Accuracy scorecard
Every forecast is stored the moment it is made and scored against the settled outturn once Elexon publishes it. This is the number that decides whether a paid market-data feed is worth buying: if the error here is dominated by where the market was trading, a live EPEX or Montel feed will help. If it is dominated by balancing actions, it will not.
| Horizon | Scored | Mean abs error | Bias | Sign match |
|---|---|---|---|---|
| Current period | 33 | £36.70 | -19.13 | 100% |
| +1 period | 32 | £40.56 | -26.26 | 100% |
| +2 periods | 31 | £44.49 | -30.89 | 100% |
| +3 periods | 30 | £46.55 | -29.84 | 100% |
| +4 periods | 29 | £45.00 | -31.93 | 100% |
Baseline (market index price used as the forecast): £41.05 mean absolute error. The model is only useful if it beats this.
Current-period forecast vs settled outturn (7 days)
Model fits
| Fitted | R² | σ | Samples |
|---|---|---|---|
| 02 Nov, 03:15 | 0.309 | £35.9 | 1399 |
| 02 Nov, 03:15 | 0.423 | £42.4 | 1397 |
| 01 Nov, 03:15 | 0.306 | £36.2 | 1399 |
| 01 Nov, 03:15 | 0.421 | £42.5 | 1397 |
| 31 Oct, 03:15 | 0.304 | £36.1 | 1399 |
| 31 Oct, 03:15 | 0.419 | £42.6 | 1397 |
| 30 Oct, 03:15 | 0.307 | £36.1 | 1399 |
| 30 Oct, 03:15 | 0.425 | £42.4 | 1397 |
| 29 Oct, 03:15 | 0.312 | £36.5 | 1399 |
| 29 Oct, 03:15 | 0.431 | £42.6 | 1397 |
Collector runs
- 03 Nov, 00:50settled=49 forecasts=5
- 03 Nov, 00:45settled=48 forecasts=5
- 03 Nov, 00:40settled=48 forecasts=5
- 03 Nov, 00:35settled=48 forecasts=5
- 03 Nov, 00:30settled=48 forecasts=5
- 03 Nov, 00:25settled=48 forecasts=5
- 03 Nov, 00:20settled=48 forecasts=5
- 03 Nov, 00:15settled=47 forecasts=5
- 03 Nov, 00:10settled=47 forecasts=5
- 03 Nov, 00:05settled=47 forecasts=5
- 03 Nov, 00:00settled=47 forecasts=5
- 02 Nov, 23:55settled=95 forecasts=5