Command Palette
Search for a command to run...
Backtesting
32 forecasting models — 8 pure mathematical baselines, a sentiment-only and a sentiment + crop-calendar variant of each (the same three families the live Forecasts page picks between), and 8 real library-grade models (AutoARIMA, Prophet, Chronos-Bolt, LightGBM…) — replayed walk-forward against real FRED history (24 months, no look-ahead) at 1, 2, and 3 months ahead. Toggle the horizon to compare.
Library-model snapshot: 74 days ago
Forecast horizon:(walk-forward over 24 months, no look-ahead)
Model leaderboard
Lower mean error = better. Sorted by Avg MAPE (lower better). The current production model is tagged. * = fewer than 300 total backtested points across all benchmarks — treat rank order with caution at this sample size.
Momentum-driftwins with 6.2% avg MAPE
At 1 month ahead, Momentum-drift has the lowest mean error (6.2% MAPE across 14 benchmarks) and won 3/14 outright. Accuracy typically degrades at longer horizons — toggle to compare.
| Rank | Model | Avg MAPE | Avg MASE | Avg Dir | Band | Wins | n |
|---|---|---|---|---|---|---|---|
Momentum-driftTS-nativecurrent 60% EWMA momentum (3-month weighted) + 40% structural drift, clamped ±15%. | 6.2% | — | 20.8% | 87.5% | 3/14 | 280* | |
| 2 | Momentum-drift + SentimentTS-native 60% EWMA momentum (3-month weighted) + 40% structural drift, clamped ±15%. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 6.3% | — | 30.5% | 85.2% | 4/15 | 299* |
| 3 | Momentum-drift + Sentiment + CalendarTS-native 60% EWMA momentum (3-month weighted) + 40% structural drift, clamped ±15%. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 6.3% | — | 33.2% | 85.2% | 3/15 | 299* |
| 4 | Chronos-BoltChronos Amazon's zero-shot pretrained time-series foundation model (amazon/chronos-bolt-small). No per-benchmark training — the same pretrained weights forecast every series. Band: native 10%/90% predicted quantiles (an 80% interval — Chronos-Bolt's trained quantile range tops out at 0.9, so bandCoverage for this model targets 80%, not the 95% other models use). | 6.3% | — | 24.1% | 84.6% | 2/14 | 266* |
| 5 | Auto-ARIMA + SentimentTS-native Automatic ARIMA — tries AR orders 0-2 on differenced series, picks the best by AIC. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 6.6% | — | 35.2% | 82.9% | 2/15 | 299* |
| 6 | AutoETS (StatsForecast)StatsForecast Automatic exponential smoothing (error/trend/seasonal). | 6.6% | — | 0% | 87.7% | 4/14 | 154* |
| 7 | Auto-ARIMATS-native Automatic ARIMA — tries AR orders 0-2 on differenced series, picks the best by AIC. | 6.7% | — | 31.6% | 87.1% | 1/14 | 280* |
| 8 | Naive (drift) + SentimentTS-native Random walk with drift — predicts last price plus the average monthly change over the window. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 6.7% | — | 37% | 83.9% | 1/15 | 299* |
| 9 | Auto-ARIMA + Sentiment + CalendarTS-native Automatic ARIMA — tries AR orders 0-2 on differenced series, picks the best by AIC. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 6.7% | — | 35.8% | 82.6% | 1/15 | 299* |
| 10 | Naive (drift)TS-native Random walk with drift — predicts last price plus the average monthly change over the window. | 6.8% | — | 33.2% | 86.4% | 1/14 | 280* |
| 11 | Moving avgTS-native Adaptive moving average — tries windows of 2/3/4/6 months and picks the best by in-sample error. | 6.9% | — | 28.9% | 83.9% | 0/14 | 280* |
| 12 | Naive (drift) + Sentiment + CalendarTS-native Random walk with drift — predicts last price plus the average monthly change over the window. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 6.9% | — | 38.4% | 82.9% | 2/15 | 299* |
| 13 | Theta (StatsForecast)StatsForecast Theta decomposition method. | 6.9% | — | 0% | 81.8% | 0/14 | 154* |
| 14 | Moving avg + SentimentTS-native Adaptive moving average — tries windows of 2/3/4/6 months and picks the best by in-sample error. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 7% | — | 29.1% | 81.5% | 2/15 | 299* |
| 15 | Moving avg + Sentiment + CalendarTS-native Adaptive moving average — tries windows of 2/3/4/6 months and picks the best by in-sample error. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 7.1% | — | 29.9% | 81.5% | 0/15 | 299* |
| 16 | LinReg (Darts)Darts Darts autoregressive linear model (lags=1). Band: 95% interval from in-sample lag-1 residual stddev (Darts has no native prediction interval for this model). | 7.4% | — | 30% | 85% | 1/14 | 266* |
| 17 | AutoARIMA (StatsForecast)StatsForecast Automatic ARIMA order selection (AIC). | 7.5% | — | 0% | 87% | 1/14 | 154* |
| 18 | WLS regressionTS-native Weighted least squares linear fit (exponential decay λ=0.95) — recent prices weighted more heavily. | 8% | — | 41.9% | 77.5% | 0/14 | 280* |
| 19 | WLS regression + SentimentTS-native Weighted least squares linear fit (exponential decay λ=0.95) — recent prices weighted more heavily. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 8.1% | — | 44.1% | 76.1% | 1/15 | 299* |
| 20 | WLS regression + Sentiment + CalendarTS-native Weighted least squares linear fit (exponential decay λ=0.95) — recent prices weighted more heavily. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 8.2% | — | 43.7% | 76.1% | 1/15 | 299* |
| 21 | Holt-Winters + SentimentTS-native Triple exponential smoothing — captures level, trend, and 12-month seasonality with grid-searched parameters. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 8.7% | — | 42.5% | 80.9% | 0/15 | 299* |
| 22 | Holt linear + SentimentTS-native Holt's linear trend method (level + slope), one-step ahead. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 8.7% | — | 42.5% | 80.9% | 0/15 | 299* |
| 23 | Holt-WintersTS-native Triple exponential smoothing — captures level, trend, and 12-month seasonality with grid-searched parameters. | 8.8% | — | 39.1% | 84.3% | 0/14 | 280* |
| 24 | Holt linearTS-native Holt's linear trend method (level + slope), one-step ahead. | 8.8% | — | 39.1% | 84.3% | 0/14 | 280* |
| 25 | Holt-Winters + Sentiment + CalendarTS-native Triple exponential smoothing — captures level, trend, and 12-month seasonality with grid-searched parameters. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 8.8% | — | 44% | 80.5% | 0/15 | 299* |
| 26 | Holt linear + Sentiment + CalendarTS-native Holt's linear trend method (level + slope), one-step ahead. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 8.8% | — | 44% | 80.5% | 0/15 | 299* |
| 27 | Seasonal naiveTS-native Predicts the price from one cycle ago — tries both 6-month and 12-month cycles, picks the best. | 9.7% | — | 17.5% | 75% | 0/14 | 280* |
| 28 | Seasonal naive + SentimentTS-native Predicts the price from one cycle ago — tries both 6-month and 12-month cycles, picks the best. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon only). | 10.1% | — | 18.4% | 72.5% | 0/15 | 299* |
| 29 | ProphetProphet Facebook Prophet — additive trend + seasonality with uncertainty. Walk-forward every 3rd month (offline run-time constraint). | 10.1% | — | 36.3% | 70% | 0/14 | 70* |
| 30 | Seasonal naive + Sentiment + CalendarTS-native Predicts the price from one cycle ago — tries both 6-month and 12-month cycles, picks the best. + real GDELT news-sentiment nudge (grid-searched weight, 1-month horizon) + crop-calendar seasonality across Hassad's sourcing regions (fixed weight, all horizons). | 10.2% | — | 22.7% | 72.1% | 0/15 | 299* |
| 31 | SeasonalNaive (StatsForecast)StatsForecast Forecast = last observed value from the same season. | 16.3% | — | 0% | 75.9% | 0/14 | 154* |
| 32 | LightGBM (MLForecast)MLForecast Gradient-boosted trees (LightGBM) on lag/rolling features via Nixtla's MLForecast. Band: 95% interval from in-sample one-step residual stddev. | 31% | — | 40.3% | 79.6% | 1/14 | 98* |
Model accuracy matrix · 1 month ahead
MAPE % per benchmark × model (lower is better). ★ = that model won the benchmark (lowest MAPE in the row) at this horizon.
| Benchmark | Naive (drift)TS-native | Seasonal naiveTS-native | Moving avgTS-native | Holt-WintersTS-native | Holt linearTS-native | WLS regressionTS-native | Auto-ARIMATS-native | Momentum-driftcurrent · TS-native | Naive (drift) + SentimentTS-native | Seasonal naive + SentimentTS-native | Moving avg + SentimentTS-native | Holt-Winters + SentimentTS-native | Holt linear + SentimentTS-native | WLS regression + SentimentTS-native | Auto-ARIMA + SentimentTS-native | Momentum-drift + SentimentTS-native | Naive (drift) + Sentiment + CalendarTS-native | Seasonal naive + Sentiment + CalendarTS-native | Moving avg + Sentiment + CalendarTS-native | Holt-Winters + Sentiment + CalendarTS-native | Holt linear + Sentiment + CalendarTS-native | WLS regression + Sentiment + CalendarTS-native | Auto-ARIMA + Sentiment + CalendarTS-native | Momentum-drift + Sentiment + CalendarTS-native | AutoARIMA (StatsForecast)StatsForecast | AutoETS (StatsForecast)StatsForecast | Theta (StatsForecast)StatsForecast | SeasonalNaive (StatsForecast)StatsForecast | ProphetProphet | LinReg (Darts)Darts | Chronos-BoltChronos | LightGBM (MLForecast)MLForecast |
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Alfalfa Hay (US PPI Index) Feed Additives & Forage | 9.1% | 9.1% | 7.1% | 10.4% | 10.4% | 8.2% | 10.9% | 8.9% | 8.9% | 8.9% | 7% | 10.2% | 10.2% | 8% | 10.7% | 8.7% | 9.3% | 9.1% | 7.4% | 10.7% | 10.7% | 8.4% | 11% | 8.9% | 7.8% | 7.8% | 8.1% | 10.6% | 10.2% | 7.5% | 7.1% | 14.1% |
| Australia Chilled Meat Export Protein | 2.7% | 5.6% | 3.7% | 2.8% | 2.8% | 5.1% | 2.8% | 2.5% | 2.7% | 5.2% | 3.5% | 2.8% | 2.8% | 5.1% | 2.8% | 2.5% | 3.1% | 5.4% | 3.7% | 3% | 3% | 5.3% | 3.1% | 2.8% | 2.4% | 1.7% | 1.9% | 27.5% | 4.9% | 4.2% | 3.5% | 2.2% |
| Barley (IMF Global Price) Grains, Oilseeds & Fats | 3.5% | 3.7% | 4% | 3.9% | 3.9% | 5.7% | 4% | 3% | 3.2% | 3.6% | 3.9% | 3.9% | 3.9% | 5.1% | 3.7% | 3% | 3.5% | 3.6% | 4% | 3.9% | 3.9% | 5.3% | 3.9% | 3.2% | 1.7% | 1.8% | 2.7% | 8.8% | 4.3% | 3.6% | 2.9% | 3.6% |
| Brazil Chilled/Frozen Export Protein | — | — | — | — | — | — | — | — | 7.6% | 18.1% | 10.8% | 9.3% | 9.3% | 11.6% | 7.5% | 7.8% | 7.1% | 17.9% | 10.7% | 9% | 9% | 11.1% | 7.3% | 7.3% | — | — | — | — | — | — | — | — |
| Broiler Chicken (IMF Poultry Proxy) Protein | 1.4% | 2.7% | 1.6% | 1.7% | 1.7% | 1.8% | 1.4% | 1.4% | 1.4% | 2.7% | 1.6% | 1.7% | 1.7% | 1.8% | 1.4% | 1.4% | 1.4% | 2.7% | 1.6% | 1.7% | 1.7% | 1.8% | 1.4% | 1.4% | 1.9% | 1.7% | 1.8% | 3.9% | 2.5% | 1.7% | 1.5% | 3.8% |
| Butter (US PPI Proxy) Protein | 9.1% | 14.3% | 9.2% | 10.6% | 10.6% | 14.1% | 7.9% | 7.2% | 8.8% | 14.3% | 9.2% | 10.4% | 10.4% | 13.6% | 7.6% | 7.2% | 9% | 14.3% | 9.4% | 10.5% | 10.5% | 14.1% | 7.9% | 7.2% | 10.7% | 10.4% | 11.7% | 30.6% | 18.6% | 10.9% | 8.5% | 10.4% |
| Corn (IMF Global Price) Grains, Oilseeds & Fats | 2.9% | 4.3% | 4.2% | 3% | 3% | 4.6% | 2.9% | 2.8% | 2.9% | 4.3% | 4% | 3% | 3% | 4.5% | 2.9% | 2.8% | 3.1% | 4.5% | 4.2% | 3.2% | 3.2% | 4.4% | 3.2% | 3% | 3.3% | 2.7% | 2.8% | 4.7% | 4.6% | 4.3% | 3.3% | 3% |
| Dairy Products Price Index (GDT Proxy) Protein | 1.1% | 2.4% | 1.7% | 1.1% | 1.1% | 1.8% | 1.1% | 1% | 1.1% | 2.4% | 1.7% | 1.1% | 1.1% | 1.8% | 1% | 1% | 1.6% | 2.6% | 2.1% | 1.8% | 1.8% | 2.4% | 1.5% | 1.5% | 1.5% | 1.3% | 1.3% | 4.3% | 2% | 1.5% | 1% | 1.3% |
| Egg Price Index (US PPI Proxy) Protein | 31.3% | 36.4% | 27.6% | 53.6% | 53.6% | 30.9% | 29% | 27.2% | 31% | 36.4% | 27.6% | 53.2% | 53.2% | 30.9% | 28.7% | 27.1% | 31% | 36.4% | 27.6% | 53.2% | 53.2% | 30.9% | 28.7% | 27.1% | 35.5% | 25.8% | 26.9% | 51% | 41.2% | 32.1% | 26.8% | 345.4% |
| Lean Hogs (IMF Swine Proxy) Protein | 4.9% | 4.8% | 5.8% | 6.4% | 6.4% | 7.4% | 4.6% | 4.4% | 4.9% | 4.6% | 5.6% | 6.3% | 6.3% | 7.4% | 4.4% | 4.3% | 4.9% | 4.6% | 5.6% | 6.3% | 6.3% | 7.4% | 4.4% | 4.3% | 6.8% | 5.7% | 6.5% | 9.2% | 11.2% | 5.2% | 4.5% | 4.8% |
| Live Cattle (IMF Beef Proxy) Protein | 1.5% | 6.2% | 2.1% | 1.7% | 1.7% | 1.6% | 1.5% | 1.5% | 1.4% | 5.7% | 1.8% | 1.6% | 1.6% | 1.3% | 1.4% | 1.4% | 1.4% | 5.7% | 1.8% | 1.6% | 1.6% | 1.3% | 1.4% | 1.4% | 1.9% | 1.8% | 1.8% | 16.8% | 1.4% | 1.7% | 2% | 2.1% |
| Skim Milk Powder (US PPI Proxy) Protein | 2.6% | 3.8% | 3.3% | 3.2% | 3.2% | 3.3% | 2.4% | 2.5% | 2.6% | 3.6% | 3% | 3.2% | 3.2% | 3.2% | 2.3% | 2.5% | 2.6% | 3.6% | 3% | 3.2% | 3.2% | 3.1% | 2.1% | 2.6% | 2.1% | 2% | 2.6% | 3.7% | 2.1% | 3.8% | 2.3% | 3.3% |
| Soybean Meal (IMF Global Price) Grains, Oilseeds & Fats | 6% | 6.3% | 6.3% | 6% | 6% | 6% | 6.4% | 5.7% | 5.8% | 6.3% | 6.2% | 6% | 6% | 5.8% | 6.3% | 5.7% | 6.1% | 6.1% | 6.3% | 6.1% | 6.1% | 6.1% | 6.5% | 5.7% | 8.8% | 8.4% | 7.5% | 12.7% | 13.4% | 5.5% | 5.5% | 13% |
| Urea / NPN (US PPI Proxy) Feed Additives & Forage | 3.4% | 13.2% | 5.1% | 3.3% | 3.3% | 5.3% | 3.4% | 3.2% | 3.2% | 12.7% | 4.5% | 3.2% | 3.2% | 5.1% | 3.3% | 3% | 3.2% | 12.7% | 4.5% | 3.2% | 3.2% | 5.1% | 3.3% | 3% | 2.8% | 3.6% | 3.4% | 24.1% | 3.8% | 3.4% | 3.9% | 2.4% |
| Vegetable Price Index (US PPI Proxy) Fruits, Vegetables & Juices | 15.5% | 22.8% | 14.3% | 15.1% | 15.1% | 16.1% | 15.3% | 15.5% | 15.3% | 22.7% | 13.9% | 14.9% | 14.9% | 15.7% | 15.1% | 15.4% | 15.5% | 23.2% | 14.1% | 15.1% | 15.1% | 15.8% | 15.1% | 15.7% | 18.1% | 17% | 18.1% | 20.2% | 21.5% | 18.7% | 15.1% | 24.1% |
Actual vs predicted · 1 month ahead
FRED / BLS PPI (alfalfa hay) · drift 4%/yr · vol 2.5%
Naive (drift) 9.1%(n=20)Seasonal naive 9.1%(n=20)Moving avg 7.1%(n=20)Holt-Winters 10.4%(n=20)Holt linear 10.4%(n=20)WLS regression 8.2%(n=20)Auto-ARIMA 10.9%(n=20)Momentum-drift 8.9%(n=20)Naive (drift) + Sentiment 8.9%(n=20)Seasonal naive + Sentiment 8.9%(n=20)Moving avg + Sentiment 7%(n=20)Holt-Winters + Sentiment 10.2%(n=20)Holt linear + Sentiment 10.2%(n=20)WLS regression + Sentiment 8%(n=20)Auto-ARIMA + Sentiment 10.7%(n=20)Momentum-drift + Sentiment 8.7%(n=20)Naive (drift) + Sentiment + Calendar 9.3%(n=20)Seasonal naive + Sentiment + Calendar 9.1%(n=20)Moving avg + Sentiment + Calendar 7.4%(n=20)Holt-Winters + Sentiment + Calendar 10.7%(n=20)Holt linear + Sentiment + Calendar 10.7%(n=20)WLS regression + Sentiment + Calendar 8.4%(n=20)Auto-ARIMA + Sentiment + Calendar 11%(n=20)Momentum-drift + Sentiment + Calendar 8.9%(n=20)AutoARIMA (StatsForecast) 7.8%(n=11)AutoETS (StatsForecast) 7.8%(n=11)Theta (StatsForecast) 8.1%(n=11)SeasonalNaive (StatsForecast) 10.6%(n=11)Prophet 10.2%(n=5)LinReg (Darts) 7.5%(n=19)Chronos-Bolt 7.1%(n=19)LightGBM (MLForecast) 14.1%(n=7)
FRED / IMF (global lamb) · drift 5%/yr · vol 2%
Naive (drift) 2.7%(n=20)Seasonal naive 5.6%(n=20)Moving avg 3.7%(n=20)Holt-Winters 2.8%(n=20)Holt linear 2.8%(n=20)WLS regression 5.1%(n=20)Auto-ARIMA 2.8%(n=20)Momentum-drift 2.5%(n=20)Naive (drift) + Sentiment 2.7%(n=20)Seasonal naive + Sentiment 5.2%(n=20)Moving avg + Sentiment 3.5%(n=20)Holt-Winters + Sentiment 2.8%(n=20)Holt linear + Sentiment 2.8%(n=20)WLS regression + Sentiment 5.1%(n=20)Auto-ARIMA + Sentiment 2.8%(n=20)Momentum-drift + Sentiment 2.5%(n=20)Naive (drift) + Sentiment + Calendar 3.1%(n=20)Seasonal naive + Sentiment + Calendar 5.4%(n=20)Moving avg + Sentiment + Calendar 3.7%(n=20)Holt-Winters + Sentiment + Calendar 3%(n=20)Holt linear + Sentiment + Calendar 3%(n=20)WLS regression + Sentiment + Calendar 5.3%(n=20)Auto-ARIMA + Sentiment + Calendar 3.1%(n=20)Momentum-drift + Sentiment + Calendar 2.8%(n=20)AutoARIMA (StatsForecast) 2.4%(n=11)AutoETS (StatsForecast) 1.7%(n=11)Theta (StatsForecast) 1.9%(n=11)SeasonalNaive (StatsForecast) 27.5%(n=11)Prophet 4.9%(n=5)LinReg (Darts) 4.2%(n=19)Chronos-Bolt 3.5%(n=19)LightGBM (MLForecast) 2.2%(n=7)
FRED / IMF (global barley, proxy) · drift -1%/yr · vol 2.5%
Naive (drift) 3.5%(n=20)Seasonal naive 3.7%(n=20)Moving avg 4%(n=20)Holt-Winters 3.9%(n=20)Holt linear 3.9%(n=20)WLS regression 5.7%(n=20)Auto-ARIMA 4%(n=20)Momentum-drift 3%(n=20)Naive (drift) + Sentiment 3.2%(n=20)Seasonal naive + Sentiment 3.6%(n=20)Moving avg + Sentiment 3.9%(n=20)Holt-Winters + Sentiment 3.9%(n=20)Holt linear + Sentiment 3.9%(n=20)WLS regression + Sentiment 5.1%(n=20)Auto-ARIMA + Sentiment 3.7%(n=20)Momentum-drift + Sentiment 3%(n=20)Naive (drift) + Sentiment + Calendar 3.5%(n=20)Seasonal naive + Sentiment + Calendar 3.6%(n=20)Moving avg + Sentiment + Calendar 4%(n=20)Holt-Winters + Sentiment + Calendar 3.9%(n=20)Holt linear + Sentiment + Calendar 3.9%(n=20)WLS regression + Sentiment + Calendar 5.3%(n=20)Auto-ARIMA + Sentiment + Calendar 3.9%(n=20)Momentum-drift + Sentiment + Calendar 3.2%(n=20)AutoARIMA (StatsForecast) 1.7%(n=11)AutoETS (StatsForecast) 1.8%(n=11)Theta (StatsForecast) 2.7%(n=11)SeasonalNaive (StatsForecast) 8.8%(n=11)Prophet 4.3%(n=5)LinReg (Darts) 3.6%(n=19)Chronos-Bolt 2.9%(n=19)LightGBM (MLForecast) 3.6%(n=7)
FRED / IMF (global poultry, quarterly) · drift -2%/yr · vol 2%
Naive (drift) + Sentiment 7.6%(n=19)Seasonal naive + Sentiment 18.1%(n=19)Moving avg + Sentiment 10.8%(n=19)Holt-Winters + Sentiment 9.3%(n=19)Holt linear + Sentiment 9.3%(n=19)WLS regression + Sentiment 11.6%(n=19)Auto-ARIMA + Sentiment 7.5%(n=19)Momentum-drift + Sentiment 7.8%(n=19)Naive (drift) + Sentiment + Calendar 7.1%(n=19)Seasonal naive + Sentiment + Calendar 17.9%(n=19)Moving avg + Sentiment + Calendar 10.7%(n=19)Holt-Winters + Sentiment + Calendar 9%(n=19)Holt linear + Sentiment + Calendar 9%(n=19)WLS regression + Sentiment + Calendar 11.1%(n=19)Auto-ARIMA + Sentiment + Calendar 7.3%(n=19)Momentum-drift + Sentiment + Calendar 7.3%(n=19)
FRED / IMF (global poultry, proxy) · drift 3%/yr · vol 2%
Naive (drift) 1.4%(n=20)Seasonal naive 2.7%(n=20)Moving avg 1.6%(n=20)Holt-Winters 1.7%(n=20)Holt linear 1.7%(n=20)WLS regression 1.8%(n=20)Auto-ARIMA 1.4%(n=20)Momentum-drift 1.4%(n=20)Naive (drift) + Sentiment 1.4%(n=20)Seasonal naive + Sentiment 2.7%(n=20)Moving avg + Sentiment 1.6%(n=20)Holt-Winters + Sentiment 1.7%(n=20)Holt linear + Sentiment 1.7%(n=20)WLS regression + Sentiment 1.8%(n=20)Auto-ARIMA + Sentiment 1.4%(n=20)Momentum-drift + Sentiment 1.4%(n=20)Naive (drift) + Sentiment + Calendar 1.4%(n=20)Seasonal naive + Sentiment + Calendar 2.7%(n=20)Moving avg + Sentiment + Calendar 1.6%(n=20)Holt-Winters + Sentiment + Calendar 1.7%(n=20)Holt linear + Sentiment + Calendar 1.7%(n=20)WLS regression + Sentiment + Calendar 1.8%(n=20)Auto-ARIMA + Sentiment + Calendar 1.4%(n=20)Momentum-drift + Sentiment + Calendar 1.4%(n=20)AutoARIMA (StatsForecast) 1.9%(n=11)AutoETS (StatsForecast) 1.7%(n=11)Theta (StatsForecast) 1.8%(n=11)SeasonalNaive (StatsForecast) 3.9%(n=11)Prophet 2.5%(n=5)LinReg (Darts) 1.7%(n=19)Chronos-Bolt 1.5%(n=19)LightGBM (MLForecast) 3.8%(n=7)
FRED / BLS PPI (butter proxy) · drift 8%/yr · vol 3.5%
Naive (drift) 9.1%(n=20)Seasonal naive 14.3%(n=20)Moving avg 9.2%(n=20)Holt-Winters 10.6%(n=20)Holt linear 10.6%(n=20)WLS regression 14.1%(n=20)Auto-ARIMA 7.9%(n=20)Momentum-drift 7.2%(n=20)Naive (drift) + Sentiment 8.8%(n=20)Seasonal naive + Sentiment 14.3%(n=20)Moving avg + Sentiment 9.2%(n=20)Holt-Winters + Sentiment 10.4%(n=20)Holt linear + Sentiment 10.4%(n=20)WLS regression + Sentiment 13.6%(n=20)Auto-ARIMA + Sentiment 7.6%(n=20)Momentum-drift + Sentiment 7.2%(n=20)Naive (drift) + Sentiment + Calendar 9%(n=20)Seasonal naive + Sentiment + Calendar 14.3%(n=20)Moving avg + Sentiment + Calendar 9.4%(n=20)Holt-Winters + Sentiment + Calendar 10.5%(n=20)Holt linear + Sentiment + Calendar 10.5%(n=20)WLS regression + Sentiment + Calendar 14.1%(n=20)Auto-ARIMA + Sentiment + Calendar 7.9%(n=20)Momentum-drift + Sentiment + Calendar 7.2%(n=20)AutoARIMA (StatsForecast) 10.7%(n=11)AutoETS (StatsForecast) 10.4%(n=11)Theta (StatsForecast) 11.7%(n=11)SeasonalNaive (StatsForecast) 30.6%(n=11)Prophet 18.6%(n=5)LinReg (Darts) 10.9%(n=19)Chronos-Bolt 8.5%(n=19)LightGBM (MLForecast) 10.4%(n=7)
USD/bu
FRED / IMF (global maize, proxy) · drift -5%/yr · vol 3%
Naive (drift) 2.9%(n=20)Seasonal naive 4.3%(n=20)Moving avg 4.2%(n=20)Holt-Winters 3%(n=20)Holt linear 3%(n=20)WLS regression 4.6%(n=20)Auto-ARIMA 2.9%(n=20)Momentum-drift 2.8%(n=20)Naive (drift) + Sentiment 2.9%(n=20)Seasonal naive + Sentiment 4.3%(n=20)Moving avg + Sentiment 4%(n=20)Holt-Winters + Sentiment 3%(n=20)Holt linear + Sentiment 3%(n=20)WLS regression + Sentiment 4.5%(n=20)Auto-ARIMA + Sentiment 2.9%(n=20)Momentum-drift + Sentiment 2.8%(n=20)Naive (drift) + Sentiment + Calendar 3.1%(n=20)Seasonal naive + Sentiment + Calendar 4.5%(n=20)Moving avg + Sentiment + Calendar 4.2%(n=20)Holt-Winters + Sentiment + Calendar 3.2%(n=20)Holt linear + Sentiment + Calendar 3.2%(n=20)WLS regression + Sentiment + Calendar 4.4%(n=20)Auto-ARIMA + Sentiment + Calendar 3.2%(n=20)Momentum-drift + Sentiment + Calendar 3%(n=20)AutoARIMA (StatsForecast) 3.3%(n=11)AutoETS (StatsForecast) 2.7%(n=11)Theta (StatsForecast) 2.8%(n=11)SeasonalNaive (StatsForecast) 4.7%(n=11)Prophet 4.6%(n=5)LinReg (Darts) 4.3%(n=19)Chronos-Bolt 3.3%(n=19)LightGBM (MLForecast) 3%(n=7)
FRED / BLS PPI (dairy proxy) · drift 6%/yr · vol 3%
Naive (drift) 1.1%(n=20)Seasonal naive 2.4%(n=20)Moving avg 1.7%(n=20)Holt-Winters 1.1%(n=20)Holt linear 1.1%(n=20)WLS regression 1.8%(n=20)Auto-ARIMA 1.1%(n=20)Momentum-drift 1%(n=20)Naive (drift) + Sentiment 1.1%(n=20)Seasonal naive + Sentiment 2.4%(n=20)Moving avg + Sentiment 1.7%(n=20)Holt-Winters + Sentiment 1.1%(n=20)Holt linear + Sentiment 1.1%(n=20)WLS regression + Sentiment 1.8%(n=20)Auto-ARIMA + Sentiment 1%(n=20)Momentum-drift + Sentiment 1%(n=20)Naive (drift) + Sentiment + Calendar 1.6%(n=20)Seasonal naive + Sentiment + Calendar 2.6%(n=20)Moving avg + Sentiment + Calendar 2.1%(n=20)Holt-Winters + Sentiment + Calendar 1.8%(n=20)Holt linear + Sentiment + Calendar 1.8%(n=20)WLS regression + Sentiment + Calendar 2.4%(n=20)Auto-ARIMA + Sentiment + Calendar 1.5%(n=20)Momentum-drift + Sentiment + Calendar 1.5%(n=20)AutoARIMA (StatsForecast) 1.5%(n=11)AutoETS (StatsForecast) 1.3%(n=11)Theta (StatsForecast) 1.3%(n=11)SeasonalNaive (StatsForecast) 4.3%(n=11)Prophet 2%(n=5)LinReg (Darts) 1.5%(n=19)Chronos-Bolt 1%(n=19)LightGBM (MLForecast) 1.3%(n=7)
FRED / BLS PPI (eggs proxy) · drift -1%/yr · vol 2%
Naive (drift) 31.3%(n=20)Seasonal naive 36.4%(n=20)Moving avg 27.6%(n=20)Holt-Winters 53.6%(n=20)Holt linear 53.6%(n=20)WLS regression 30.9%(n=20)Auto-ARIMA 29%(n=20)Momentum-drift 27.2%(n=20)Naive (drift) + Sentiment 31%(n=20)Seasonal naive + Sentiment 36.4%(n=20)Moving avg + Sentiment 27.6%(n=20)Holt-Winters + Sentiment 53.2%(n=20)Holt linear + Sentiment 53.2%(n=20)WLS regression + Sentiment 30.9%(n=20)Auto-ARIMA + Sentiment 28.7%(n=20)Momentum-drift + Sentiment 27.1%(n=20)Naive (drift) + Sentiment + Calendar 31%(n=20)Seasonal naive + Sentiment + Calendar 36.4%(n=20)Moving avg + Sentiment + Calendar 27.6%(n=20)Holt-Winters + Sentiment + Calendar 53.2%(n=20)Holt linear + Sentiment + Calendar 53.2%(n=20)WLS regression + Sentiment + Calendar 30.9%(n=20)Auto-ARIMA + Sentiment + Calendar 28.7%(n=20)Momentum-drift + Sentiment + Calendar 27.1%(n=20)AutoARIMA (StatsForecast) 35.5%(n=11)AutoETS (StatsForecast) 25.8%(n=11)Theta (StatsForecast) 26.9%(n=11)SeasonalNaive (StatsForecast) 51%(n=11)Prophet 41.2%(n=5)LinReg (Darts) 32.1%(n=19)Chronos-Bolt 26.8%(n=19)LightGBM (MLForecast) 345.4%(n=7)
FRED / IMF (global swine, proxy) · drift -3%/yr · vol 3%
Naive (drift) 4.9%(n=20)Seasonal naive 4.8%(n=20)Moving avg 5.8%(n=20)Holt-Winters 6.4%(n=20)Holt linear 6.4%(n=20)WLS regression 7.4%(n=20)Auto-ARIMA 4.6%(n=20)Momentum-drift 4.4%(n=20)Naive (drift) + Sentiment 4.9%(n=20)Seasonal naive + Sentiment 4.6%(n=20)Moving avg + Sentiment 5.6%(n=20)Holt-Winters + Sentiment 6.3%(n=20)Holt linear + Sentiment 6.3%(n=20)WLS regression + Sentiment 7.4%(n=20)Auto-ARIMA + Sentiment 4.4%(n=20)Momentum-drift + Sentiment 4.3%(n=20)Naive (drift) + Sentiment + Calendar 4.9%(n=20)Seasonal naive + Sentiment + Calendar 4.6%(n=20)Moving avg + Sentiment + Calendar 5.6%(n=20)Holt-Winters + Sentiment + Calendar 6.3%(n=20)Holt linear + Sentiment + Calendar 6.3%(n=20)WLS regression + Sentiment + Calendar 7.4%(n=20)Auto-ARIMA + Sentiment + Calendar 4.4%(n=20)Momentum-drift + Sentiment + Calendar 4.3%(n=20)AutoARIMA (StatsForecast) 6.8%(n=11)AutoETS (StatsForecast) 5.7%(n=11)Theta (StatsForecast) 6.5%(n=11)SeasonalNaive (StatsForecast) 9.2%(n=11)Prophet 11.2%(n=5)LinReg (Darts) 5.2%(n=19)Chronos-Bolt 4.5%(n=19)LightGBM (MLForecast) 4.8%(n=7)
FRED / IMF (global beef, proxy) · drift 4%/yr · vol 2.5%
Naive (drift) 1.5%(n=20)Seasonal naive 6.2%(n=20)Moving avg 2.1%(n=20)Holt-Winters 1.7%(n=20)Holt linear 1.7%(n=20)WLS regression 1.6%(n=20)Auto-ARIMA 1.5%(n=20)Momentum-drift 1.5%(n=20)Naive (drift) + Sentiment 1.4%(n=20)Seasonal naive + Sentiment 5.7%(n=20)Moving avg + Sentiment 1.8%(n=20)Holt-Winters + Sentiment 1.6%(n=20)Holt linear + Sentiment 1.6%(n=20)WLS regression + Sentiment 1.3%(n=20)Auto-ARIMA + Sentiment 1.4%(n=20)Momentum-drift + Sentiment 1.4%(n=20)Naive (drift) + Sentiment + Calendar 1.4%(n=20)Seasonal naive + Sentiment + Calendar 5.7%(n=20)Moving avg + Sentiment + Calendar 1.8%(n=20)Holt-Winters + Sentiment + Calendar 1.6%(n=20)Holt linear + Sentiment + Calendar 1.6%(n=20)WLS regression + Sentiment + Calendar 1.3%(n=20)Auto-ARIMA + Sentiment + Calendar 1.4%(n=20)Momentum-drift + Sentiment + Calendar 1.4%(n=20)AutoARIMA (StatsForecast) 1.9%(n=11)AutoETS (StatsForecast) 1.8%(n=11)Theta (StatsForecast) 1.8%(n=11)SeasonalNaive (StatsForecast) 16.8%(n=11)Prophet 1.4%(n=5)LinReg (Darts) 1.7%(n=19)Chronos-Bolt 2%(n=19)LightGBM (MLForecast) 2.1%(n=7)
FRED / BLS PPI (skim milk proxy) · drift -2%/yr · vol 2.5%
Naive (drift) 2.6%(n=20)Seasonal naive 3.8%(n=20)Moving avg 3.3%(n=20)Holt-Winters 3.2%(n=20)Holt linear 3.2%(n=20)WLS regression 3.3%(n=20)Auto-ARIMA 2.4%(n=20)Momentum-drift 2.5%(n=20)Naive (drift) + Sentiment 2.6%(n=20)Seasonal naive + Sentiment 3.6%(n=20)Moving avg + Sentiment 3%(n=20)Holt-Winters + Sentiment 3.2%(n=20)Holt linear + Sentiment 3.2%(n=20)WLS regression + Sentiment 3.2%(n=20)Auto-ARIMA + Sentiment 2.3%(n=20)Momentum-drift + Sentiment 2.5%(n=20)Naive (drift) + Sentiment + Calendar 2.6%(n=20)Seasonal naive + Sentiment + Calendar 3.6%(n=20)Moving avg + Sentiment + Calendar 3%(n=20)Holt-Winters + Sentiment + Calendar 3.2%(n=20)Holt linear + Sentiment + Calendar 3.2%(n=20)WLS regression + Sentiment + Calendar 3.1%(n=20)Auto-ARIMA + Sentiment + Calendar 2.1%(n=20)Momentum-drift + Sentiment + Calendar 2.6%(n=20)AutoARIMA (StatsForecast) 2.1%(n=11)AutoETS (StatsForecast) 2%(n=11)Theta (StatsForecast) 2.6%(n=11)SeasonalNaive (StatsForecast) 3.7%(n=11)Prophet 2.1%(n=5)LinReg (Darts) 3.8%(n=19)Chronos-Bolt 2.3%(n=19)LightGBM (MLForecast) 3.3%(n=7)
FRED / IMF (global soybean meal, proxy) · drift 2%/yr · vol 3%
Naive (drift) 6%(n=20)Seasonal naive 6.3%(n=20)Moving avg 6.3%(n=20)Holt-Winters 6%(n=20)Holt linear 6%(n=20)WLS regression 6%(n=20)Auto-ARIMA 6.4%(n=20)Momentum-drift 5.7%(n=20)Naive (drift) + Sentiment 5.8%(n=20)Seasonal naive + Sentiment 6.3%(n=20)Moving avg + Sentiment 6.2%(n=20)Holt-Winters + Sentiment 6%(n=20)Holt linear + Sentiment 6%(n=20)WLS regression + Sentiment 5.8%(n=20)Auto-ARIMA + Sentiment 6.3%(n=20)Momentum-drift + Sentiment 5.7%(n=20)Naive (drift) + Sentiment + Calendar 6.1%(n=20)Seasonal naive + Sentiment + Calendar 6.1%(n=20)Moving avg + Sentiment + Calendar 6.3%(n=20)Holt-Winters + Sentiment + Calendar 6.1%(n=20)Holt linear + Sentiment + Calendar 6.1%(n=20)WLS regression + Sentiment + Calendar 6.1%(n=20)Auto-ARIMA + Sentiment + Calendar 6.5%(n=20)Momentum-drift + Sentiment + Calendar 5.7%(n=20)AutoARIMA (StatsForecast) 8.8%(n=11)AutoETS (StatsForecast) 8.4%(n=11)Theta (StatsForecast) 7.5%(n=11)SeasonalNaive (StatsForecast) 12.7%(n=11)Prophet 13.4%(n=5)LinReg (Darts) 5.5%(n=19)Chronos-Bolt 5.5%(n=19)LightGBM (MLForecast) 13%(n=7)
FRED / BLS PPI (urea proxy) · drift -6%/yr · vol 4%
Naive (drift) 3.4%(n=20)Seasonal naive 13.2%(n=20)Moving avg 5.1%(n=20)Holt-Winters 3.3%(n=20)Holt linear 3.3%(n=20)WLS regression 5.3%(n=20)Auto-ARIMA 3.4%(n=20)Momentum-drift 3.2%(n=20)Naive (drift) + Sentiment 3.2%(n=20)Seasonal naive + Sentiment 12.7%(n=20)Moving avg + Sentiment 4.5%(n=20)Holt-Winters + Sentiment 3.2%(n=20)Holt linear + Sentiment 3.2%(n=20)WLS regression + Sentiment 5.1%(n=20)Auto-ARIMA + Sentiment 3.3%(n=20)Momentum-drift + Sentiment 3%(n=20)Naive (drift) + Sentiment + Calendar 3.2%(n=20)Seasonal naive + Sentiment + Calendar 12.7%(n=20)Moving avg + Sentiment + Calendar 4.5%(n=20)Holt-Winters + Sentiment + Calendar 3.2%(n=20)Holt linear + Sentiment + Calendar 3.2%(n=20)WLS regression + Sentiment + Calendar 5.1%(n=20)Auto-ARIMA + Sentiment + Calendar 3.3%(n=20)Momentum-drift + Sentiment + Calendar 3%(n=20)AutoARIMA (StatsForecast) 2.8%(n=11)AutoETS (StatsForecast) 3.6%(n=11)Theta (StatsForecast) 3.4%(n=11)SeasonalNaive (StatsForecast) 24.1%(n=11)Prophet 3.8%(n=5)LinReg (Darts) 3.4%(n=19)Chronos-Bolt 3.9%(n=19)LightGBM (MLForecast) 2.4%(n=7)
FRED / BLS PPI (fresh vegetables) · drift 2%/yr · vol 3%
Naive (drift) 15.5%(n=20)Seasonal naive 22.8%(n=20)Moving avg 14.3%(n=20)Holt-Winters 15.1%(n=20)Holt linear 15.1%(n=20)WLS regression 16.1%(n=20)Auto-ARIMA 15.3%(n=20)Momentum-drift 15.5%(n=20)Naive (drift) + Sentiment 15.3%(n=20)Seasonal naive + Sentiment 22.7%(n=20)Moving avg + Sentiment 13.9%(n=20)Holt-Winters + Sentiment 14.9%(n=20)Holt linear + Sentiment 14.9%(n=20)WLS regression + Sentiment 15.7%(n=20)Auto-ARIMA + Sentiment 15.1%(n=20)Momentum-drift + Sentiment 15.4%(n=20)Naive (drift) + Sentiment + Calendar 15.5%(n=20)Seasonal naive + Sentiment + Calendar 23.2%(n=20)Moving avg + Sentiment + Calendar 14.1%(n=20)Holt-Winters + Sentiment + Calendar 15.1%(n=20)Holt linear + Sentiment + Calendar 15.1%(n=20)WLS regression + Sentiment + Calendar 15.8%(n=20)Auto-ARIMA + Sentiment + Calendar 15.1%(n=20)Momentum-drift + Sentiment + Calendar 15.7%(n=20)AutoARIMA (StatsForecast) 18.1%(n=11)AutoETS (StatsForecast) 17%(n=11)Theta (StatsForecast) 18.1%(n=11)SeasonalNaive (StatsForecast) 20.2%(n=11)Prophet 21.5%(n=5)LinReg (Darts) 18.7%(n=19)Chronos-Bolt 15.1%(n=19)LightGBM (MLForecast) 24.1%(n=7)
Methodology
Each model predicts the price 1 month ahead using only the history observed up to that point — a walk-forward replay over the last 24 live FRED observations per benchmark, with no look-ahead leakage. Use the horizon toggle to compare 1, 2, and 3 months ahead; accuracy typically degrades at longer horizons as uncertainty compounds. Metrics: MAPE = mean absolute % error. Directional accuracy = share of months the predicted sign matched actual (±1% deadband). Band coverage = share of actuals inside the 95% band. Quarterly series and non-live benchmarks are excluded. n(shown on each model chip below, and as “Wins” sample size in the leaderboard) = the number of walk-forward months a model was actually evaluated over — this differs across the Python library models due to differing minimum-training requirements and, for Prophet, a reduced every-3rd-month cadence (an offline runtime constraint). A model with a lower MAPE over fewer months is not necessarily more accurate than one tested over more months — check n before comparing across libraries.
TS-nativeNaive (drift)TS-nativeSeasonal naiveTS-nativeMoving avgTS-nativeHolt-WintersTS-nativeHolt linearTS-nativeWLS regressionTS-nativeAuto-ARIMATS-nativeMomentum-driftTS-nativeNaive (drift) + SentimentTS-nativeSeasonal naive + SentimentTS-nativeMoving avg + SentimentTS-nativeHolt-Winters + SentimentTS-nativeHolt linear + SentimentTS-nativeWLS regression + SentimentTS-nativeAuto-ARIMA + SentimentTS-nativeMomentum-drift + SentimentTS-nativeNaive (drift) + Sentiment + CalendarTS-nativeSeasonal naive + Sentiment + CalendarTS-nativeMoving avg + Sentiment + CalendarTS-nativeHolt-Winters + Sentiment + CalendarTS-nativeHolt linear + Sentiment + CalendarTS-nativeWLS regression + Sentiment + CalendarTS-nativeAuto-ARIMA + Sentiment + CalendarTS-nativeMomentum-drift + Sentiment + CalendarStatsForecastAutoARIMA (StatsForecast)StatsForecastAutoETS (StatsForecast)StatsForecastTheta (StatsForecast)StatsForecastSeasonalNaive (StatsForecast)ProphetProphetDartsLinReg (Darts)ChronosChronos-BoltMLForecastLightGBM (MLForecast)