How an August NHL Power Ranking Actually Gets Built
NHL power rankings consist of panel votes, weighted ballots, computer composites, and odds boards to create a 1-32 team list with notable discrepancies.
Nobody has played a game since June. Every 1-32 you will read between now and October is built from opinions about transactions, and the method used to collect those opinions changes the answer substantially. Knowing which method you are looking at is most of the work.
The panel poll
ESPN runs the best-known version. A panel of hockey commentators, analysts, reporters and editors each submits a 1-32 ballot, and the published list is the aggregate. In season the ballots are cast on games through a cutoff date. In June, with nothing to score, the same panel voted on predictions for 2026-27.
The strength is coverage. A dozen people who follow different beats will collectively catch things a single writer misses, and the averaging smooths out any one voter’s hobby horse. The weakness is that consensus lags. A panel that voted on June 14 was locked in before the June 26-27 draft and before free agency opened July 1, and no aggregate updates itself.
The weighted ballot
Bleacher Report uses a scored variant. A first-place vote is worth 32 points, a last-place vote is worth one, the totals are added and the teams sorted. Mechanically it is the same aggregation, but the scoring makes the gaps legible. Two teams separated by one spot might be separated by four points or by forty, and the second case means something the ordered list alone does not show.
The computer composite
PowerRankingsGuru builds a different animal entirely, averaging the outputs of algorithmic systems including DRatings, Massey, Sagarin and MoneyPuck into one composite number. No humans vote. The tradeoff is straightforward: a model cannot be talked into a narrative, and it also cannot know that a captain who missed a full season is healthy again, or that a general manager was replaced in June.
That limitation bites hardest in August. Rating systems are built on results, and the most recent results describe rosters that in several cases no longer exist.
The market
Sportsbooks and prediction markets produce an implicit ranking that nobody sits down to write. It reprices continuously, it has money enforcing it, and it responds to news within hours. It also compresses badly at the bottom, where the difference between the 28th and 32nd best team disappears into a rounded longshot price.
The same team, four answers
Florida is the case study of the summer. ESPN’s June panel had the Panthers ninth. theScore, ranking on July 4 after the draft and free agency, had them second. By July 30 the futures board had them as the outright 2027 Stanley Cup favourite at +640, ahead of the defending champions, and Kalshi traders had reached the same conclusion on July 8 at 13 percent.
None of those sources is wrong. They were built at different moments using different inputs, and the June number simply predates Brady Tkachuk arriving and Aleksander Barkov getting healthy. A reader who compares a June panel ranking to a July market price and concludes that one of them is stupid has misread what both are for.
What to check before you trust a list
Look for the date it was compiled rather than the date it was posted, because those are frequently weeks apart in the offseason. Look for whether it is a vote, a model or a price. Look for whether the write-ups mention transactions that have since resolved, which is the fastest way to spot a stale list.
And treat any August ranking as a statement about roster construction, not about hockey. The actual evidence starts in October.
Odds and market references describe the situation at the time of writing. You must meet the legal gambling age in your province or territory. See responsible gambling.