HUMANITYVECTOR

Forecast lab

Forecasts

Modeled values, never measurements. Every forecast is an immutable record with its data cutoff, model, baseline comparison, drivers (statistical associations, not causes), assumptions and falsifiers, and is scored against the observation that later arrives. A model becomes champion only by beating the last-value baseline out of sample; where none does, the baseline is the champion and this page says so.

Modeled

models_v1, targets_v1. Families: last value, seasonal naive, drift, linear trend, Holt, ridge on lags and drivers. Intervals: empirical residual quantiles at 80 percent. Promotion: at least 5 percent lower MAE than last value over at least 24 monthly or 8 annual cutoffs. Full scorecard.

Annual targets

horizons 1, 2 and 5 years beyond the last observed year
WLD · annual

World population

Modeled
HorizonChampionSkill vs last valueLatest forecast
1 yearHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 95.7 percent over 16 cutoffs (minimum 8, margin 5 percent).
95.7% over 16 cutoffs, MAE 3,458,2492026: 8,291,408,074
80% 8,285,853,010 to 8,294,904,994 people · cutoff 2026-09-07
2 yearsHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 94.5 percent over 15 cutoffs (minimum 8, margin 5 percent).
94.5% over 15 cutoffs, MAE 8,984,9942027: 8,367,893,531
80% 8,345,970,059 to 8,375,440,699 people · cutoff 2026-09-07
5 yearsHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 91.2 percent over 12 cutoffs (minimum 8, margin 5 percent).
91.2% over 12 cutoffs, MAE 35,742,3002030: 8,601,609,314
80% 8,536,707,706 to 8,601,609,314 people · cutoff 2026-09-07
66 observations to 2025 · backtest data: pseudo real-time (first ingested values)
Modeled
HorizonChampionSkill vs last valueLatest forecast
1 yearRidge regression on lags and drivers champion
ridge beats last_value on out-of-sample MAE by 35.2 percent over 16 cutoffs (minimum 8, margin 5 percent).
35.2% over 16 cutoffs, MAE 0.27492025: 73.69
80% 72.84 to 73.83 years · cutoff 2026-09-07
2 yearsRidge regression on lags and drivers champion
ridge beats last_value on out-of-sample MAE by 41.1 percent over 15 cutoffs (minimum 8, margin 5 percent).
41.1% over 15 cutoffs, MAE 0.41832026: 73.89
80% 73.02 to 74.26 years · cutoff 2026-09-07
5 yearsHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 47.0 percent over 12 cutoffs (minimum 8, margin 5 percent).
47.0% over 12 cutoffs, MAE 0.52162029: 75.05
80% 74.24 to 75.05 years · cutoff 2026-09-07
65 observations to 2024 · backtest data: pseudo real-time (first ingested values)
WLD · annual

Cereal yield

Modeled
HorizonChampionSkill vs last valueLatest forecast
1 yearHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 13.1 percent over 16 cutoffs (minimum 8, margin 5 percent).
13.1% over 16 cutoffs, MAE 62.362025: 4,054
80% 3,939 to 4,093 kg per hectare · cutoff 2026-09-07
2 yearsRidge regression on lags and drivers champion
ridge beats last_value on out-of-sample MAE by 29.3 percent over 15 cutoffs (minimum 8, margin 5 percent).
29.3% over 15 cutoffs, MAE 72.742026: 4,026
80% 3,927 to 4,113 kg per hectare · cutoff 2026-09-07
5 yearsRidge regression on lags and drivers champion
ridge beats last_value on out-of-sample MAE by 44.4 percent over 12 cutoffs (minimum 8, margin 5 percent).
44.4% over 12 cutoffs, MAE 142.22029: 4,170
80% 4,012 to 4,364 kg per hectare · cutoff 2026-09-07
64 observations to 2024 · backtest data: pseudo real-time (first ingested values)
Modeled
HorizonChampionSkill vs last valueLatest forecast
1 yearHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 87.2 percent over 16 cutoffs (minimum 8, margin 5 percent).
87.2% over 16 cutoffs, MAE 0.31392026: 428.3
80% 428.1 to 429.3 ppm · cutoff 2026-09-07
2 yearsHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 88.5 percent over 15 cutoffs (minimum 8, margin 5 percent).
88.5% over 15 cutoffs, MAE 0.56322027: 431
80% 430.8 to 432.5 ppm · cutoff 2026-09-07
5 yearsHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 88.4 percent over 12 cutoffs (minimum 8, margin 5 percent).
88.4% over 12 cutoffs, MAE 1.412030: 439.2
80% 438.8 to 441.9 ppm · cutoff 2026-09-07
47 observations to 2025 · backtest data: pseudo real-time (first ingested values)
Modeled

No backtest has run for this target yet. Until one does, any forecast would use the last-value baseline; none is shown.

2 observations to 2025

Monthly targets

horizons 1, 3, 6 and 12 months beyond the last observed month
Modeled
HorizonChampionSkill vs last valueLatest forecast
1 monthLast value baseline is champion
Baseline is champion: the best challenger (ridge) beats last_value by -0.7 percent, below the 5 percent margin.
0.0% over 47 cutoffs, MAE 0.43162026-08: 103
80% 102.5 to 103.8 index 2017=100 · cutoff 2026-09-07
3 monthsLast value baseline is champion
Baseline is champion: the best challenger (ridge) beats last_value by -1.9 percent, below the 5 percent margin.
0.0% over 45 cutoffs, MAE 0.642026-10: 103
80% 102.4 to 104 index 2017=100 · cutoff 2026-09-07
6 monthsLast value baseline is champion
Baseline is champion: the best challenger (ridge) beats last_value by -18.8 percent, below the 5 percent margin.
0.0% over 42 cutoffs, MAE 0.65942027-01: 103
80% 102.2 to 104.6 index 2017=100 · cutoff 2026-09-07
12 monthsLast value baseline is champion
Baseline is champion: the best challenger (seasonal_naive) beats last_value by 0.0 percent, below the 5 percent margin.
0.0% over 36 cutoffs, MAE 0.91582027-07: 103
80% 102 to 104.7 index 2017=100 · cutoff 2026-09-07
1291 observations to 2026-07 · backtest data: pseudo real-time (first ingested values)
Modeled
HorizonChampionSkill vs last valueLatest forecast
1 monthSeasonal naive champion
seasonal_naive beats last_value on out-of-sample MAE by 65.0 percent over 47 cutoffs (minimum 24, margin 5 percent).
65.0% over 47 cutoffs, MAE 10,8512026-06: 393,285
80% 389,395 to 415,611 million kilowatthours · cutoff 2026-09-07
3 monthsSeasonal naive champion
seasonal_naive beats last_value on out-of-sample MAE by 79.2 percent over 45 cutoffs (minimum 24, margin 5 percent).
79.2% over 45 cutoffs, MAE 10,9312026-08: 420,242
80% 416,085 to 444,099 million kilowatthours · cutoff 2026-09-07
6 monthsSeasonal naive champion
seasonal_naive beats last_value on out-of-sample MAE by 70.6 percent over 42 cutoffs (minimum 24, margin 5 percent).
70.6% over 42 cutoffs, MAE 11,2922026-11: 334,986
80% 331,672 to 354,003 million kilowatthours · cutoff 2026-09-07
12 monthsDrift champion
drift beats last_value on out-of-sample MAE by 13.6 percent over 36 cutoffs (minimum 24, margin 5 percent).
13.6% over 36 cutoffs, MAE 9,3122027-05: 360,015
80% 350,484 to 374,292 million kilowatthours · cutoff 2026-09-07
641 observations to 2026-05 · backtest data: pseudo real-time (first ingested values)
USA · monthly

US housing starts

Modeled
HorizonChampionSkill vs last valueLatest forecast
1 monthHolt (exponential smoothing with additive trend) champion
holt beats last_value on out-of-sample MAE by 16.0 percent over 47 cutoffs (minimum 24, margin 5 percent).
16.0% over 47 cutoffs, MAE 79.792026-08: 1,269
80% 1,159 to 1,404 thousands of units, seasonally adjusted annual rate · cutoff 2026-09-07
3 monthsLast value baseline is champion
Baseline is champion: the best challenger (holt) beats last_value by 3.5 percent, below the 5 percent margin.
0.0% over 45 cutoffs, MAE 98.112026-10: 1,239
80% 1,078 to 1,363 thousands of units, seasonally adjusted annual rate · cutoff 2026-09-07
6 monthsLast value baseline is champion
Baseline is champion: the best challenger (holt) beats last_value by 2.0 percent, below the 5 percent margin.
0.0% over 42 cutoffs, MAE 93.792027-01: 1,239
80% 1,109 to 1,350 thousands of units, seasonally adjusted annual rate · cutoff 2026-09-07
12 monthsLast value baseline is champion
Baseline is champion: the best challenger (ridge) beats last_value by 2.2 percent, below the 5 percent margin.
0.0% over 36 cutoffs, MAE 93.692027-07: 1,239
80% 1,073 to 1,348 thousands of units, seasonally adjusted annual rate · cutoff 2026-09-07
811 observations to 2026-07 · backtest data: pseudo real-time (first ingested values)
WLD · monthly

Copper price

Modeled
HorizonChampionSkill vs last valueLatest forecast
1 monthLast value baseline is champion
Baseline is champion: the best challenger (drift) beats last_value by 0.4 percent, below the 5 percent margin.
0.0% over 47 cutoffs, MAE 317.92026-08: 13,543
80% 12,977 to 14,529 US dollars per metric ton · cutoff 2026-09-07
3 monthsDrift champion
drift beats last_value on out-of-sample MAE by 7.0 percent over 45 cutoffs (minimum 24, margin 5 percent).
7.0% over 45 cutoffs, MAE 566.12026-10: 13,725
80% 13,040 to 15,680 US dollars per metric ton · cutoff 2026-09-07
6 monthsDrift champion
drift beats last_value on out-of-sample MAE by 7.9 percent over 42 cutoffs (minimum 24, margin 5 percent).
7.9% over 42 cutoffs, MAE 895.62027-01: 13,909
80% 12,702 to 16,899 US dollars per metric ton · cutoff 2026-09-07
12 monthsDrift champion
drift beats last_value on out-of-sample MAE by 16.5 percent over 36 cutoffs (minimum 24, margin 5 percent).
16.5% over 36 cutoffs, MAE 1,1372027-07: 14,285
80% 13,324 to 18,946 US dollars per metric ton · cutoff 2026-09-07
415 observations to 2026-07 · backtest data: pseudo real-time (first ingested values)
Modeled
HorizonChampionSkill vs last valueLatest forecast
1 monthRidge regression on lags and drivers champion
ridge beats last_value on out-of-sample MAE by 47.4 percent over 46 cutoffs (minimum 24, margin 5 percent).
47.4% over 46 cutoffs, MAE 0.45942026-08: 333.5
80% 333 to 334.3 index 1982-1984=100 · cutoff 2026-09-07
3 monthsDrift champion
drift beats last_value on out-of-sample MAE by 63.9 percent over 44 cutoffs (minimum 24, margin 5 percent).
63.9% over 44 cutoffs, MAE 0.89192026-10: 335.7
80% 333.8 to 336.6 index 1982-1984=100 · cutoff 2026-09-07
6 monthsDrift champion
drift beats last_value on out-of-sample MAE by 78.7 percent over 41 cutoffs (minimum 24, margin 5 percent).
78.7% over 41 cutoffs, MAE 1.022027-01: 338.6
80% 336.2 to 339.1 index 1982-1984=100 · cutoff 2026-09-07
12 monthsRidge regression on lags and drivers champion
ridge beats last_value on out-of-sample MAE by 81.2 percent over 36 cutoffs (minimum 24, margin 5 percent).
81.2% over 36 cutoffs, MAE 1.752027-07: 341
80% 340.2 to 344.7 index 1982-1984=100 · cutoff 2026-09-07
954 observations to 2026-07 · backtest data: pseudo real-time (first ingested values)