Radar's Top 10 Went 10 for 10. The Players Who Weren't Playing Taught Us More.

TL;DR. On the October 2 roster update, the top of the Diamond Radar board was nearly perfect: the top 10 went 10 for 10 and the top 20 went 19 of 20. Cards scored 90 or better upgraded 95% of the time, against 23% for any card. Every score band lined up in order, high to low. The softer spot, against September, was the 80s. And the players who weren't playing, including one we kept on the board on purpose, showed us a clear blind spot: the model doesn't let their scores fall far enough.

---

That's the best the top of the board has looked since the first graded update in May.

San Diego Studio published the October 2 roster update, which it called its "Final Regular Season Roster Update." 745 cards had ratings changed, and 512 of them moved overall. Here's how the board from the day before did against it.

The top of the board

The accuracy page counts the top of the board on its own, because that's how the board actually gets read. So start there.

• Top 5: 5 of 5 upgraded.
• Top 10: 10 of 10.
• Top 20: 19 of 20.

The one miss in the top 20 was Kyle Schwarber, with a Diamond Radar score of 95. He stayed at 85.

On September 11 the top 20 went 15 of 20. Across the season it had ranged from 70% to 100%. This one is a single card short of May's 20 for 20.

Every band, in order

Each band higher than the one below it, all the way up. That hadn't happened yet this season.

1,454 cards were graded and 328 went up. So any card had about a 23% chance, whatever Radar thought of it. That's the line to beat.

Two things I'd point at.

The staircase climbs in every step. On September 11 the 90s came in just under the 80s, 76% to 79%. This time the top band is clear of everything.

And the lower bands were right too. Every band under 50 sat at or below the line. A low score said "probably not," and it was.

Same model, second time

This was the second update graded on v11, the version of the model running now. The first was September 11.

The very top got sharper. The 70s and 80s got softer. Two updates is still two updates.

The very top got sharper: 76% to 95% in the 90s, and 15 to 19 in the top 20. The 70s and 80s got softer. The 80s went from 79% to 67%.

Put together, everything v11 scored 70 or better across its two updates: 342 cards, 71% upgraded, against about 22% for any card. That's two updates in a row at three times chance or better. It's a good start. It's still two updates.

The hits

Victor Mesa Jr. had a Diamond Radar score of 100 and went 67 to 76, Bronze to Silver. That tied for the biggest jump in the whole update.

The other was Leo Bernal (92), 58 to 67, Common to Bronze. Ray Kerr (100) went 59 to 65, and Jared Jones (99) went 78 to 80, Silver to Gold. Three more cards scored 90 or better crossed Silver to Gold: Nolan Arenado, Brandon Lowe and Brice Turang.

The October 2 page of calls that landed. "Radar called" is the overall rating Radar predicted, not the Radar score. Mesa Jr. was predicted to reach 70 and got 76.

71 cards crossed a rarity line upward. 56 of them had a score of 50 or better the day before. Two crossed into Diamond: Mookie Betts (84) and Cade Smith (62).

The misses

Only two cards scored 90 or better didn't upgrade: Schwarber at 95 and Brett Baty at 91, who stayed at 68. None of the 90s went down.

The 80s are where it slipped against September: 79% then, 67% now. 24 of the 72 didn't go up, including Mike Trout (87) and Zach Neto (88). Two went down a point.

And one name from last time. Cal Raleigh had a score of 100 on September 11 and didn't move. This time he scored 77 and went down, 81 to 79, Gold to Silver.

The player we left on the board on purpose

Nick Kurtz has been on the 60-day injured list since August 21. He hasn't played since July 31. We could have stepped in, and we didn't, on purpose. I wanted to see what the model does with a player who stops playing, because that's a case it has to handle on its own. The last report card is where I said I was watching him.

Two other players ended up in the same spot. Kendry Rojas hasn't pitched in the majors since August 2, and Chase Petty since August 7. Neither was on the major-league injured list when he stopped pitching. Both were sent down to the minors, and Rojas spent three weeks on the injured list there. Here's what the board showed for all three.

Diamond Radar scores, as the board showed them each day.

Under the older model, their scores fell off the way you'd want, by about half or more across August. Then on August 28 we shipped a new version that reads recent form from the last big roster update. When a player's recent stretch goes empty, it sets that stretch aside and leans on what's left. All three jumped back up within ten days. Since mid-September they've barely moved. Kurtz dropped to 84 the day after the September 11 update and sat there every day through October 2.

So the answer to "how fast does a player who isn't playing fall off the board?" is: right now, not nearly fast enough.

The updates show what that costs. Rated 70 or better, these players missed five times across September and October. Kurtz twice, Rojas once and Petty twice. The one hit was Rojas this time, who went up 61 to 66 anyway, and I'm not counting that as the model being right. Across everything scored 50 or better, players who hadn't played in the majors for a month went up once in 51 tries. Every other card scored 50 or better went up more than half the time.

That's what keeping Kurtz there was meant to show, and it gives us three things to tune:

• How fast a score falls once a player stops playing.
• A short injured-list stint isn't a lost season. A player back in two weeks can still earn an upgrade. A player out for the year shouldn't hold a high score at all.
• Injured isn't the only way to stop playing. Rojas and Petty weren't on the major-league injured list when they stopped pitching in the majors. They were sent down. "Hasn't played in the majors lately" is the thing to watch.

This is the part of Radar that's hard to do by hand. It scores every Live Series card it has this season's stats for, on every scoring day, and then it gets graded on every one of them, so a blind spot like this shows up in the numbers instead of hiding in a highlight reel. Until the model learns this one, a high score on a player who isn't playing is the first thing I'd check myself.

Diamonds are still the hard end

Radar grades every card on the board, so 50 Diamonds were graded. It also flagged 38 of them as upgrade candidates, including 15 that scored under 50. Only 9 of those 38 went up. Across all 424 cards it flagged, more than half did. Only 10 Diamonds scored 70 or better, and 4 of those went up, the worst of any rarity.

The fair test is band by band. A Diamond should go up about as often as any other card with the same score. If they had, about 17 of the 50 would have upgraded. 10 did. On September 11 it was 5 of 52 against about 14, and Radar had flagged 37 of them.

That's two updates in a row of Diamonds coming in well under what their scores said. Most score bands hold fewer than ten Diamonds, so it's a pattern to watch, not a settled fact. But it's the same story both times: the score tells you least at the top of the market. Three lost Diamond this time: Jacob deGrom, Corey Seager and Yandy Díaz.

What I can't tell you yet

Prices. The update landed on October 2, and a settled price two days after it doesn't exist until the evening of October 4. Whether these calls were worth buying is a different question from whether they were right, and it has its own post. I'll come back to it.

And it's one update. The 90s were 41 cards. 95% on 41 cards is a great result and a wide one. Read it as "the top of the board was very good," not "the top of the board is 95%."

Graded by Core against San Diego Studio's October 2, 2026 roster update, using the Diamond Radar board from the day before. The accuracy page adds it to the season record.

I'm Shaun. If something is wrong, tell me.