The scoring engine is algorithmic by design. Every portfolio action follows a documented sequence of inputs, weights, and thresholds — not a black box. This page explains that sequence.
When Coincruze evaluates whether to buy, hold, or rebalance an asset, it runs the same deterministic process every time. There is no human trader making judgment calls. There is no model that adapts its goals based on business incentives.
Each asset in your universe receives a composite score across seven dimensions. Scores are normalised, weighted by your chosen preset, and compared against your portfolio's current allocation. If the deviation exceeds your configured drift threshold, the system proposes a rebalance and executes it within the position limits you approved.
Raw price, fundamental, and news data pulled from FMP, Benzinga, and GDELT on a scheduled cycle.
Each asset is scored across 7 dimensions and normalised to a 0–100 scale.
Scores are weighted by your preset's dimension priorities and combined into a single composite score.
Composite scores are compared against your current allocation and target weights.
If portfolio drift exceeds your configured band, a rebalance candidate is generated.
Proposed trades are submitted to IBKR within your approved position limits and allocation ranges.
Coincruze does not manufacture its own market data. All inputs come from established, independently operated data providers. Here is exactly what each one supplies.
Provides real-time and end-of-day price data, earnings figures, balance sheet snapshots, P/E ratios, revenue growth rates, and sector classifications for equities in your universe. FMP is the primary source for all fundamental scoring dimensions.
Delivers structured financial news and press releases tied to specific tickers. Each article is timestamped, tagged to one or more equities, and rated by editorial relevance. Benzinga feeds Coincruze's news sentiment pipeline alongside GDELT.
The Global Database of Events, Language, and Tone monitors worldwide media — over 100 languages, 65 billion+ records. Coincruze uses GDELT's CAMEO-coded sentiment scores and geographic tone signals to capture macro-level risk context that ticker-specific feeds miss.
Data is cached and refreshed on a scheduled cycle. Coincruze does not stream live tick data or act on sub-minute price movements. The system is designed for long-term, rules-based portfolio management — not high-frequency trading.
Two AI systems contribute to Coincruze's process. Neither has discretionary authority over your portfolio. Both operate within hard-coded rule boundaries.
FinBERT is a BERT-based language model fine-tuned on financial text (earnings calls, financial news, analyst reports). It reads each incoming Benzinga and GDELT article and classifies the sentiment as positive, negative, or neutral for the associated equity. The output is a sentiment score that feeds the News Sentiment scoring dimension.
Claude Haiku (Anthropic's lightweight production model) is used to convert the system's numerical scoring output into plain-English explanations. When your portfolio changes, Haiku writes the rationale you see in your decision log. It does not choose which trades to make — it explains the trades the algorithm already determined.
Neither AI model can override your configured position limits, drift thresholds, or allocation ranges. All trades are generated by the deterministic scoring pipeline first. AI is applied only after the algorithm has already produced a candidate action, never before.
No scoring model can predict market movements. The system identifies signals — it does not guarantee outcomes.
Automation and diversification reduce certain risks but cannot eliminate the possibility of loss.
The system cannot trade outside the position limits and allocation ranges you approved during setup.
If a data feed is unavailable, the system holds rather than acting on incomplete information.
Every triggered action is logged with the score, rationale, and threshold that caused it.
You can disconnect automation at any time. Your broker account and holdings remain accessible.
The Internet Adviser Exemption permits certain investment advisers to operate without state registration when the advisory business is conducted exclusively through an interactive website that delivers advice based on personal information supplied by the client through the website.
Documenting that Coincruze's process is algorithmic — not discretionary — is part of satisfying this standard. The scoring pipeline described above operates identically for all users. No human makes case-by-case investment decisions on your behalf.
For the full regulatory picture, see the Disclosures page.
Six presets, seven scoring dimensions, and exactly how FinBERT and Claude Haiku interact — all documented.
Investing involves risk. Algorithmic scoring and AI-assisted analysis do not guarantee profit or prevent loss. Past performance of any scoring model does not guarantee future results. Review all disclosures before enabling automation.