How These Signals Work
Each outlook card combines five independent inputs. We deliberately lead with the four technical signals — not the dollar forecast — because multi-day point forecasts on commodity prices are noisy by their nature, while the underlying signals are far more interpretable.
1. Trend regime (50DMA vs 200DMA)
The relationship between the 50-day and 200-day simple moving averages is the textbook way to label a market's long-term regime. A 50DMA above the 200DMA is a golden-cross regime (bullish bias); a 50DMA well below the 200DMA is a death-cross regime (bearish bias). Range-bound markets show the two averages crossing back and forth.
2. Momentum (RSI 14)
The 14-day Relative Strength Index gives a fast read on whether buyers or sellers are in control. Readings above 70 are overbought (often a contrarian sell signal), below 30 are oversold (often a contrarian buy signal), and the 45–55 range is genuinely neutral.
3. MACD cross (12/26/9)
The MACD histogram (MACD line minus signal line) is the most-watched short-term cross signal in technical analysis. We chip it as above signal, below signal, or flat — a positive histogram is consistent with bullish momentum, a negative histogram with bearish momentum.
4. Statistical model (damped Holt)
The dollar forecast comes from a damped-trend Holt exponential smoother (Gardner & McKenzie 1985), with the smoothing parameters (α, β, φ) chosen by grid search. The damped trend is well-known to outperform classic Holt on multi-step horizons because it prevents the trend from extrapolating indefinitely. We back-adjust the underlying futures series for contract rolls before fitting so the model is not penalised for unforecastable contract-roll discontinuities.
5. Out-of-sample backtest
The "backtest credibility" block under each card is the most important number on the page. We replay the same model across the most recent 30 trading sessions, scoring 7-day-ahead forecasts against the actual prices that printed, and report the mean absolute percentage error (MAPE). We compare it to a naive "no change" baseline — if the model can't beat that baseline, the card honestly says so.
Why we surface this instead of a single point forecast
Predicting energy prices a week out is genuinely hard. Anyone publishing a confident-sounding price target is either selling something or hasn't backtested. We'd rather give you a transparent dashboard of independent signals plus an honest accuracy number, and let you decide how much weight the dollar forecast deserves. The signal stack is also independently useful: when trend, RSI, MACD and the 50/200 DMA regime all agree, the directional setup is materially stronger than when they disagree, regardless of any model output.