SeaHow we evaluate and present information.
What we use, how we calculate it, how we label it, and what it can't tell you. We do not publish performance statistics, win rates or returns unless they come from verifiable records over a clearly stated period.
Where the numbers come from
| Data | Source | Notes |
|---|---|---|
| Crypto prices and candles | Apex Omni public market data | The platform's charts and scanners use closed candles only |
| Macro: S&P 500, Nasdaq 100, VIX, US 10-year, gold | Yahoo Finance; FRED where configured | Each value states its source and session |
| DXY | Computed from ECB reference rates (ICE DXY formula) | Daily reference fix |
| BTC dominance | Conduit, or CoinGecko | |
| News | Conduit and public feeds | Each headline names its publisher |
| Economic calendar | ForexFactory (USD events) | With forecast, previous and actual where published |
| Stablecoin supply and DeFi TVL | DefiLlama | |
| Website demonstrations | Coinbase public market data | Calculated in your browser; labelled on each demo |
Every value says when it was taken
Each live value is shown with its source and an "as of" time in Eastern Time. Market values carry their session (for example a live quote, or the last close). When a source fails, its last good value is kept and labelled with its time, or the value is shown as unavailable; it is never replaced with an estimate.
The formulas
Deans
Four signed measurements per candle (upper wick, body, lower wick, full range), each from its own starting price. Full definition.
Efficiency ratio and regimes
Trending with the 20 / 50-period averages aligned and ER ≥ 0.30; range-bound below 0.20; transitioning otherwise; confirmed after 3 agreeing observations.
Realised volatility
Ranked against the last 180 observations; high at or above the 80th percentile, low at or below the 20th.
Correlation
Daily log returns on common dates; 7, 30 and 90-day windows with minimum samples of 5, 15 and 40; treated as significant only when |t| ≥ 2.
Reward-to-risk and sizing
Statistical intervals
Win rates in the ICC research are shown with a 95% Wilson score interval and their sample size; results under 10 samples are flagged and not ranked.
How sequences are recorded
Every ICC sequence the analyst detects gets one permanent Global ID. Each phase is measured once, from the closed candles available at that moment, and never overwritten; later changes are appended to its history. After an entry, the outcome is read from later closed candles: the target or the stop, whichever came first; a candle that touches both counts as the stop; maximum favourable and adverse excursion are recorded. Detected, completed and executed counts are kept apart, and trade results come only from closed trades the customer logged against that sequence.
Historical tests, labelled as such
Backtests run on real exchange history. They report raw historical performance first, then what survived further testing. Simulations are labelled as simulations, and small samples are flagged. Backtesting tools cannot place orders.
Findings from historical data are checked on a later, held-out period before they are presented as holding up, and they are always described as observations, not rules.
Four kinds of result, never mixed
Live results
Closed trades actually placed on an exchange and recorded.
Historical results
What happened to recorded setups in the market, measured from real candles after the fact.
Backtested results
Rules applied to past data. Useful for research; they did not happen as trades.
Simulated results
Hypothetical scenarios, such as the website's execution preview. Always labelled.
What the models can't do
- Past behaviour does not guarantee future behaviour. Regimes change and relationships between markets break.
- Correlation shows co-movement, not cause.
- Backtests can overfit and ignore costs such as fees, slippage and funding.
- Small samples are unreliable, which is why they are flagged.
- Data sources can be late, revised or unavailable; values are labelled with their time and source.
- Technical concepts such as fair value gaps or structure breaks are definitions, applied consistently — not predictions.
- Nothing on the website or the platform is financial advice.