Giorgos Diakos

Observed in more than one division. D2 2022-2023 at 26 involvements in 26 nineties (1.00/90), then D1 2024-2025 at 0.89/90. Surname is unique in the population, so the match is high confidence. These two rates are not directly comparable, which is exactly what the level model is being built to resolve.

Career

Season	Div	Team	Pos	GP	90s	G	A	raw/90	adj/90	sched	rank	pct
2025	D1	Cornell	FWD	15	8.6	3	5	0.739	0.705	0.94	91	97
2024	D1	Cornell	FWD	18	6.0	2	3	0.656	0.654	1.00	157	94
2023	D2	Charleston (WV)	MID	21	12.4	10	5	0.874	0.789	0.89	95	96
2022	D2	Charleston (WV)	MID	22	13.7	6	5	0.632	0.543	0.84	316	87
Career	D1 + D2	2 teams		76	40.7	21	18	0.597	0.550

The career row totals only what can be totalled. Schedule, rank and percentile are blank because each one is a position inside a single division-season, so there is no career version of them, and the best season's percentile is not one. The career raw/90 and adj/90 are every season put on the D1 scale and weighted by minutes, which is the only way to total a rate across divisions. The level ladder is measured from 665 players observed in two divisions, in both directions, so it is a within-player comparison rather than a guess. A goal in D2 counts as 0.72 of a D1 goal. Treat it as a translation, not a projection. The ladder is an average over players who moved, and most of them were mid-table, so applying it to someone far out in his division's tail assumes the gap is the same size up there. The mover sample cannot check that.

2025 · every game by opponent strength

Each circle is one game, placed by how strong the opponent was and how many goal involvements he had. Filled circles are games he scored or assisted in. He drew 100% of his involvements against the weaker half where 69% was expected given the schedule he played (p = 0.053). That is within the normal range.

2025 game log