Foldyield Ratings processes order-book and sentiment data to produce ranked trading signals every few seconds. Instant withdrawals apply at all times — capital is never held under lock-up periods.
Each pillar of the platform addresses a specific point of failure in discretionary trading: slow data, inconsistent risk judgement, and delayed execution.
Market feeds, order-book depth, and news sentiment are ingested continuously and normalised into a single data stream, with processing latency measured in milliseconds rather than seconds.
Models estimate volatility clustering and drawdown probability for each position before capital is allocated, rather than reacting to losses after they occur.
Signals that clear the confidence threshold are routed directly to execution, removing the delay between analysis and order placement that affects manual trading.
Most automated trading platforms require funds to sit under a notice period, often between seven and thirty days, before withdrawal requests are processed. This arrangement protects the platform's liquidity management, not the trader's.
Foldyield Ratings separates account liquidity from position management. The AI engine can hold, adjust, or close positions continuously while account capital remains withdrawable at any time, with no minimum holding period attached to a withdrawal request.
Each signal is the output of a three-stage pipeline. The process is designed to be auditable at every step, rather than a single opaque output.
Tick-level price data, order-book depth, and macro announcements are collected across multiple venues and time-stamped to a common clock for comparison.
Natural-language models score news and social sentiment, which is then cross-referenced against volatility clustering patterns to flag periods of elevated risk.
A ranking model weighs non-linear correlations between instruments to produce a single confidence score, which determines whether a signal is executed, sized down, or discarded.
Foldyield Ratings was built on the premise that a trading signal is only useful if its inputs can be inspected. The interface surfaces the confidence score, the contributing factors, and the historical accuracy of each model, so a trader can decide how much weight to give any single recommendation.
The underlying models are retrained on a defined schedule as new market data arrives, rather than left static after initial deployment.
Intraday traders use confidence-ranked signals to prioritise entries during high-liquidity windows, which tightens the effective spread paid on fast-moving instruments.
Risk desks apply the predictive drawdown estimates as an additional input to existing position limits, helping to reduce drawdown on correlated exposures before volatility spikes.
Discretionary investors use the non-linear correlation output to identify fragmented pockets of alpha across asset classes that standard covariance models tend to miss.
Account capital and open AI-managed positions are tracked separately. A withdrawal request only draws from unallocated or settled capital, which allows the platform to process requests on a rolling basis, 24 hours a day, without waiting for a fixed settlement window.
Core models are retrained on a scheduled cycle using recent market data, with additional retraining triggered if model accuracy drifts below a defined threshold. This keeps the signal engine aligned with current volatility conditions rather than relying on a single historical training set.
Account access is protected by encrypted authentication, and fund movements require verification before processing. Infrastructure is segmented so that data analysis systems do not have direct write access to custodial fund records.