Improvement on rolling cross-validation
▲ +27.1%
Ensemble_Horizon scored 0.8161 versus 1.1194 for Naive. Selection uses pre-holdout evidence only.
Selected output vs. CV-best naive baseline · Bias-Adjusted WAPE · lower is better
Improvement on untouched holdout
▼ -23.0%
The selected output scored 1.1049 versus 0.8980 for the same preselected baseline on unseen data.
Selected CV
0.8161
Ensemble_Horizon
Baseline CV
1.1194
Naive
Selected holdout
1.1049
12 periods, unseen
Best observed holdout
0.7947
stat__simpleexponentialsmoothingoptimized · diagnostic only
Individual series · forecast explorer
“All series” is summed from every series in the run. Individual exploration includes 30 representative high-volume and high-error series out of 2000; exports contain the displayed series only.
Holdout evaluation · summed across all series
Bias-Adjusted WAPE
—
lower is better
MAE
—
units per period
MASE
—
below 1 beats naive
Bias
—
Percentage bias
—
aim for 0%
Coverage
—
Demand type
—
Historical actuals
Holdout actuals
Selected forecast
Interval
Future forecast
Rolling CV windows
Range:
Select a series to inspect its forecast.
Open exact selected-series values
All model CV scores · Bias-Adjusted WAPE · lower is better
Ensemble_Horizon is the delivered output; the baseline marker is fixed before holdout review.
Open full accuracy and efficiency table
| Model | CV Bias-Adjusted WAPE | Holdout Bias-Adjusted WAPE | Holdout VN1 | Runtime | Pareto |
|---|---|---|---|---|---|
| stat__simpleexponentialsmoothingoptimized | 1.1383 | 0.7947 | 0.7947 | 126.12s | Pareto |
| stat__dynamicoptimizedtheta | 1.2120 | 0.7982 | 0.7982 | 126.12s | — |
| stat__tsb_alpha_0p7 | 1.0978 | 0.8042 | 0.8042 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p7 | 1.0981 | 0.8047 | 0.8047 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p8 | 1.1004 | 0.8073 | 0.8073 | 126.12s | — |
| stat__tsb_alpha_0p8 | 1.1002 | 0.8077 | 0.8077 | 126.12s | — |
| MLF_LGBM_Tweedie_1p1 | 2.1231 | 0.8124 | 0.8124 | 36.74s | Pareto |
| stat__tsb_alpha_0p6 | 1.1032 | 0.8164 | 0.8164 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p6 | 1.1035 | 0.8170 | 0.8170 | 126.12s | — |
| stat__tsb_alpha_0p5 | 1.1183 | 0.8244 | 0.8244 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p5 | 1.1186 | 0.8251 | 0.8251 | 126.12s | — |
| stat__holt | 1.4307 | 0.8286 | 0.8286 | 126.12s | — |
| stat__tsb_alpha_0p4 | 1.1454 | 0.8286 | 0.8286 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p4 | 1.1455 | 0.8292 | 0.8292 | 126.12s | — |
| stat__tsb_alpha_0p3 | 1.1905 | 0.8311 | 0.8311 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p3 | 1.1899 | 0.8315 | 0.8315 | 126.12s | — |
| stat__adida | 1.1914 | 0.8322 | 0.8322 | 126.12s | — |
| stat__imapa | 1.1927 | 0.8322 | 0.8322 | 126.12s | — |
| MLF_LGBM | 2.0885 | 0.8339 | 0.8339 | 36.74s | — |
| stat__dynamictheta | 1.1531 | 0.8349 | 0.8349 | 126.12s | — |
| stat__theta | 1.1717 | 0.8369 | 0.8369 | 126.12s | — |
| Basic_LGBM_prior120 | 2.0937 | 0.8381 | 0.8381 | 36.74s | — |
| stat__autotheta | 1.2170 | 0.8445 | 0.8445 | 126.12s | — |
| stat__optimizedtheta | 1.2193 | 0.8454 | 0.8454 | 126.12s | — |
| stat__simpleexponentialsmoothing_alpha_0p9 | 1.1083 | 0.8529 | 0.8529 | 126.12s | — |
| stat__tsb_alpha_0p9 | 1.1082 | 0.8531 | 0.8531 | 126.12s | — |
| MLF_LGBM_Tweedie_1p5 | 2.0399 | 0.8532 | 0.8532 | 36.74s | — |
| WindowAverage_8 | 1.2811 | 0.8540 | 0.8540 | 3.38s | Pareto |
| MLF_LGBM_Tweedie_1p9 | 1.7688 | 0.8720 | 0.8720 | 36.74s | — |
| stat__simpleexponentialsmoothing | 1.2760 | 0.8781 | 0.8781 | 126.12s | — |
| stat__tsb | 1.2794 | 0.8900 | 0.8900 | 126.12s | — |
| Naive | 1.1194 | 0.8980 | 0.8980 | 3.38s | — |
| baseline__naive | 1.1194 | 0.8980 | 0.8980 | 3.38s | — |
| stat__naive | 1.1194 | 0.8980 | 0.8980 | 126.12s | — |
| stat__crostonoptimized | 1.2005 | 0.9085 | 0.9085 | 126.12s | — |
| stat__randomwalkwithdrift | 1.2578 | 0.9105 | 0.9105 | 126.12s | — |
| SeasonalNaive_7 | 1.4406 | 0.9445 | 0.9445 | 3.38s | — |
| WindowAverage_13 | 1.3912 | 1.0700 | 1.0700 | 3.38s | — |
| baseline__moving_average | 1.3912 | 1.0700 | 1.0700 | 3.38s | — |
| stat__windowaverage | 1.3912 | 1.0700 | 1.0700 | 126.12s | — |
| stat__autoets | 1.3804 | 1.0711 | 1.0711 | 126.12s | — |
| Ensemble_Horizon | 0.8161 | 1.1049 | 1.1049 | 239.73s | — |
| MLForecast_Dispersion_LGBM | 0.9545 | 1.1409 | 1.1409 | 36.74s | — |
| MetaCategorical_LGBM | 0.9317 | 1.1475 | 1.1475 | 36.74s | — |
| Custom_CatBoost_MAE | 1.0038 | 1.1606 | 1.1606 | 36.74s | — |
| Basic_LGBM_reg90__SeasonalBlend_25 | 1.0822 | 1.1668 | 1.1668 | 36.74s | — |
| Basic_LGBM_reg90__SeasonalIndexBlend_25 | 1.1571 | 1.1668 | 1.1668 | 36.74s | — |
| Basic_LGBM_reg90 | 0.7480 | 1.1737 | 1.1737 | 36.74s | — |
| MetaCategorical_LGBM__SeasonalBlend_25 | 1.2201 | 1.1878 | 1.1878 | 36.74s | — |
| MLForecast_Dispersion_LGBM__SeasonalBlend_25 | 1.2379 | 1.1916 | 1.1916 | 36.74s | — |
| MLForecast_Dispersion_LGBM__SeasonalIndexBlend_25 | 1.3117 | 1.1916 | 1.1916 | 36.74s | — |
| SeasonalNaive_14 | 1.6309 | 1.2826 | 1.2826 | 3.38s | — |
| stat__simpleexponentialsmoothing_alpha_0p1 | 1.5455 | 1.2928 | 1.2928 | 126.12s | — |
| stat__tsb_alpha_0p1 | 1.5596 | 1.3496 | 1.3496 | 126.12s | — |
| Basic_LGBM_prior120__SeasonalBlend_25 | 2.0927 | 1.5886 | 1.5886 | 36.74s | — |
| Basic_LGBM_prior120__SeasonalIndexBlend_25 | 2.1709 | 1.5886 | 1.5886 | 36.74s | — |
| MLF_LGBM__SeasonalBlend_25 | 2.0886 | 1.5908 | 1.5908 | 36.74s | — |
| MLF_LGBM__SeasonalIndexBlend_25 | 2.1673 | 1.5908 | 1.5908 | 36.74s | — |
| stat__crostonsba | 1.5593 | 1.6658 | 1.6658 | 126.12s | — |
| baseline__seasonal_ma | 2.2706 | 1.7333 | 1.7333 | 3.38s | — |
| stat__mstl | 3.4329 | 1.7752 | 1.7752 | 126.12s | — |
| stat__crostonclassic | 1.6801 | 1.7834 | 1.7834 | 126.12s | — |
| MetaCategorical_LGBM__SeasonalBlend_50 | 1.5342 | 2.2721 | 2.2721 | 36.74s | — |
| stat__holtwinters | 1.5227 | 2.4056 | 2.4056 | 126.12s | — |
| SeasonalNaive_28 | 2.4917 | 2.4997 | 2.4997 | 3.38s | — |
| MLF_LGBM__SeasonalBlend_50 | 2.1061 | 2.5490 | 2.5490 | 36.74s | — |
| baseline__seasonal_blend | 2.1783 | 3.1478 | 3.1478 | 3.38s | — |
| HistoricAverage | 2.5874 | 4.1714 | 4.1714 | 3.38s | — |
| stat__historicaverage | 2.5874 | 4.1714 | 4.1714 | 126.12s | — |
| SameWeekLastYear_52 | 2.2055 | 4.6126 | 4.6126 | 3.38s | — |
| SeasonalNaive_52 | 2.2055 | 4.6126 | 4.6126 | 3.38s | — |
| baseline__seasonal_naive | 2.2055 | 4.6126 | 4.6126 | 3.38s | — |
| stat__seasonalnaive | 2.2055 | 4.6126 | 4.6126 | 126.12s | — |
| SeasonalWindowAverage_52_3 | 2.5447 | 4.6126 | 4.6126 | 3.38s | — |
| stat__seasonalexponentialsmoothing | 4.8487 | 4.6126 | 4.6126 | 126.12s | — |
| stat__seasonalexponentialsmoothingoptimized | 4.8487 | 4.6126 | 4.6126 | 126.12s | — |
Forecast error by step ahead · Bias-Adjusted WAPE
Bias-Adjusted WAPE rises 41% in the later half of the horizon.
Open exact horizon metrics
Highest-error displayed series on holdout · select to inspect
Ensemble composition · mean share across horizons
Each horizon uses its own optimized simplex. The bars summarize mean share only; forecasts never use the displayed average.
Open exact per-horizon weights
Configuration
| Granularity | W-MON |
| Forecast horizon | 12 periods |
| CV windows / step | 4 / 12 |
| Series count | 2000 |
| Displayed series | 30 plus total |
| Confidence levels | 80%, 95% |
| Primary metric | Bias-Adjusted WAPE |
| Requested ensemble | best_cv |
| Selected ensemble | slsqp_horizon |
| legacy_slsqp_horizon CV | 0.8745 |
| slsqp_horizon CV | 0.8736 · selected |
| median_horizon CV | 1.0937 |
| Models evaluated | 76 this run · 79 reference context |
| Pipeline runtime | 2m 16s |
Completed pipeline progress · 7 of 7 steps finished
Step durations show where this forecast spent its time. The customer-facing run-status API reports the same ordered steps while a run is active.
-
1
0.5sLoad and prepare data
-
2
0.2sRun rolling cross-validation
-
3
3.2sSelect and score the ensemble
-
4
1m 03sEvaluate the untouched holdout
-
5
1m 04sGenerate the production forecast
-
6
0.1sBuild the leaderboard and metadata
-
7
4.2sWrite artifacts and render the report
Diagnostics, run settings, and methodology
| Source rows | 174550 |
| Normalized rows | 174550 |
| Missing periods added | 0 |
| Stockout periods detected | 0 |
| Zero share | 24.0% |
| Coverage | 24 Sep 2018–22 Jun 2020 |
| Model profile | quick |
| Lags | 1, 2, 3, 4, 7 |
| Moving windows | 4, 8 |
| EWM alphas | 0.3 |
| Parallel families | On |
| Inner jobs | 4 |
| Conformal windows | 4 rolling windows |
| Conformal score | Forecast-scaled residual |
| Joint scenarios | 4 aligned residual paths |
| Interval aggregation | Pathwise reconciliation at every hierarchy depth |
| Item holdout CRPS | 17.7873 |
| Total holdout CRPS | 32578.9252 |
| Weight scope | separate_per_horizon |
| Holdout use | Final diagnostic only; never used for selection |
| Future forecast | Included |