FROZENFORECAST

Gold 2014

Year	Elite	Platinum	Upper Gold	Gold	Gold Grp 1	Gold Grp 2
2019	-	-	-	-	-	-
2018	Elite2018	Platinum2018	Upper Gold2018	-	-	-
2017	Elite2017	Platinum2017	Upper Gold2017	Gold2017	-	-
2016	Elite2016	Platinum2016	Upper Gold2016	Gold2016	-	-
2015	Elite2015	Platinum2015	Upper Gold2015	Gold2015	-	-
2014	Elite2014	Platinum2014	Upper Gold2014	Gold2014	-	-
2013	Elite2013	Platinum2013	Upper Gold2013	Gold2013	-	-
2012	Elite2012	Platinum2012	Upper Gold2012	-	Gold Grp 12012	Gold Grp 22012

The Information Ladder

What our model knows — and what each piece is worth. This is also the answer to "why not just use win-loss records?"

No-skill baseline

+50.5%

A coin flip that ignores the teams entirely.

Team identity (results only)

+6.7%

Knowing who is playing — win/loss records, no scores — closes most of the gap.

Elo with margins

+2.1%

Adding goal margins to the rating recovers most of the rest.

Full Dixon-Coles

reference

Full goal modeling is the reference — the last ~2% over Elo.

Knowing who is playing closes most of the gap from a coin flip. Goal margins recover most of the rest. Full goal modeling — the Dixon-Coles model this site runs — is the last couple of percent. Each rung is measured against the one below it (relative accuracy vs Dixon-Coles).