledgement. UNKNOWN simulation demonstrates the reconcile-before-retry rule; it does not contact an exchange.
PAPER EXECUTION JOURNAL
Orders & positions
| Client ID | Created | Strategy / market | State | Quantity | Entry / stop / target or trail trigger | Realized PnL | Fill quality · simulated only | Reconciliation |
|---|---|---|---|---|---|---|---|---|
| No paper order intents | ||||||||
07 / ECONOMIC TRUTH
Costs, tax & cash flow
Three separate ledgers: market PnL, scenario liability estimates, and withholding liquidity.
EXPECTANCY SCENARIO
Illustrative per-trade model
LEDGER SEPARATION
Accounting treatment
Price changes, partial paper fills, realized sells, exchange fees, spread, slippage, impact, and financing.
Conservative stress is labeled as an assumption. Legal classification, gain base, loss offsets, and taxpayer circumstances require professional review.
Withholding/TDS reduces deployable cash until credited or refunded; do not count it automatically as permanent cost.
TRADE-BY-TRADE ECONOMIC GATE
After-tax expected value & profit threshold
Selected symbol: —. Fees, spread and slippage are saved per symbol as assumptions. Illustrative arithmetic from the entered probability, stop, target, tax stress and cost assumptions. Probability calibration, measured costs, and a one-standard-error margin are not yet available, so this panel cannot authorize a trade.
08 / PER-USER CLOUD BACKUP
Cloud backup
Authenticate with Supabase to back up and restore your own work across browsers.
PROJECT CONNECTION
Supabase client settings
The URL and publishable key are saved in this browser. For the hosted site, configure them as Render build variables so visitors do not need to enter them.
ACCOUNT
Not signed in
Supabase sends an email link. Open it in the same browser after adding the deployed site URL to Supabase Auth’s redirect allow-list.
WORKSPACE BACKUP
Save or restore your private workspace
Each signed-in user can access only their own row. Restore replaces this browser’s current local work after a confirmation prompt. Cloud backup is not a tamper-proof audit ledger.
09 / GOVERNANCE & RECOVERY
Audit & approvals
Append-only local event journal, exportable experiment manifests, lifecycle state, and no self-promotion.
DEPLOYMENT LIFECYCLE
Promotion control
Human approval manifest
Recording this note does not grant live permission. It only appends an auditable local review record.
MONITORING & RECOVERY
Current system checks
APPEND-ONLY ACTIVITY