Predictive Analytics for Canadian Markets
Vexquant applies predictive models to large sets of market and portfolio data, then publishes the resulting calls in a public log so you can assess predictive accuracy before committing capital.
Methodology
Vexquant's process is built to be inspected, not taken on faith. Each stage below produces an intermediate output that is recorded, so the path from raw data to recommendation can be traced after the fact.
Price, volume, and on-chain or market-depth data are pulled from multiple sources at regular intervals and checked for gaps or inconsistencies before use.
Models trained on historical data identify recurring relationships between variables that are difficult to isolate manually across large datasets.
Each detected pattern is tested against out-of-sample data to filter signals that do not hold up beyond the period in which they were found.
Surviving signals are converted into a risk-adjusted recommendation, timestamped, and written to the public performance log before any outcome is known.
Core Capabilities
Each tool below addresses a distinct question a risk manager or investor typically has to answer manually. Vexquant does not replace that judgment; it narrows the data down to what is statistically relevant.
Assets and positions are scored against historical volatility bands, flagging exposures that deviate from a portfolio's stated risk tolerance.
Portfolios can be run against modeled market downturns and liquidity shocks to estimate drawdown before capital is at risk.
Identifies when previously uncorrelated holdings begin moving together, a common early indicator of concentrated risk.
Proposes weighting adjustments based on predictive accuracy scores for each asset class, not on static rules of thumb.
Every recommendation carries a confidence rating derived from backtested accuracy, so low-confidence signals can be weighted accordingly.
Exportable reports summarize model inputs, assumptions, and outcomes in a format suited for internal risk committees.
Performance Transparency
Rather than asserting a track record, Vexquant publishes one. Every recommendation is logged with a timestamp before the outcome is known, and the result is recorded once it resolves, whether favourable or not.
| Date Logged | Asset Class | Recommendation | Confidence | Outcome Status |
|---|---|---|---|---|
| 2024-11-03 | Digital assets, large-cap | Reduce exposure | Moderate | Resolved — aligned |
| 2024-11-18 | Equities, cross-sector | Hold, monitor volatility | High | Resolved — aligned |
| 2024-12-02 | Digital assets, mid-cap | Increase allocation | Low | Resolved — diverged |
| 2024-12-21 | Fixed income proxy | Maintain position | High | Pending resolution |
Each log entry links back to the data snapshot and model version used to generate it. Entries are append-only; past recommendations are never edited or removed after the fact, including cases where the signal did not hold.
Outcome labels are assigned automatically once the resolution window for a given recommendation closes, using the same data sources cited in the original entry, so the grading criteria stay consistent across the log.
Use Cases
Two recurring patterns among Vexquant users illustrate how the same underlying models are applied to different mandates.
Corporate Treasury
Problem statement: A treasury team holding a portion of reserves in digital assets needed a repeatable way to size position adjustments without relying solely on internal sentiment or ad hoc analyst calls.
Approach and result: Risk scoring and scenario modeling were applied to the existing holdings, producing weekly confidence-rated adjustment ranges. Decisions remained with the treasury committee; the model's role was narrowing the range of reasonable actions and documenting the reasoning behind each one.
Private Portfolio
Problem statement: An investor managing a personal portfolio across several digital assets wanted an independent, data-based check against emotional decision-making during periods of high volatility.
Approach and result: Correlation mapping flagged when holdings presumed to be diversified began moving in tandem, prompting a rebalance ahead of a broader drawdown. The public log allowed the investor to compare the signal's stated confidence against its actual outcome before acting on similar signals again.
Frequently Asked Questions
No. Vexquant produces data-driven recommendations with stated confidence levels, logged publicly for review. It does not provide personalized financial advice, and no outcome is guaranteed. Decisions based on these outputs remain the responsibility of the user or their advisor.
The outcome is recorded in the same log as "diverged," without alteration or removal of the original entry. This is part of how predictive accuracy is measured honestly over time, rather than only showcasing favourable results.
Signals are validated against out-of-sample periods before being logged, and confidence ratings are recalibrated as new outcome data accumulates. This does not eliminate the risk of a pattern breaking down, but it reduces reliance on patterns that only worked in hindsight.
Portfolio-specific data is used only to generate recommendations for the account it belongs to. Aggregate, anonymized market data contributes to model training; individual holdings are not shared or disclosed to other users.
The model operates within predefined risk thresholds and does not execute trades directly; it produces recommendations for human review. It also does not account for information outside its data sources, such as unannounced regulatory changes.
A full description of model limitations and risk factors is available in our risk disclosure documentation.
Access to the full performance log and model documentation is available without requiring a long-term commitment upfront. Most accounts are configured for initial review within a few business days.