thesisPrototype
methodprediction pack · v0.1.0

Base-rate-first

Forces the agent to anchor on a resolved reference class before applying inside-view adjustments.

28
runs
27
targets
2
agents
1
versions

Purpose

Keep the agent anchored to an explicit outside-view reference class before it applies target-specific adjustments.

Mechanism

  • Identify a resolved historical base rate for the target series.
  • State the base rate before any inside-view adjustment.
  • Show how each adjustment moves the forecast away from that anchor.

Checks

  • A reference class is named before the final forecast.
  • The public trace includes a base-rate calculation or lookup.
  • Inside-view adjustments are signed and small enough to audit.

Inputs

  • Resolved historical observations
  • Forecast horizon and unit
  • Recent trend window selected by the run

Limitations

  • Does not choose the best reference class by itself.
  • Can underreact when the current regime has genuinely changed.

Versions

v0.1.0base-rate-first@0.1.028 runs

Runs using this pack

ASEC income pack set, Australia May CPI pack set, Australia May labour pack set, CPI annual-average pack set, Canada May CPI pack set, Census official-poverty pack set, June core CPI pack set, June labor-market pack set, June monthly CPI pack set, June weekly claims pack set, May BEA personal-income pack set, May BEA social-benefits pack set, May JOLTS labor pack set, May core PCE bridge pack set, May housing starts pack set, May retail sales pack set
CPI-U annual average inflation, 2026
brier-1.packedJun 12, 2026Brier-1 · CPI packs · Jun 12
pack_set.cpi_annual_2026.v1pack-enabled
2.6%
+0.1% vs no-pack

Earlier packed run before the June 14 refresh; same agent and pack set, older source context.

80% interval 2.0% to 3.4%
CPI-U annual average inflation, 2026
brier-1.packedJun 14, 2026Brier-1 · CPI packs
pack_set.cpi_annual_2026.v1pack-enabled
2.7%
+0.2% vs no-pack

June 14 refresh with the base-rate, component decomposition, and tariff pass-through packs enabled.

80% interval 2.1% to 3.4%
Official poverty rate, 2025
brier-1.packedJun 14, 2026Brier-1 · Census cash-income packs
pack_set.census_official_poverty_2025.v1pack-enabled
10.3%
-0.3% vs no-pack

Pack-enabled run that bridges PolicyEngine cash income to the Census official poverty release.

80% interval 9.7% to 11.0%
Median household income, 2025
brier-1.packedJun 14, 2026Brier-1 · ASEC income packs
pack_set.asec_income_2025.v1pack-enabled
$80,800
+$700.00 vs no-pack

Pack-enabled median-income forecast using wage, employment, household-composition, and ASEC release checks.

80% interval $78,900 to $82,900
Nonfarm payroll change, June 2026 (first print)
brier-1.packedJun 15, 2026Brier-1 - labor packs
pack_set.labor_june_2026.v1pack-enabled
150k
+25k vs no-pack

Pack-enabled run with labor-market momentum and first-print calibration.

80% interval 35k to 260k
Unemployment rate, June 2026 (first print)
brier-1.packedJun 15, 2026Brier-1 - labor packs
pack_set.labor_june_2026.v1pack-enabled
4.3%
matches no-pack control

Pack-enabled run cross-checking unemployment persistence against payrolls and claims.

80% interval 4.1% to 4.5%
CPI-U month-over-month, June 2026 (first print, SA)
brier-1.packedJun 15, 2026Brier-1 - CPI energy packs
pack_set.cpi_monthly_june_2026.v1pack-enabled
+0.5%
+0.1% vs no-pack

Pack-enabled headline CPI run with energy nowcast, component decomposition, and tariff tail checks.

80% interval +0.1% to 0.8%
Core CPI month-over-month, June 2026 (first print, SA)
brier-1.packedJun 15, 2026Brier-1 - core CPI packs
pack_set.core_cpi_monthly_june_2026.v1pack-enabled
+0.3%
+0.0% vs no-pack

Pack-enabled core CPI run using component checks and tariff pass-through tails.

80% interval +0.1% to 0.5%
Retail sales, May 2026 month-over-month
brier-1.packedJun 15, 2026Brier-1 - spending packs
pack_set.retail_may_2026.v1pack-enabled
+0.3%
+0.1% vs no-pack

Pack-enabled run using consumer-spending nowcast and release-vintage checks.

80% interval -0.3% to 0.8%
US housing starts, May 2026, total SAAR (millions)
brier-1.packedJun 15, 2026Brier-1 - housing packs
pack_set.housing_starts_may_2026.v1pack-enabled
1.5M
+0.1M vs no-pack

Pack-enabled run using permits, mortgage-rate, builder-sentiment, and preliminary-release checks.

80% interval 1.3M to 1.6M
JOLTS job openings, May 2026 (first print)
brier-1.packedJun 15, 2026Brier-1 - JOLTS packs
pack_set.jolts_may_2026.v1pack-enabled
7.4M
+0.2M vs no-pack

Pack-enabled run using payroll, claims, openings, and JOLTS release-noise checks.

80% interval 7M to 7.8M
Canada CPI all-items year-over-year inflation, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.canada_cpi_may_2026.v1pack-enabled
2.7%
+0.1% vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval 2.1% to 3.4%
Canada CPI annual inflation, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.canada_cpi_may_2026.v1pack-enabled
2.7%
+0.1% vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval 2.1% to 3.3%
Australia CPI annual inflation, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.australia_cpi_may_2026.v1pack-enabled
4.6%
+0.4% vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval 3.8% to 5.4%
Australia employment change, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.australia_labor_may_2026.v1pack-enabled
20k
+10k vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval -35k to 85k
Australia unemployment rate, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.australia_labor_may_2026.v1pack-enabled
4.5%
matches no-pack control

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval 4.2% to 4.8%
Core PCE, May 2026 month-over-month
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.core_pce_may_2026.v1pack-enabled
+0.3%
+0.1% vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval +0.1% to 0.4%
US disposable personal income, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_personal_income_may_2026.v1pack-enabled
$23,520B
+$20B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $23,445B to $23,620B
US government social benefits, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_social_benefits_may_2026.v1pack-enabled
$4,998B
+$8B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $4,950B to $5,055B
US Medicaid benefits, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_social_benefits_may_2026.v1pack-enabled
$1,041B
+$3B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $1,025B to $1,060B
US Medicare benefits, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_social_benefits_may_2026.v1pack-enabled
$1,332B
+$2B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $1,320B to $1,344B
US personal current taxes, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_personal_income_may_2026.v1pack-enabled
$3,268B
+$6B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $3,248B to $3,301B
US Social Security benefits, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_social_benefits_may_2026.v1pack-enabled
$1,651B
+$1B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $1,640B to $1,663B
US wages and salaries, May 2026
brier-1.shadowJun 20, 2026Brier-1 - packs
pack_set.bea_personal_income_may_2026.v1pack-enabled
$13,362B
+$17B vs no-pack

Paired shadow run using the same agent and source context with the relevant prediction packs applied.

80% interval $13,315B to $13,410B