Each entry here documents a specific analytical scenario — what Rimvaro's tool was asked to do, what the data looked like, and where the model performed well or showed its limits. No summaries, just the actual work.
A portfolio manager working with 60 mid-cap equities across the region wanted to know whether the model could identify volatility clustering before earnings periods. The tool was fed 14 months of daily OHLCV data alongside options flow snapshots. It flagged 38 instances of pre-earnings compression patterns, of which 29 preceded significant post-announcement moves.
The question was whether sector rotation signals could be detected early enough to be useful, or whether the model was simply confirming what had already happened. Testing across technology, industrials, consumer staples, energy, financials, and healthcare over a 24-month window showed average signal lead time of roughly 6 trading days ahead of visible price momentum, with shorter leads in fast-moving sectors like energy.
An analyst needed to track how individual equities within a Singapore-listed index shifted their correlation to interest rate movements during a period of rapid monetary tightening. The model ran rolling 30-day correlation windows and flagged 17 companies whose sensitivity coefficient changed by more than 0.3 during the observation window — useful for identifying regime-change risk at the individual stock level.
Each scenario above required more than running data through a model. The preparation work — cleaning feeds, deciding on lookback windows, choosing which signals to weight — took as long as the analysis itself. That context matters when reading the outcomes.
The tool does not produce trade recommendations. What it generates are structured observations about pattern behaviour, correlation shifts, and signal timing. Interpretation and decision-making remain with the analyst.
Across the three scenarios, signal accuracy was strongest where the underlying data was clean and the lookback window matched the natural cycle of the asset class. Equity volatility patterns held up well with 60 to 90 day windows. Sector rotation signals degraded noticeably when tested against periods of macro shock, where historical correlations broke down temporarily.
The macro correlation mapping case was the most instructive in terms of limitations. The model identified shifts accurately, but the timing of those shifts relative to actual price impact varied considerably. A 0.3 coefficient change over 30 days does not translate directly into a usable signal without additional context about the underlying driver.
Signal reliability
Weakest conditions
Optimal lookback
