Signals & Intelligence

HL Vaults provides two distinct layers of vault intelligence. The first is purely deterministic — signals computed from on-chain data, no AI involved, zero configuration required. The second is AI-powered analysis, which produces free-form insights once you connect an LLM provider. Both are ★ Pro features.

Deterministic Signals

Seven signal cards deliver structured diagnostics the moment you’re on the Pro plan. No setup, no API keys, no AI — just computation against on-chain data. Each card presents a plain-English explanation of what it means.

How is the vault’s trading style classified?

The Style Fingerprint analyzes the vault’s trade history — hold times, fill frequency, and win rate — and assigns one of four archetypes:

Scalper — many small trades, sub-hour hold times
Intraday — positions opened and closed within a day
Swing — medium-duration positions, up to a few days
Position — few large positions held for weeks or more

The fingerprint also measures how consistently the vault’s actual behavior matches its declared approach. A vault that calls itself a position trader but churns positions weekly may be misaligned — a useful signal in itself.

Where does leverage fall on the risk spectrum?

The Leverage Profile captures average and maximum leverage and shows the distribution of leverage used across all positions, binned into six buckets:

0-1x — fully collateralized or no positions
1-2x — modest leverage
2-5x — active leverage
5-10x — aggressive leverage
10-20x — very aggressive
>20x — extreme, near-liquidation territory

The raw categorization is useful, but the real signal is in the deviation from the vault’s own history. A vault that normally runs 1.5x but sits at 4x today is behaving unusually. That delta often tells you more than the absolute leverage figure.

How does the vault recover from drawdowns?

Drawdown Recovery examines every significant drawdown and measures three dimensions: how long recovery took, the average recovery speed, and whether the vault tends to bounce back fully or only partially.

A vault that recovers quickly and consistently demonstrates disciplined risk management. One that lingers in drawdown territory for months carries latent risk — even if its headline return numbers look acceptable. Use this to separate vaults that manage risk from those that have simply been lucky so far.

Is depositor capital concentrated or widely held?

Depositor Concentration applies a Herfindahl-Hirschman Index to the vault’s depositor base:

Well Distributed — no single depositor dominates
Moderately Concentrated — the top depositor holds a significant share
Highly Concentrated — one or two wallets control most of the capital

High concentration introduces structural risk. If the dominant depositor withdraws, the vault’s capital base — and its ability to execute its strategy — can be impaired. Flag this during position sizing, particularly for larger allocations.

How often does the vault flip direction?

Trading Velocity counts how many times the vault has flipped between long and short bias, reported over the trailing 30-day and 90-day windows.

Frequent flips tend to generate more fees, which can drag on net returns. They may reflect active management, or they may indicate reactive, trend-chasing behavior. Infrequent flips suggest a longer time horizon and lower trading costs. Neither is inherently better or worse — the value is in understanding which pattern fits your expectations for the strategy.

How correlated is this vault with others in your portfolio?

Cross-Vault Correlations computes the Pearson R between this vault’s returns and those of every other vault on the platform. High correlation means two vaults move together — similar strategies, overlapping market exposure, or both. Low correlation means their return streams are largely independent.

This signal exists to guard against false diversification. Two vaults with different labels can still correlate heavily if they trade the same assets or use similar frameworks. If you hold three vaults that all correlate above 0.8, you’ve concentrated your exposure — not diversified it.

How fast does the leader act on their own fills?

Fill-to-Mark Lag measures the median delay (in milliseconds) between the vault leader’s fills and the corresponding mark price moves, along with how many samples the measurement is based on. Persistent lag can hint at slower execution or information delays relative to the market.

AI-Powered Analysis

Deterministic signals tell you what is happening. AI analysis attempts to address why. It requires you to connect an LLM provider — OpenAI, Anthropic, or any OpenAI-compatible API — through your settings. Once configured, the AI generates three types of additional analysis.

What does the AI reveal about the vault leader?

The Personality Dossier is an LLM-generated profile of the vault leader’s trading psychology. It examines:

Decision-making patterns — aggressive positioning vs. conservative capital preservation
Risk management behavior — how they size positions and cut losses
Behavioral biases — overconfidence, loss aversion, recency bias, and others
Consistency and discipline across different market regimes

This is not a clinical evaluation. It’s a structured framework for understanding the person behind the strategy. Two vaults can produce similar returns while having very different risk personalities — and that distinction tends to matter most during market stress.

Can the AI forecast returns?

The Return Forecast analyzes historical performance, current market conditions, and recent trading patterns to produce a directional outlook. The card shows:

Outlook — a bullish / neutral / bearish badge
Confidence rating — a percentage meter for how strongly the model holds the view
Rationale — the key reasons behind the call, including the risks the AI identifies as most likely to affect performance

This is directional, not predictive. The forecast is a pattern-recognition exercise — it surfaces what has typically happened after similar conditions in the past, not what will happen now. Use it as one input among many, not as a standalone reason to enter or exit a position.

What warning signs does the AI detect?

Red Flag Detection scans for patterns that have historically preceded adverse events:

Unusual trading behavior relative to the vault’s own history
Sudden shifts in leverage, position sizing, or strategy
Concentrated positions in volatile or illiquid assets
Behavioral patterns that have preceded losses in the past

Each flag includes a severity rating. These are early warnings, not definitive judgments. Treat them as triggers for deeper investigation — when the AI flags something, examine the underlying data yourself before acting on it.

How do you connect an AI provider?

Navigate to Settings → AI.
Select your provider — OpenAI, Anthropic, or a custom OpenAI-compatible endpoint.
Enter your API key.
Click Test Connection to verify everything works.

Once configured, AI analysis becomes available on every vault detail page. You can regenerate individual reports on demand — each analysis type has its own refresh button, so you’re not forced to recalculate everything at once.

Column Reference — the metrics behind each signal
Performance Metrics — LP and DP grades in depth

Alerts

SQL Analyzer