FORECAST EDGE
A forecast only earns trust where it disagrees with the dumb baseline — a random walk with drift, built from the same window the model saw. Agreement adds nothing to bet on; divergence is the model's actual call.
NAMES FORECAST
6
CONTRARIAN TO DRIFT
2
WIDEST DIVERGENCE
PTT
NAMEKRONOSDRIFTEDGECALL
PTT-16.7%+7.4%-24.2%fights trend
KBANK-21.9%-5.8%-16.2%amplifies
SET-13.0%+1.3%-14.3%fights trend
BBL-12.5%-2.1%-10.3%amplifies
SCB-8.3%-2.7%-5.6%amplifies
ADVANC-3.8%-0.8%-3.0%amplifies
THE REPORT CARD · FORECASTS GRADED ON WHAT HAPPENED
Every forecast is logged the day it is made and graded the day its horizon arrives. The 2026-06-24 exam is in: the model went 3 for 12 on direction, and beat the dumb drift baseline 1 times. A famous model is not a right model — that is the whole lesson of this page.
NAMEKRONOSDRIFTREALITYGRADE
SET-10.3%+5.4%+1.3%miss
PTT-9.0%+1.2%+7.5%miss
KBANK-11.9%+3.9%+15.3%miss
ADVANC-10.0%-2.1%+4.2%miss
SCB-7.2%+7.6%+8.2%miss
BBL-8.7%+7.2%+7.2%miss
SET-15.9%+3.3%-1.0%hit
PTT-14.5%-2.1%+8.5%miss
KBANK-26.3%+7.3%+1.2%miss
ADVANC-9.9%+0.5%-5.9%hit
SCB-16.0%-5.1%-0.7%hit
BBL-17.0%-8.5%+0.0%miss
graded 2026-06-24 → 2026-07-23 · today's forecast above is on record the same way — next grading ~2026-10-10 · we publish the misses