Methodology: how the market structure read is made

How bias is computed

The engine takes the candles for one coin and one timeframe (1h, 4h or 1d) and finds swing highs and swing lows. It labels the last 30 swings as higher high, higher low, lower high or lower low. It then counts the pattern. A run of higher highs and higher lows points to a bullish bias. A run of lower highs and lower lows points to a bearish bias. The last break of structure is used as a second input.

Two checks stop weak reads. A staleness check ignores a break that is too old to describe the current market. A range check marks structure as neutral when price is moving sideways between swings. A break counts as stale when it happened more than 35% of the recent swing history ago (at least 5 candles). Structure counts as a range when the last six swing points sit inside a band narrower than 6% of price, or when the last three swings made no new high or low and price sits near the middle of that band. If both swing sides are roughly equal, the last break is stale and price is ranging, the bias is neutral.

The result is a bias (bullish, bearish or neutral) and a context (trend, pullback, correction or range). The "Why" bullets on each coin page are generated by the same rules, so each bullet maps to a condition that was true when the page was built. No AI model is involved.

What BOS and CHOCH mean in our engine

A break of structure (BOS) is a candle close beyond the last swing in the direction of the current bias. A change of character (CHOCH) is a candle close beyond the last protected swing against it. Each event stores its type, direction, level, timestamp and scope. The events table on each coin page shows the last 10.

Scope is internal or external. An internal event breaks a smaller swing inside the current range. An external event breaks a larger swing that defines the range. In our engine, external events are closes beyond the major swing highs and lows. Internal events are closes beyond the smaller swings between them. A CHOCH is only ever internal: it is an internal break in the opposite direction to the current external trend. Treat external events as more significant, and neither as a signal on its own.

Guides: break of structure · change of character

Where the data comes from

Candles come from Binance spot USDT pairs. When a result is not in cache, the system requests fresh candles from Binance, runs the analysis and stores the result for one hour. Up to 400 candles are loaded per timeframe. The analysis needs at least 50 candles, so a very new listing may show no read.

Prices and 24h change in page headers come from the Binance public ticker and are not cached for an hour. Each analysis shows the time it was generated.

Key levels come from two sources: the previous day's high and low, and liquidity levels, which are clusters of equal highs or lows. Each level has a strength value from the engine.

How volatility is measured

Volatility is Average True Range with a 14-candle period (ATR 14), shown as a percent of price so coins can be compared. The label maps ATR % to Low, Medium or High.

The same thresholds apply on every timeframe, so a daily chart will usually read High. Compare labels within one timeframe, not across timeframes. The label describes recent movement. It does not say which direction price will go.

The Trading Sessions tool uses a second measure: the average high-to-low range inside a fixed time window, taken over a number of past sessions. Windows are defined in UTC and shown in your timezone: London prep 06:30 to 07:00, London main 07:20 to 08:30, New York observe 13:00 to 13:30, New York sweep 13:30 to 14:10, New York main entry 14:30 to 15:30. When an AI paragraph is shown comparing today with the average, it is labelled as AI-generated.

How news impact is classified

This is the one place AI is used. BiasNex reads public crypto RSS feeds. A language model classifies each article: sentiment (positive, negative or neutral), impact (high, medium or low), the tickers it concerns, its topic (for example regulation, ETF, hack or exchange), a risk-on or risk-off lean, and a one-line reason. If the model is unavailable, simple keyword rules provide a fallback classification.

On a coin page, the News impact module takes the last 72 hours for that ticker. It shows the highest impact level, the majority sentiment, the overall risk lean and up to five headlines. Headlines link to the publisher. We do not republish articles. Full crypto news is on our sister site, RIX News.

Limits of this module: a model can misread an article. Ticker tagging can miss or over-tag. Coins with little coverage may have no result. Impact describes how important a story looks, not what price will do.

What uses AI

Not financial advice

BiasNex publishes education and analysis tools. Nothing here is a recommendation to buy or sell. Crypto trading carries a high risk of loss, and leveraged trading can lose more than a small stake quickly. Do your own research and only risk money you can afford to lose. See the disclaimer.