Search papers, labs, and topics across Lattice.
This paper investigates the phenomenon of forecast collapse in time-series foundation models (TSFMs) when predicting hourly returns for US equities, revealing that predictions become nearly flat and poorly rank stocks due to low predictability and per-series objectives. The authors identify a calibration-ranking tradeoff, where optimizing for squared error leads to flat forecasts, while optimizing for cross-sectional correlation inflates forecast amplitude significantly. To address this issue, they introduce CalibRank, an objective that effectively balances calibration and ranking, resulting in nearly tripled cross-sectional correlation without sacrificing amplitude on the Finance1K dataset.
Forecast collapse in time-series models reveals a critical calibration-ranking tradeoff that could mislead financial decision-making.
When forecasting hourly returns for 1,000 US equities, we observe an unexpected phenomenon: predictions become nearly flat and show poor stock ranking, as measured by cross-sectional correlation. We call this forecast collapse. Surprisingly, the phenomenon largely disappears when forecasting trading volume under the same setting. We investigate forecast collapse across time-series foundation models (TSFMs), twelve deep-learning forecasting models, and 97 public benchmark configurations, and find that it is closely tied to target predictability. We identify two distinct reasons behind it: low predictability limits the amplitude of calibrated point forecasts, while per-series objectives leave cross-series structure unidentified. These findings reveal a calibration-ranking tradeoff: optimizing squared error leads to flat predictions, whereas directly optimizing cross-sectional correlation improves ranking but can inflate forecast amplitude by more than an order of magnitude. To address this tradeoff, we introduce CalibRank, a simple objective that balances calibration and ranking. On Finance1K, CalibRank nearly triples cross-sectional correlation while keeping amplitude close to the target, and improves correlation on all tested models. Our results reveal a blind spot in conventional time-series evaluation: per-series metrics can hide failures in cross-series structure needed by downstream decisions.
