Premier League Predictor: FastAPI & PostgreSQL — Chapter 1, Exercise 2
====================================================

TASK
Explain why this app averages multiple guest predictors into a single
"guest" figure for weeks with more than one guest, rather than tracking
each guest as a permanently separate predictor across the season.

SOLUTION

The "guest" slot is genuinely different from the other three prediction
sources. The user, the BBC's expert, and the BBC's own AI predictions
are all real, continuous predictors — the same identity, week after
week, for the entire 38-gameweek season, which is exactly what the
prediction league table (Chapter 8) needs to rank performance fairly
over time.

The guest slot has no such continuity. Some weeks have one guest, some
weeks have more than one, and there's no guarantee any single guest ever
appears again in a later gameweek. Tracking each individual guest as
their own permanent row in the prediction league table would mean most
"guest predictors" only ever have one or two gameweeks of data on
record — not enough to meaningfully compare against a predictor who has
been tracked continuously across the whole season, and it would also
mean the table's own row count grows unpredictably as more one-off
guests show up.

Averaging every week's guest predictions into a single "guest" figure
keeps that slot behaving like the other three — one consistent row in
the prediction league table, updated every gameweek — while still
genuinely reflecting whatever the actual guest(s) predicted that week,
whether it was one person or several.

WHY THIS WORKS AS AN ANSWER
----------------------------
It identifies the real structural difference between the guest slot and
the other three predictors (no persistent identity across gameweeks),
explains the concrete problem that would arise from tracking guests
individually (a table with unstable, mostly-single-data-point rows), and
connects the averaging decision directly to what the prediction league
table in Chapter 8 actually needs to work correctly.