<div class="section-label"><span class="num">05</span> Solo Distribution</div>
<div class="section-title">The distribution of <em>information</em>.</div>
A **solo prediction** is a prediction formed from a single observation, scaled onto the same expected scale as this task. The actual forecast is a weighted average of those solo predictions. Together they form a **distribution of information** for this prediction, not a classical error-variance band.
If the information is mutually supportive, the distribution is tight and you should be more confident. If it is mutually contradictory, the distribution is wide and you should be less confident. Unlike a residual variance that is the same for every task, this distribution is specific to *these* circumstances.
Censored observations receive zero weight, so they do not appear in the distribution. Observations with very low relevance can produce extreme solo values; their weights are tiny, and the optional histogram typically folds those tails so the display is not dominated by noise.
```text
result.solo_distribution
├── y_solo, xi_solo
├── statistics sigma, skewness, kurtosis, modality…
└── distribution optional histogram
```
## Vectors
| Field | Shape | Role |
| --- | --- | --- |
| <span class="mono">y_solo</span> | N | Solo prediction from each observation |
| <span class="mono">xi_solo</span> | 2-D | Weight of that solo prediction in the actual forecast |
The weights are the observation’s share of squared relevance among retained observations. They are nonnegative and sum to 1.
Example: <span class="mono">result.solo_distribution.y_solo</span>
## Summary statistics
When filled (length typically T), these describe the shape of the information distribution:
| Field | Role |
| --- | --- |
| <span class="mono">sigma</span> | Spread of the solo predictions |
| <span class="mono">skewness</span> | Skewness |
| <span class="mono">kurtosis</span> | Kurtosis |
| <span class="mono">pearson_modality_index</span> | Pearson modality index |
| <span class="mono">bimodal_index</span> | Bimodal index |
## Optional histogram
A view of the information distribution under <span class="mono">solo_distribution.distribution</span> (<span class="mono">bin_edges</span>, <span class="mono">bin_centers</span>, <span class="mono">bin_widths</span>, <span class="mono">count</span>). Extreme tails are typically IQR-filtered so low-relevance outliers do not dominate the plot.
On Grid, the histogram can pool solo predictions across cells: each cell is a weighted average of solos, and the composite is a fit-weighted average of cells.
| Path | When it appears |
| --- | --- |
| Predict | When the run includes it |
| Grid | When you retain <span class="mono">ysolo_distribution</span> (pooled across cells) |
## Related
- [[Results/Grid Insights|Grid Insights]]: composite solo weights on Grid (<span class="mono">xi_solo_composite</span>)
- [[Results/Grid Cells|Grid Cells]]: per-combination solo tables
- [[Settings/Options#Retain grid objects|Retain grid objects]]: how to keep the Grid histogram
<div class="btn-row center">
<a class="btn-primary" href="/Results/Grid%20Insights">Grid Insights</a>
<a class="btn-ghost" href="/Results/Overview">Results</a>
</div>