Readout 14 Nov 2026 · interim
23 of 64 forecast dates matured · 551 company-days
All companies
551 company-days · 24 companies · 23 forecast dates (14 Jul 2026 – 14 Aug 2026)
| Series | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|
| Model panel median | −4.60 | 31.52 | 43.0 |
| Analyst mean target | +8.30 | 14.47 | 53.2 |
| Pre-set blend (model weight 0.20) | +5.72 | 14.26 | 54.4 |
H1 · bias · |analyst mean error| − |model mean error| = +3.70 · 90 % interval [−11.18, +10.56] → no difference established
H2 · pre-set blend · analyst MAE − blend MAE = +0.21 · 90 % interval [−2.21, +2.66] → no difference established
Same metrics for the model series re-computed with the all-class share count and the dividend subtracted.
| Series | n | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|---|
| Model panel median | 545 | −12.55 | 28.44 | 48.1 |
| Analyst mean target | 545 | +8.39 | 14.57 | 53.0 |
| Pre-set blend (model weight 0.20) | 545 | +4.20 | 13.61 | 53.6 |
H1 · bias · |analyst mean error| − |model mean error| = −4.16 · 90 % interval [−17.43, +7.67] → no difference established
H2 · pre-set blend · analyst MAE − blend MAE = +0.96 · 90 % interval [−1.60, +3.43] → no difference established
| Series | n | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|---|
| Without companies pinned at an engine limit | 482 | −6.97 | 27.55 | 40.9 |
| Unrolled estimate (present value, no 12-month roll-forward) | 551 | −3.46 | 41.27 | 45.2 |
| 70/30 research blend | 551 | −0.73 | 23.04 | 44.5 |
| Dates when the median company rose — Model panel median | 551 | −4.60 | 31.52 | 43.0 |
| Dates when the median company rose — Analyst mean target | 551 | +8.30 | 14.47 | 53.2 |
Best in-sample model weight 0.10 (90 % interval 0.00–0.30) — fitted on this same data, so it cannot be worse than either series alone.
Finland
275 company-days · 12 companies · 23 forecast dates (14 Jul 2026 – 14 Aug 2026)
| Series | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|
| Model panel median | +10.72 | 32.88 | 45.5 |
| Analyst mean target | +5.48 | 14.29 | 35.3 |
| Pre-set blend (model weight 0.10) | +6.01 | 14.27 | 41.8 |
H1 · bias · |analyst mean error| − |model mean error| = −5.23 · 90 % interval [−25.07, +6.90] → no difference established
H2 · pre-set blend · analyst MAE − blend MAE = +0.02 · 90 % interval [−1.98, +1.81] → no difference established
Same metrics for the model series re-computed with the all-class share count and the dividend subtracted.
| Series | n | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|---|
| Model panel median | 275 | +2.43 | 27.89 | 48.7 |
| Analyst mean target | 275 | +5.48 | 14.29 | 35.3 |
| Pre-set blend (model weight 0.10) | 275 | +5.18 | 14.22 | 38.9 |
H1 · bias · |analyst mean error| − |model mean error| = +3.06 · 90 % interval [−17.92, +8.41] → no difference established
H2 · pre-set blend · analyst MAE − blend MAE = +0.07 · 90 % interval [−1.83, +1.64] → no difference established
| Series | n | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|---|
| Without companies pinned at an engine limit | 252 | +1.95 | 26.13 | 44.0 |
| Unrolled estimate (present value, no 12-month roll-forward) | 275 | +17.21 | 43.79 | 52.0 |
| 70/30 research blend | 275 | +9.15 | 24.85 | 44.0 |
| Dates when the median company rose — Model panel median | 132 | +8.16 | 32.42 | 52.3 |
| Dates when the median company rose — Analyst mean target | 132 | +4.89 | 13.29 | 44.7 |
| Dates when the median company fell — Model panel median | 143 | +13.08 | 33.31 | 39.2 |
| Dates when the median company fell — Analyst mean target | 143 | +6.03 | 15.21 | 26.6 |
Best in-sample model weight 0.05 (90 % interval 0.00–0.30) — fitted on this same data, so it cannot be worse than either series alone.
United States
276 company-days · 12 companies · 23 forecast dates (14 Jul 2026 – 14 Aug 2026)
| Series | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|
| Model panel median | −19.85 | 30.16 | 40.6 |
| Analyst mean target | +11.10 | 14.65 | 71.0 |
| Pre-set blend (model weight 0.35) | +0.27 | 14.76 | 55.8 |
H1 · bias · |analyst mean error| − |model mean error| = −8.75 · 90 % interval [−26.61, +9.58] → no difference established
H2 · pre-set blend · analyst MAE − blend MAE = −0.11 · 90 % interval [−5.59, +5.81] → no difference established
Same metrics for the model series re-computed with the all-class share count and the dividend subtracted.
| Series | n | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|---|
| Model panel median | 270 | −27.80 | 29.01 | 47.4 |
| Analyst mean target | 270 | +11.35 | 14.85 | 71.1 |
| Pre-set blend (model weight 0.35) | 270 | −2.35 | 12.51 | 58.9 |
H1 · bias · |analyst mean error| − |model mean error| = −16.45 · 90 % interval [−30.97, −2.47] → contradicted (the difference runs the other way)
H2 · pre-set blend · analyst MAE − blend MAE = +2.34 · 90 % interval [−3.98, +8.98] → no difference established
| Series | n | Mean error, pp | Mean absolute error, pp | Direction match, % |
|---|---|---|---|---|
| Without companies pinned at an engine limit | 230 | −16.73 | 29.10 | 37.4 |
| Unrolled estimate (present value, no 12-month roll-forward) | 276 | −24.06 | 38.76 | 38.4 |
| 70/30 research blend | 276 | −10.57 | 21.25 | 44.9 |
| Dates when the median company rose — Model panel median | 276 | −19.85 | 30.16 | 40.6 |
| Dates when the median company rose — Analyst mean target | 276 | +11.10 | 14.65 | 71.0 |
Best in-sample model weight 0.20 (90 % interval 0.00–0.45) — fitted on this same data, so it cannot be worse than either series alone.