Duvet Fondange — interface for market analysis and predictive models
Public performance logs

Predictive models whose results are verifiable, not promised.

Duvet Fondange aggregates market feeds in real time, calculates portfolio optimization scenarios and publishes its entire execution logs. No winning guarantee displayed, only searchable data.

Transparency

Every engine decision is logged and viewable.

The flows below illustrate the data display format. Dashboard access provides access to actual, time-stamped, exportable logs.

EUR/USD+0.12%
CAC40-0.34%
BTC/EUR+1.05%
DAX+0.22%
OAT 10A-0.08%
WTI+0.47%

Display format for illustrative purposes. Actual values ​​are available after authentication.

History of executed recommendations

S-9S-6S-3S-0

Schematic representation. The detailed interactive chart — with filters by instrument and by period — is available in the connected space.

Raw log

Each recommendation, its timestamp, its instrument and its closing result are exportable in CSV format, without aggregation or statistical smoothing.

View Logs
Architecture

An engine designed for decision latency and robustness.

Three separate components process the raw data through to the executable recommendation, each audited separately.

01 — INGESTION

Ultra-low latency

Connections to marketplaces use direct binary streams. The average time between reception and processing remains less than ten milliseconds on instruments monitored continuously.

02 — MODELING

Hybrid predictive models

The models combine time series and volume signals to produce movement probabilities, recalculated with each significant market update.

03 — CONTROL

Risk management logic

Each recommendation is bounded by position size and volatility thresholds defined at the account level, before any sending to execution.

Portfolio optimization under constraints

The optimization module does not seek the theoretical maximum yield. It arbitrates between exposure, correlation between positions and available liquidity, according to the parameters set by the user.

Recalculations are triggered by market event rather than at a fixed interval, which limits decisions made on already outdated data.

Duvet Fondange — dashboard for portfolio analysis and predictive models
Methodology

From raw flow to executed decision, in four traced steps.

Each step produces an independent log, viewable separately from the final result.

01

Aggregation

Market flows, order books and macroeconomic data are collected continuously and standardized in a single format.

02

Analysis

Predictive models evaluate probable short- and medium-term scenarios for each instrument monitored.

03

Optimization

The signals retained are weighted according to the risk and correlation constraints defined for the portfolio.

04

Execution

The final recommendation is transmitted to the connected broker, with execution confirmation logged in real time.

Verification

Results are verified by the community, not just the publisher.

The verification protocol is based on the systematic publication of logs, which can be consulted by any connected user, regardless of their subscription.

Verification protocol

Each published recommendation receives a unique identifier and a timestamp. Users can compare the initial recommendation to the closing result, without the possibility of retroactive modification of the log.

Active community workforce viewable in real time from the dashboard after connection.

API Connectivity

Access to recommendations and logs is via standard interfaces, compatible with the existing infrastructures of quantitative traders.

  • REST API
  • WebSocket flow
  • FIX protocol
  • CSV/JSON export
Access

Three levels of access, depending on the volume of data processed.

Pricing is communicated after verification of the profile and volume of instruments monitored. The features below are fixed, regardless of the final price.

Analyst
On quoteIndividual use
  • Access to public newspapers
  • 10 instruments monitored continuously
  • CSV export of recommendations
  • Standard latency
Request access
As for
On quoteProfessional use
  • All Analyst level
  • Unlimited Instruments
  • REST API and WebSocket connection
  • Priority latency
  • Customizable risk thresholds
Access the Dashboard
Institution
On quoteDedicated infrastructure
  • All Quant level
  • Dedicated FIX protocol
  • Co-located hosting
  • Extended audit log
Request access
Technical questions

Answers to common objections about latency and data integrity.

How is latency measured and guaranteed?

The announced latency corresponds to the time between receipt of market data and updating of the model concerned. It is measured continuously and can be consulted in the connected space technical journal.

Can performance logs be edited after publication?

No. Each entry is timestamped and linked to a fixed identifier. Any corrections appear as additional entries, never as a modification of the existing history.

What market data feeds the models?

Feeds come from marketplaces and brokers connected via API. The exact list of sources depends on the instruments followed and is detailed in the technical documentation provided upon account activation.

What happens if the data flow is interrupted?

The engine suspends the issuance of new recommendations on the instrument concerned and records the incident in the technical log, immediately visible on the dashboard.

Contact technical support

Check the newspapers before deciding on a subscription.

Encrypted connection, two-factor authentication available, no wallet data shared without explicit user action.