The Radar Scored an Injured Player 96. He Didn't Move.

TL;DR: the September 11 roster update, graded by score band. 1,424 cards scored, 294 upgraded, so chance was about 1 in 5. Cards the Radar scored 80–89 upgraded 79% of the time, 3.8× chance. 63 of the 77 rarity crossings scored 50 or higher. And a 96 on Nick Kurtz, who hadn't played since July, went 86→86.

---

The most useful thing Diamond Radar did this update was get Nick Kurtz wrong in a way I watched happen. The rest of the update went well, and I'll show all of it band by band. But Kurtz is the one that taught me something, so he goes first.

🩼 Nick Kurtz: a 96 on the 60-day IL

Kurtz first went on the injured list this season on July 10. He came back on July 25, played his last game on July 31, went back on the IL retroactive to August 1, and on August 21 moved to the 60-day IL, done for the season. I started watching him on the Radar that first time he went down in July, because I wanted to see what the model would do with a player who'd stopped playing.

For a while it did the sensible thing. At the August 14 roster update, two weeks into his second IL stint, he scored a 39.

Then, on August 31, he went from 40 to 96 overnight. That was three days after we shipped an upgrade to the scoring model, and exactly a month after his last game. By September 3 he was still a 96, flagged Hot, while sitting on the 60-day IL. The card even said so with a straight face, right next to a "Very low confidence" badge.

Here's why, and it's the most useful thing I can tell you about how the Radar reads a player. It looks at his real production across several stretches, recent and full-season, and measures the gap between that and what his card says he is. When a player stops playing, the recent stretches go empty. The model doesn't read empty as "bad." It sets them aside and leans on what's left. By the end of August his last game was a month old, and every recent stretch was empty. For Kurtz, what was left was his season line: 99 games of .256/.396/.483. With nothing else to go on, that one line went from a small slice of his score to all of it. He didn't fade. He got stickier. And there's no injury data in the Radar at all, so nothing told it why the stretches were empty.

He went into the September 11 update at 96 and came out of it 86→86. No upgrade, no downgrade. The day after, his score dropped to 84, and it stayed there through September 16.

His Radar score every day. Each update is graded on the last score from before update day, so the August 14 update used his August 13 score of 39. The line breaks on August 2, the one day with no score recorded.

Nobody picks a Radar score by hand, including me. Every night it re-learns how much each attribute matters from the roster updates I've approved, and it pings me when a new update is waiting to be graded. What it can't do yet is notice that one of its own scores makes no sense. Seeing him at 96 on September 3 is what sent me digging into why, and how to fix it. I filed the fix that day. Eight days later Kurtz was still a 96 on update day, because the fix isn't built yet. That's the gap between automated and unattended, and it's why I watch it all season.

The fix isn't to push injured players down. My hunch is that SDS leaves an injured player's card alone, and if that's right, "sell Kurtz" would be the same mistake pointed the other way. So the fix is two things. A player's season line shouldn't swallow his whole score just because his recent stretches are empty. And a player who can't play shouldn't get a score at all. Before I build that second one, I'm checking SDS's own history to see whether injured cards really do sit still. Right now it's a hunch from one card, and one card is an anecdote.

A score is a chance. Here's what every band did.

A Radar score runs 0 to 100, which makes it look like a percentage. It isn't. It's a ranking, and the only way to know what a 70 is worth is to check. So here's every card we scored on September 11, grouped by score:

Three things to read off that.

The bottom is as useful as the top. 213 cards scored under 10. One upgraded. When the Radar shrugs at a card, it means it.

It climbs all the way to 80–89. Every band from 60 up was at least 2× chance. Every band under 40 was under 1×. The only step down is at the very top, where 90–100 came in just under 80–89.

The top is three in four, not ten in ten. The 90–100 band went 28 for 37, a small sample, and 80–89 did a touch better at 45 for 57. Nine of those 37 top-band cards didn't upgrade: Kurtz at 96, Cal Raleigh at a flat 100, and seven more, one of which, Connor Norby at 97, actually dropped a point. That's not the model breaking. That's what a top score means: most move, some don't.

Every card the Radar scored before the September 11 update, grouped by score. The line is what you'd get picking at random: 21% of all cards upgraded.

📈 Rarity crossings: 63 of 77 scored 50+

Crossing a rarity line is the upgrade I watch most. 77 cards did it on September 11, and 63 of them had a Radar score of 50 or higher going in. 29 scored 80 or higher.

The 90–100 band holds only 37 cards, and 15 of them crossed a line. Four cards made Diamond, and three had scored 50+: Jesús Luzardo (scored 87, 84→87, and the Radar projected 87 exactly), Elly De La Cruz (75, 83→85) and Jackson Chourio (66, 84→85). The fourth, Michael Harris II (84→85), scored under 50. Silver→Gold was the sharpest group: Rafael Devers (92), Alex Bregman (92), Oneil Cruz (92) and Randy Arozarena (99) were among six top-band cards that made Gold.

The +8s

Four cards went up eight points, the biggest jumps in the update: Nate Lavender (57→65), Héctor Rodríguez (60→68), Pedro Ramirez (63→71) and Julian Garcia (57→65). All four were Common→Bronze, and all four scored 83 or higher: 90, 89, 88 and 83.

The Radar projected every one of them at +3. It's a lot better at "will he" than "by how much," and a +8 is where that shows most.

What the prices did

These are daily closing buy prices from our market history: the average over the week before the update (September 4–10), the close the day before, and the best close in the days after, through September 15. The best close is hindsight, because nobody knows the top while it's happening. And this way of measuring flatters some cards a lot more than others: cards whose ratings didn't change at all typically hit a best close about 2× their week-before average if they cost 30 to 100 stubs, but only about 1.0× at 100 stubs or more. Read these as what happened, not as what a trade would have made.

Measured from the week before, the two biggest moves in this table, relative to price, were Silver→Gold cards. Cavalli went from a 127 average to 646, and Stewart nearly tripled. Not every Silver→Gold card paid: Jac Caglianone made Gold and still never got back to his pre-update average (639 the week before, 451 at best through September 15).

The Diamonds tell a different story, and it's the one to remember. Luzardo averaged 3,433 the week before the update and was already 4,596 the day before, so most of his move happened before the new ratings dropped. De La Cruz and Chourio did the same thing. If you only buy the night before, you're paying for an upgrade the market has already priced in.

Each line starts at the card's average price the week before the update. "3×" means three times that price. The last point is the highest price in the five days after, which you only know looking back.

What a miss actually cost

A top-band miss isn't automatically a disaster, but it isn't free either. Kurtz barely moved: 4,242 the day before the update, 4,185 two days after, down about 1%. Raleigh fell harder. He spent most of early September closing between about 800 and 970, closed at 1,100 the day before the update, and two days later he was down to 764, below where he'd spent most of the month.

I've been pulling these numbers for every band from 50 up, across all five updates we've graded this season, and it's getting its own post, including how much the price you pay matters next to the score.

Diamonds are the hard end

Only 5 of the 52 Diamonds we scored upgraded, about 1 in 10, half the window's chance. That's also where a miss hurts. A wrong 96 on a 44-stub Common is a rounding error. A wrong 96 on Kurtz put 4,242 stubs at risk. If you're spending real stubs on a Radar score, weigh the tier as much as the number.

The whole window, in one line

1,424 cards scored, 294 upgraded, chance about 1 in 5, and every band from 60 up at least double that. Of the five updates we've graded this season, only the first one, back in May, ranked cards slightly better.

This might be the last major roster update of 2026. SDS hasn't said, and I'd call it a coin flip. Either way, the injured-list fix is on the offseason list, and there are now five graded updates to test it against before next year's game.

Every graded update, band by band, checked against SDS's roster updates: /radar/track-record.

---

Diamond Radar is a forecast model: forecasts, not guarantees. Markets move. Stub sales, flash sales and demand swings can shift prices at any time. Nothing here is financial advice.