Why is this data important?
In major exchanges, active buy/sell order data is limited to VIP customers with a fee of 5-15 million VND/month (Fiintrade, Vietstock Pro...). They hide this information for a reason:
Vnstock.io.vn decodes this algorithm, offering it for free in part and in full with the Premium package for 299K VND/month.
Heuristic formula and reasoning
factor = (Close - Low) / (High - Low) // closing price weight within the range
buy_active = Volume × factor // active buy
sell_active = Volume × (1 - factor) // active sell
net_flow_VND = (buy - sell) × close // net cash flow in VNDReasoning: When most orders are matched at a high price in a session (close near high), it means buyers are accepting the ask price, indicating active buying power. Conversely, a close near low indicates active selling.
Shark threshold: |net_flow| ≥ 1 billion VND in a session = "Shark detected".
Accumulation over 5/20 periods — detecting long-term buying
- 🦈 BUYING ≥5 periods: a very strong signal — institutions are pushing the price up
- 🦈 SELLING ≥5 periods: a warning sign — distribution is occurring
- Streak detection: automatic label display ("Vietcombank (VCB) has net bought 18 billion VND in the last 5 periods")
- Velocity: order matching speed — detecting robot algorithms that split orders to hide
Detecting 'Pump and Dump' Schemes — Sharks' Tactics to Deceive F0
When retail investors (sheep_noise) rush to buy at the peak but sharks (shark_follow) are selling, the AI warns of a risk score of 85/100 + a label
This is the biggest risk for F0 — Vnstock detects it early to prevent buying at the peak.
AI Layer — Storing Shark History in the Database
Vnstock stores shark_flow_history in the database daily. When asking the AI Analyst
This is something only AI with real data can do — not a generic GPT.