The sleepers nobody ranked high enough

Every preseason ranking is a map of the season we think we are about to get. Then the games begin, rotations mutate, young players leap, and a few names launch themselves through the roof of that map.

We went back to the archived ESPN and Yahoo 9-category boards for 2025–26, placed each against the season’s final 9-category results, and asked one simple question: who climbed the farthest?

The honest version of a sleeper list is written after the season—with the original rankings still attached.

The three boards, side by side

These are rank gains, not points scored. A gain of 110 means the player finished 110 places better than his preseason position. ESPN and Yahoo are each measured inside the group of players available on that archived board. Ball Ranks uses its full backtest pool, so the exact rank numbers are not interchangeable across columns.

ESPN’s blind spot: Keyonte George

ESPN had George at 135th inside its surviving comparison pool. He finished 25th, a 110-place jump and the largest climb on that board.

The useful lesson is not merely that a young guard improved. It is that uncertain efficiency can bury a player whose path to volume is already visible. George did not need to become flawless to smash a rank in the 130s. He needed enough growth for the minutes, assists, scoring and threes to outweigh the parts of his profile that made preseason models nervous.

Onyeka Okongwu and Andrew Wiggins completed ESPN’s top three. Different players, same warning: established labels—timeshare big, known veteran wing—can lag behind a changing role.

Yahoo’s blind spot: Neemias Queta

Yahoo’s largest riser was Queta, from 177th to 55th inside that comparison pool. A player treated as a last-round flier produced like someone who belonged much closer to the middle of a standard draft.

Queta is the classic category-league trapdoor. Low-usage centers can look unimportant in a broad basketball conversation while rebounds, blocks and strong field-goal percentage quietly do serious fantasy work. When the playing time arrives, the category value can arrive all at once.

Kevin Porter Jr. climbed 85 places, while George appeared again with an 84-place rise. When a player shows up as a major miss on two independently assembled boards, that is less a quirky ranking decision and more a league-wide failure to price the upside.

Our blind spot: the second-year jump

Now for the part where we put our own forecast under the same fluorescent light.

Ball Ranks’ biggest miss was Reed Sheppard: 311th in the preseason backtest, 70th at season’s end. Jaylon Tyson rose 227 places and Cam Spencer rose 224. All three were second-year players.

That pattern matters more than any one name. A history-driven model naturally has the least evidence on players with one NBA season—especially when that first season offered limited minutes. Lean too hard on the small sample and the breakout never appears. Lean too hard on draft reputation and every young player becomes a fake sleeper.

The answer is not to add 200 places to every sophomore. It is to get sharper about which early-career signals predict a real role and which are just summer optimism in a nice suit.

What we would carry into the next draft

Three practical rules survived the autopsy:

1. Treat a clear path to more minutes as a ranking input, not a footnote. The biggest leaps were rarely mysterious after the role changed.
2. Look for category concentration. A player does not need a complete stat line to become valuable if he can swing scarce categories.
3. Be humbler with short résumés. Second-year players create an uncomfortable forecasting problem precisely because the evidence is thin and the range of outcomes is wide.

This is why we keep the old boards. Anyone can name last season’s sleepers in July. The useful work is preserving what we believed in September, measuring exactly how wrong it was, and making the next board a little harder to surprise.

Sources and method

We compared the archived ESPN 2025–26 roto/category rankings and Yahoo Sports 2025–26 9-category rankings with final 2025–26 9-category per-game value. Players had to reach 20 games, and first-year players were excluded because a fair historical Ball Ranks comparison cannot recreate manual rookie scouting without hindsight. Each external board was re-ranked inside its own shared player pool; the Ball Ranks section uses the complete 393-player non-rookie backtest pool.