Taiwan Government Revenues

The point forecast beat the no-change baseline overall and in a majority of the evaluated tests.

Model: Agentic Investor TSFM (foundation model)

Accuracy vs naive

+11%Beats the no-change baseline over 10 tests (10 tests, 8 won)

Direction

77%No clear edge over a coin flip (23 of 30 calls right in 10 tests)

Turning points

0 of 9Anticipated no turn, as the no-change baseline does (9 turns in 10 tests, 5 without a turn)

Error vs baseline

11% lowerError ratio 0.89; 1.00 is a tie.

Tests won

8 / 10Complete rolling backtests; ties count as no win.

Test period

1996-12-01 to 2025-12-013 observations ahead per test.

80% band coverage

83%Observed outcomes inside the band across all stored test windows.

• Band coverage uses all stored test windows, including model-selection windows when present; it does not determine the confidence label.
• Evaluated 2026-11-15 against the cached history through 2025-12-01.
• This is an evaluation of the serving model; it does not establish a winner across four model classes.