Starting goalies

How to read it

8 clubs on the slate. 8 nets are closer than 60/40, so treat them as open.

None of this is an announcement. Clubs name a starter around the morning skate, and sometimes not until warmups. Every chance here is a projection, and the column beside it says where it came from. When the books hang a saves line on exactly one goalie, that is the strongest signal there is before a club speaks: the lined goalie started 224 of 226 games this season. We still print it as 93%, not 100%, because it isn't a confirmation.

Before the books post, the number is ours. The model reads what each club has actually done: who started last night, whether tonight is a back-to-back, each goalie's share of the last ten and thirty starts, how many in a row he has made, his rest, his last five save percentages and his workload over the past year. On the 2025-26 season, which it never saw while it was fitted, its pick started 66% of the time across 2,623 games, against 57.1% for simply taking each club's most-used goalie. On back-to-backs it was right 79.3% of the time. The goalie who started the first night almost never starts the second. The model never says more than 90%, because its most confident calls were right about nine times in ten, not every time.

What it can't see: an injury nobody has reported, a goalie recalled from the minors that morning, a coach who simply changes his mind. ESPN lists probable goalies too, and we tested that list against who actually started. When it named one goalie, he started 53% of the time, so we don't use it.

The same chances drive the saves board, each club's lines page and the lineup optimizer, so they always agree. They reprice every hour.