Historical market prices, not portfolio returns.
| Market | Side | Entry | Current | P&L |
|---|
Checking paper portfolio and background scanner…
Market monitor
Kraken BTC, ETH and SOL quotes in GBP. Change compares the latest price with the hourly close at least 24 hours ago; volume is Kraken-only.
| Asset | Price | 24h | Volume | Volatility | Trend | Signal | Rule score |
|---|
Signals with evidence
Rules calculated from completed Kraken hourly candles: 20H/50H trend, price above 20H average and positive 4H momentum. AI analysis is not connected.
Simulated execution
Paper LONG trades use real GBP prices. Stops and targets are checked on each refresh while this page is open; gaps can cross the requested exit price.
| Market | Side | Size | Entry | Current | Stop | Target | P&L |
|---|
| Market | Size | Exit | P&L | Reason | Closed |
|---|
Your previous demo portfolio is preserved separately.
Saved in this browser only. Model assumptions: 0.10% fee and 0.05% adverse slippage per side; these are not an exchange fee quote. P&L includes estimated closing costs.
Research & testing
Historical opportunity replay and forward shadow testing for BTC, ETH and SOL in GBP. Gold v3.9-inspired micro rules, with separate evidence for each target.
Small trades. Measured costs.
Checking Coinbase collection…
Three micro targets and one hourly trend book compare the same Kraken signals with observed Coinbase GBP order books. Each has a separate £25,000 simulated balance and a £25 entry budget. Automatic paper entries and the portfolio pause control apply here.
Fee verification pending.
| Variant | Admitted / open / pending | Resolved / net wins | Net return | Expectancy | Profit factor | Realised drawdown | Uncertain / skipped |
|---|
Net includes both taker fees, observed spread and depth, plus 0.05% adverse slippage on each side. Slippage is already included in fill prices. Drawdown uses resolved trades only; open and uncertain outcomes are excluded. Independent variants must not be added together.
Collection coverage & latest spreads
| Market | Bid / ask | Spread | Quote timestamp | Saved / usable | Collection started |
|---|
Latest 30 decisions & fill evidence
| Decided | Variant / market | Status / reason | Estimated costs | Net result | Audit |
|---|
Method & limits
Decisions precede a later minute's entry snapshot. Fills walk up to five visible book levels and round down to Coinbase size increments. Entry checks are repeated at fill. The 0.22R round-trip cost ceiling is retained; high account fees can legitimately block every entry.
Quotes are sampled once per minute, so intermediate stop or target crossings can be missed. Stops, targets and time exits execute at the next observed eligible book price, not a guaranteed trigger price. Gaps longer than 150 seconds, unavailable exit depth or unverified exit fees produce uncertain results and pause that variant for the UTC day. No missing price path is reconstructed.
Each variant permits one active trial, up to eight entries per UTC day, 0.10% stop-plus-cost risk, 10% notional exposure and a 0.50% daily loss budget. Three consecutive net losses stop new entries. These are admission checks, not guaranteed loss limits. Research does not train itself or submit exchange orders.
Waiting for collection.
Browse and download archived evidence
100 records per page. Each download retains the next cursor. Active trials may update while browsing.
Historical research · 90-day archive
Every covered setup, explained
Checking historical collection…
Background jobs import past Kraken GBP trades and preserve new one-minute candles. Every covered five-minute boundary is evaluated. Technical candidates are replayed at 0.25R, 0.50R and 0.75R, including candidates rejected on cost. Hypothetical results stay separate from forward trials and your paper portfolio.
Coverage & progress · UTC
| Market / archive | Requested window | Historical import | Next replay boundary | Quality / freshness |
|---|---|---|---|---|
| Waiting for archive status. | ||||
Import progress measures time traversed, not a guaranteed completion time. “Not yet covered” includes the gap between the historical import and recent live candles. Zero-trade minutes are explicitly marked and cannot supply an entry fill.
All technical candidates · counterfactual outcomes
| Market | Target | Trials / resolved | Net win rate | Total net | Expectancy | Profit factor | Double-cost net | Mean MFE / MAE bounds | Ambiguous / data gaps | No entry / exit trade |
|---|---|---|---|---|---|---|---|---|---|---|
| Waiting for replay results. | ||||||||||
Method, fill assumptions & limitations
Browse all historical evaluations & download data
Up to 100 evaluations per page; downloads include full inputs and outcome paths.
| Signal time · UTC | Market | Reason | RSI | Costs in R | Outcomes | Evidence |
|---|
Waiting for the forward observation start date.
This is an evidence review date, not a promised launch date. Historical results do not replace forward testing. The engine does not train itself or promote a strategy automatically.
Forward shadow trials · observed from collection start
Sign in to your cloud portfolio to view saved research.
| Target | Admitted / open | Resolved | Net R | Expectancy | Profit factor | Drawdown | Uncertain | Review |
|---|---|---|---|---|---|---|---|---|
| Waiting for owner research data. | ||||||||
R is the initial stop distance. Each variant has its own simulated book; do not add their returns together. Review requires 100 resolved trials, positive net expectancy, profit factor ≥ 1.2, drawdown ≤ 4R and no uncertain outcomes. Review never enables live trading.
Up to 0.10% simulated stop risk per entry, capped at 10% notional exposure. One position per variant, up to 8 entries per UTC day. Entries stop after 3 consecutive net losses or when the 0.50% daily loss budget cannot cover another stop plus modelled costs.
Targets: 0.25R, 0.50R and 0.75R. Maximum holding time: 30 minutes. Enable Automatic paper entries in System settings; the portfolio pause control also pauses new shadow entries. Existing trials continue to resolve.
Model assumptions: 0.10% fee and 0.05% adverse slippage per side, not a Coinbase or Kraken fee quote. A setup is skipped if estimated round-trip costs exceed 0.22R. Actual spread, liquidity and fees need separate exchange validation.
Decisions are saved before a future minute's opening price is used as the shadow entry. Missing candles and candles touching both stop and target are marked uncertain. Data is collected for manual review; no model trains itself or changes the strategy.
| Observed | Market | Decision / reason | RSI | Costs in R | Outcomes |
|---|---|---|---|---|---|
| No observations loaded. | |||||
All observations and resolved trial paths are retained in your private cloud dataset. This view and download show the latest 30 observations.
Risk before execution
New entries check cash, position limits, daily equity loss (UTC) and cooldown. Limits can block entries; they do not guarantee a maximum loss.
Pause new simulated entries instantly.
Activity history
Latest 200 signals, paper orders and risk actions. Cloud portfolios retain a full server audit.
System configuration
Paper and future live environments remain deliberately separated.
Checking cloud setup…
Cloud and browser portfolios stay separate. A running background scan checks stops and targets once a minute even when this page is closed. Automatic entries remain off until you enable them. Real exchange trading stays locked.
Sign in to view Coinbase markets and connection status.
This amount previews costs only. It does not create a trade or change your paper settings. Quotes refresh every 60 seconds while System settings is visible, or when you press Refresh Coinbase.
Account balances pending
Coinbase GBP markets
| Market / quote time | Best bid / ask | Spread | Order minimums | Preview status | Rounded preview size |
|---|
Round-trip costs · current snapshot
| Market | Spread + depth impact | Taker fees · both sides | Total estimated cost | Latest Kraken stop | Indicative micro cost |
|---|
Successful fee checks are saved for up to 24 hours of Coinbase paper research. Open Strategy to inspect the separate Coinbase execution lab. Existing Kraken history and forward results keep their original modelled costs.
Connect your Coinbase account · setup
- Create a Coinbase Advanced / CDP key using ECDSA, restricted to the intended portfolio. Enable View only; disable Trade and Transfer.
- In Cloudflare, open Workers & Pages → ai-crypto-trader (Pages) → Settings → Variables and secrets → Production. Add two encrypted secrets:
COINBASE_API_KEY_NAME(the full organizations/…/apiKeys/… name) andCOINBASE_API_PRIVATE_KEY(the complete PEM private key, including its BEGIN/END lines). - Redeploy the Pages project, then press Refresh Coinbase. The companion verifies the key permissions before reading balances and fees.
Enter the key only in Cloudflare's encrypted secret fields. It is never entered in this dashboard, committed to GitHub or added to the paper Worker. Keys with trading or transfer permission are rejected.
Kraken public prices refresh every 60 seconds while this page is visible. Use the Cloud portfolio panel to check storage, owner sign-in and background scans. The browser portfolio runs only while this page is open; a configured cloud portfolio runs on its scheduled Worker.