Blog

Grading guide

How to tell if fantasy basketball rankings are any good

Five rules for grading any 9-category board against what actually happened, and how Ball Ranks applies them to its own rankings in public.

November 3, 2026· Ball Ranks

Every fantasy basketball site publishes a preseason board. Almost none of them tell you how last year's board turned out.

That leaves you choosing rankings on brand names and confident paragraphs. You can do better with a spreadsheet and an evening. Here is the method we use to grade our own board against ESPN, Yahoo, FantasyPros panelists, and other sites, written so you can run it on any source you're thinking of trusting.

Rule 1: grade the board people actually drafted from

A ranking only counts if it existed before the season started. Boards get edited in November after a hot start, and a December snapshot graded against April results is not a draft-day prediction.

Use the version that was live before opening night. For past seasons, archived copies on the Wayback Machine are the cleanest source. If you can only find an in-season version of a board, leave that season out instead of guessing.

Match the format, too. A points-league board graded on 9-category results will look worse than it is, and so will a salary-cap board graded on draft rank. Grade category boards against category results.

Rule 2: compare inside a shared pool

ESPN might rank 150 players. Another site might rank 300. Comparing a site's rank 180 with nothing is meaningless, so build one pool per source and season:

  1. Start with the players on the outside board. If you're comparing two outside boards, keep only players both of them ranked.
  2. Drop players who never reached a minimum number of games. We use 20. A player who played eight games tells you about injury luck, not about the ranking.
  3. Re-rank everyone left, on both boards and on the actual season results, from 1 to the size of the pool.

Now every player has three numbers in the same units: where each board put him and where he actually finished among the same players. Use per-game category value for the actual finish if your league plays per game, season totals if it plays totals.

Rule 3: weight the misses that cost you leagues

The miss for each player is the gap between his ranked position and his actual finish inside the pool. A plain average treats missing the 10th pick the same as missing the 190th, and nobody loses a league over the 190th pick.

We weight the best actual finisher in each pool at 2x, declining smoothly to 1x for the last player. There's no cliff between rank 20 and rank 21, but a whiff on a real difference-maker still counts double. Lower weighted miss is better.

Rule 4: don't let hindsight in

The easiest way to flatter a board is to reconstruct it after the fact. Rookies are where this goes wrong: rebuilding what a model "would have said" about a first-year player is hard to do without knowing how his season went.

We leave first-year players out of these comparisons on both sides, so every player graded has at least one prior NBA season. If you're grading two outside sites against each other, you don't need this rule, since both boards were published in advance. You need it the moment one of the boards is rebuilt rather than archived.

Rule 5: score each source-year on its own and publish the losses

Report a result for every source and season, not one blended number. A board that wins four seasons by a little and loses one by a lot should show you that loss.

Then publish all of it. A record that only includes wins is marketing.

What this looks like on Ball Ranks

Ball Ranks runs every projection change against completed NBA seasons before it ships, and grades its board against outside boards with the same shared pools, eligibility rules, and weighted miss, with every result published. Where a source's format couldn't be confirmed, the comparison page says so.

Run it on this season

The 2026-27 season is under way, so save the preseason boards you drafted from now, before they change. Keep a copy of each one and grade them in April with the five rules above.

The 2026-27 Ball Ranks board covers every NBA player in 9-category leagues, with per-game and season-total modes and re-ranking for up to three punted categories. The comparisons above show how it has held up against archived boards, with the losses published next to the wins.