See where risk concentrates.

What this model actually does

1. Observe closed candles and open interest. Positive changes in OI create modeled exposure at the candle’s closing price. OI includes both sides of a contract; it does not identify actual traders.
2. Distribute hypothetical exposure. Taker-volume imbalance and your leverage weights allocate that exposure to long and short scenarios. This allocation is a heuristic, not observed positioning.
3. Project and track levels. Simplified maintenance-margin formulas estimate liquidation prices. Earlier levels are swept when a later mark-price candle reaches them. Original research mode decays remaining weights; the CoinGlass-inspired mode retains them. Allocation may use the preceding candle as a causal timing hypothesis.
4. Confirm changes in CLLD. CLLD accumulates modeled long additions minus short additions. A sufficient reversal and cooldown produce a confirmed turn. The arrow is timestamped when confirmed.

What it cannot know

Actual account entries, leverage, collateral changes and voluntary closures are private. This model starts with no pre-existing positions, so early levels have a warm-up limitation. Strength is relative to the strongest remaining level; it is not a dollar liquidation amount.

What “calibrated” means

Parameters can be chosen to reduce differences against reference observations. CoinGlass has not supplied its complete algorithm here. An identical model and a profitable strategy have not been established.

Data handling

The app opens a labeled saved dataset when available. Binance data is fetched on first use or when you request a fresh load. Successful datasets are saved on this computer. There is no exchange-account login, order execution, automatic trading, or background refresh. Synthetic demo data is always labeled.

CoinGlass methodology ↗Binance market data ↗