It is usually true. That is what makes it dangerous.

I wanted to know whether it is true here, so I built the only test that can answer it: compare each young Oriole not to his own past, but to every other hitter who was the same age that season.

My thesis going in, stated plainly:

1. The young core has stalled, and it is not just the eye test.
2. It is worse than normal variance for players this age.
3. It is happening to the players the organization spent the most on, not to everyone.

Two of those held up. The third held up in a way I did not expect, and it points somewhere uncomfortable.

First I needed to know what normal looks like. So I took every hitter in baseball with 250 or more plate appearances in consecutive seasons from 2024 through 2026, and measured how his OPS changed, grouped by the age he entered the season at.

This is the shape everyone assumes and almost nobody checks. Through the early twenties the median hitter gets better and most of the cohort improves. By the early thirties the median falls and most of the cohort declines. The cliff is real, and it arrives later than fans think.

Which means a decline at 23 or 24 is not a player being young. It is a player going the wrong way while his peers go the right way.

Now put the Orioles’ young hitters against that curve.

Four of the six measurable transitions sit in the bottom quarter of their age group.

Gunnar Henderson did it twice. Tenth percentile at 23, twenty-second at 24. In the two seasons a hitter is most likely to arrive, he went backwards further than three quarters of the players his own age, then did it again.

Colton Cowser’s 2025 was the sixth percentile of every age-24 hitter in the sample. Ninety-four percent of players his age had a better year than he did.

There is an obvious out here, and it is the one every fan reaches for. Batted balls find gloves. A .280 hitter has a .240 season and nothing is actually wrong.

Statcast can test that directly. Expected wOBA is built from how hard and at what angle a hitter hits the ball, not from where the ball landed. If the expected number holds while the real one falls, it was luck. If both fall, the hitting got worse.

FIG. 1 · ACTUAL AGAINST EXPECTED wOBA

Henderson's expected numbers fell with his real ones. That is not luck.

Henderson’s expected wOBA went .374, then .341, then .322. It fell by fifty-two points over two seasons, almost exactly tracking the drop in his real production. The contact quality went with it.

Cowser’s expected numbers tell the same story through 2025 and then partially recover, which matches his cohort rebound.

Holliday and Mayo come out differently. Both improved their expected numbers across the full window. Whatever is wrong here did not reach them, and pretending otherwise would be the easy version of this piece.

Here is where my third thesis point held up in a way I did not want.

The players going backwards are Henderson, Cowser and Holliday: the number one overall pick, the fifth overall pick, and the other number one overall pick. The players going forwards are Mayo, a fourth-round pick, and Samuel Basallo, an international signing.

That is not a story about young hitters. Every one of these players is young. It is a story about the ones the organization invested the most in and expected the most from.

I can think of three explanations and I cannot separate them with this data:

• Development. Something about how these hitters are coached here is not taking, and it shows up most in the players given the least runway to fail quietly.
• Deployment. The Orioles’ own lineup data showed the premium arrivals holding stable top-of-the-order jobs through long slumps while others were moved around. Protecting a struggling hitter in a role is a defensible choice and it is not the same as developing him.
• Regression to a truer level. Henderson’s .893 at 23 may simply have been the outlier, and everything since is the real player finding his level. If that is true it is the most expensive possibility of the three.

The Orioles spent a decade acquiring the right to draft players like this. The bet was never that they would be good immediately. It was that they would get better.

Two of them are getting better. The three the whole plan rested on are not, and in the one case we can test cleanly, it is not luck.

Tell me where I have this wrong. If you think Henderson’s 2024 was the aberration rather than his 2026, make the case, because that is the version of this where the Orioles are fine and I would genuinely like to believe it.

Definitions used: age_cohort_change@v1 · expected_woba_gap@v1 · Each links to the exact filter, the thresholds, and the argument against them.

Sources: MLB Stats API for season hitting, all clubs 2024-2026; Baseball Savant for expected statistics. Cohort built from every hitter with 250 or more plate appearances in consecutive seasons. The Bottom Row No. 005 · bottomrowsports.substack.com · Written by a fan, for an argument. Bring your own numbers.

I went into this one with a list of things I was sure about. Most of them held. One of them fell apart completely, and the way it fell apart changes what I think the Orioles should do this winter. I’ll take them in order, and I’ll flag the one I got wrong when we get to it.

Here is the number that should frame everything you think about the 2026 bullpen. Félix Bautista threw one and two-thirds innings. Ryan Helsley threw fifteen and a third. Seventeen innings, combined, from the two men the season was built around.

And the bullpen was still good. Tyler Wells posted a 2.38 ERA with a 0.90 WHIP. Rico García threw 67 innings across 75 appearances with a 0.97 WHIP. Yennier Cano came back with a 2.90. Josh Walker, a thirty-one-year-old nobody drafted a parade for, gave them thirty innings at 2.08.

That is a genuinely good relief corps assembled almost entirely from men who were not the plan.

Pete Alonso did what Pete Alonso does: 43 home runs, an .885 OPS over 700 plate appearances, and by Statcast’s expected numbers he was better than that line, not worse. His expected wOBA of .402 against an actual .379 says the contact quality was even stronger than the results.

The genuine surprise was Blaze Alexander. A .772 OPS in 289 plate appearances from a utility infielder is real value at a position where most teams are running out an automatic out.

And the third-base tandem worked. Coby Mayo hit 24 home runs with a .758 OPS at twenty-four and improved on his 2025. Christian Encarnación-Strand posted an .838 in 213 plate appearances.

This is the one that rewards looking past the back of the card, and it inverts the obvious reading.

FIG. 1 · ERA AGAINST EXPECTED ERA, 2026

The staff's best ERA belongs to its weakest underlying season.

Brandon Young led the rotation with a 3.56 ERA. His expected ERA was 4.32, the worst mark on the staff. Sixty-one walks in 156 innings is not a foundation; it is a tightrope he happened not to fall off.

Trevor Rogers ran the opposite way. A 4.03 ERA that looks ordinary, sitting on a 1.18 WHIP, 157 strikeouts against 39 walks, and a 3.60 expected ERA. He was the best starter they had, and the number on the scoreboard hid it.

Dean Kremer is the same story in miniature: a 4.68 ERA against a 4.15 expected, undone mostly by 18 home runs in 90 innings.

And here is why that matters more than a trivia point: Trevor Rogers is a free agent. The Orioles are most likely losing him.

A 4.03 ERA is what the rest of baseball sees on his page too, which means whichever club signs him is buying a 3.60-xERA starter at a fourth-starter price. The Orioles are not just letting their best starter walk. They are letting him walk at a discount they created by not looking past the ERA themselves.

If the winter plan is built off ERA, they will pay to keep the wrong arm and watch someone else get the right one cheap.

FIG. 2 · TEAM OUTFIELD OPS, 2026

Only Houston's outfielders hit worse. The league median is sixty points away.

An outfield OPS of .657, twenty-ninth of thirty, against a league median of .717. Not one Oriole outfielder with 150 plate appearances cleared a .710 OPS. Taylor Ward took 607 plate appearances and slugged .311. Leody Taveras took 482 and posted a .633.

You cannot win with this. No team does.

I want to be careful here, because the lazy version of this claim is wrong.

Some young Orioles improved. Coby Mayo went from a .687 OPS to .758. Samuel Basallo went from .559 to .697 at twenty-one. Adley Rutschman rebounded from .673 to .741.

But look at what happened to the players the organization spent the most to acquire:

Gunnar Henderson has now declined in three consecutive seasons, from ages twenty-three to twenty-five, through the years a hitter is supposed to be arriving rather than leaving. His expected numbers say he was somewhat unlucky, a .246 expected average against a .226 actual, but unlucky does not explain a 178-point OPS fall over two years.

The pattern is not “young players get worse here.” It is narrower and more worrying than that: the players who arrived with the highest expectations regressed, and the ones who arrived quietly improved. That is a development question, and it points somewhere uncomfortable.

I was certain about this one. An outfield that hits like that, I assumed, must also be giving away outs in the field. It would explain a lot, and it fit the story I already believed.

It is not true.

By Outs Above Average, the Orioles’ outfield finished -2, fifteenth of thirty. Dead average. And Colton Cowser posted +10 OAA in center, one of the better defensive seasons by any outfielder in baseball.

The defensive problem the Orioles do have is somewhere else. Dylan Beavers was -9 in left, the worst mark on the team, and he is also the outfielder whose bat went backwards. He is the one position player where both arrows point down.

I said last time that I could not answer the manager question from a spreadsheet. So I pulled every lineup card and every pitching change from all 166 games, and there is an answer in there. It is not about tactics. It is about who gets the job.

Start with the batting order.

Taylor Ward hit .670 and batted in the top four essentially every day, in four different slots all season. Colton Cowser hit better than Ward and was used in all nine batting-order positions.

The obvious objection is that churn follows slumps: a struggling hitter gets moved down, so instability is the manager reacting rather than the manager causing. That objection is why the table is sorted by production. It does not survive it. Ward was the worse hitter and kept the stable premium job. Henderson kept a top-of-the-order slot through a 178-point collapse. The players who got moved around were, on average, producing as well or better.

Stability tracked status. It did not track performance.

The bullpen says the same thing in a different language.

The two relievers with the worst ERA and the worst WHIP on this list collected nineteen of the saves. Ryan Helsley got eight of them in fifteen and a third innings. Tyler Wells, who was the best reliever on the team by some distance, got six.

So I will put my conclusion the careful way. This does not show a bad manager. Brandon Young’s ERA does not show a good pitcher either, and the point of this newsletter is that the obvious number is usually the wrong one.

What it shows is a manager allocating roles by standing rather than by results. On a veteran team chasing a title, that is a virtue: it is how you keep a clubhouse together through a slump. On a team whose entire theory of value is that young players develop into stars, it is the opposite of the job. You cannot both build around development and hand the defined roles to the players who are not developing.

That is the mismatch. Not competence. Fit.

Here is the honest accounting.

What is real: Alonso’s bat. Cowser’s glove. Rogers’ underlying performance. A bullpen that proved it can be rebuilt from spare parts. Mayo and Basallo trending up.

What is hope: That Henderson’s three-year slide reverses. That Holliday’s third season was the aberration rather than the trend. That Bautista and Helsley are healthy after contributing seventeen innings. That the outfield fixes itself internally when the internal options are the ones who just posted a .657.

That is a lot of hope, and most of it is attached to players whose last full season argued against it.

The mitigations that follow from the numbers rather than from optimism are narrow: buy outfield offense, because it is the single largest gap and it will not close from within. Make Rogers a priority, not a formality. He is a free agent with a 3.60 expected ERA wearing a 4.03 on his page, and the market reads the page. Stop building a bullpen plan around two arms with injury histories, not because Bautista and Helsley are bad, but because the 2026 team already ran that experiment and needed four unplanned relievers to survive it.

And the question I cannot answer with a spreadsheet, so I will just put it plainly: if the highest-expectation young hitters get worse while the quiet ones get better, that is either terrible luck three times over, or it is about how this team develops hitters. I do not know which. But two more seasons of it and the rebuild will not have produced a contender; it will have produced a very expensive lesson.

Tell me where you think I am wrong. I got the defense wrong in this piece and only found out because I checked. If you think I am wrong about Henderson, or about the manager question I have carefully not answered here, make the case and I will run the numbers.

Definitions used: outfield_offense@v1 · outs_above_average@v1 · era_minus_xera@v1 · lineup_stability@v1 · late_inning_role@v1 · Each links to the exact filter, the thresholds, and the argument against them.

Sources: MLB Stats API (season and roster statistics); Baseball Savant (Outs Above Average, expected statistics). Full 2026 season. The Bottom Row No. 004 · bottomrowsports.substack.com · Written by a fan, for an argument. Bring your own numbers.

You know the Monday routine. Open your phone and there are already a thousand takes on yesterday. This quarterback is clutch. That one isn’t. Somebody found one stat, posted it with little or no context, and by lunchtime it’s gospel, or at least it’s all over the radio shows and the podcasts.

I’ve been guilty of it too. But the more I watched, the more I kept asking follow-up questions nobody was answering, or even asking. What was the score when that happened? Who was coaching? Did the officiating change the momentum? How much time was left? Who was on the field around him? Would the answer change if you looked at it from a different angle?

What the name means

Everyone in sports talks about the bottom line. The final score, the stat line, the W or the L. One number that settles it. It grabs headlines and it engages people quickly.

This is about the row underneath.

Sports data doesn’t arrive as a bottom line, and it can be bent toward almost any narrative if you choose your numbers selectively. I would rather approach it the way Coach Lasso put it: be curious, not judgmental. We are lucky to have access to sources this rich, and they arrive as rows. One per play, and every row carries everything that play happened inside. The score at the time. The clock. The down and the distance. Who was on the field. What play was called. How likely each team was to win before the snap.

Very few people take the time to read those rows. The takes get made off the top line, and the row underneath goes uncounted.

The top line summarizes. The row remembers.

So when someone tells you a quarterback threw for 300 yards, that’s the top line. The bottom row is that 180 of them came after his team was already down 24 against a prevent shell. Same number. Different fact.

The top line is the outcome. The bottom row is the part nobody counted.

That’s the whole idea, and everything here comes back to it.

What I’m promising

I’ll tell you what I think before I look. Every piece starts with my thesis, stated plainly, so you can hold me to it. Then I go get the numbers.

I’ll show you the method. Every stat I use has a definition, and every definition lives on a page you can read: the exact filter, the thresholds I picked, and the arguments against picking them. If you think my definition is wrong, you can see precisely what to argue with.

I’ll tell you when I’m wrong. This is the one that matters, so let me prove it rather than promise it.

The first thing I got wrong

I just published a piece on the 2026 Orioles. Going in, I was certain about something. Our outfield was a big problem. They didn’t produce offensively, and they had to be giving away outs in the field too. Because did you watch them play defense?

It wasn’t as bad as I suspected. By Statcast’s fielding numbers the Orioles outfield finished dead average, fifteenth of thirty, and Colton Cowser had one of the better defensive seasons of any outfielder in baseball. As much as I miss watching Cedric.

I could have quietly dropped that section. Instead it’s in the piece, because the data told a different story, and that story changes what 2027 should look like. If the outfield were bad at both halves of the job, you have a much bigger hole to climb out of. Since the gloves were average, the glaring problem is somewhere else: the middle infield not hitting, the bullpen, development, fundamentals, small ball. I had the severity wrong, and being wrong about it opened up more areas actually worth researching before next season.

That’s the deal here. I don’t go looking for the picture I already want. I state the theory, then I go try to break it. I suspect I’ll be wrong often, and honestly that’s what makes this fun.

What you’ll get

A piece a week during the season, less often in the summer. Baltimore first, Ravens and Orioles, plus the occasional piece that’s just me chasing a number until it gives up.

There’s also a site at thebottomrow.com, where every chart is built from the data you can check, every definition has its own page, and there are tools you can actually use. One of them lets you ask whether a hitter is as dominant in baseball as a quarterback is in football, which turns out to be a harder and more interesting question than it sounds.

Now your turn

The best thing you can do here is disagree with me.

Reply to this email with a question you actually want answered, or a take you think I’ll get wrong. I’ll pull the numbers on the best ones and publish what comes back, credited to whoever asked. If your definition beats mine, I’ll rerun it with yours and show both.

Free, all of it, for now. Bring your own numbers.

Chris

Sources: MLB Stats API; Baseball Savant. Full 2026 season. The Bottom Row No. 000 · bottomrowsports.substack.com · Written by a fan, for an argument. Bring your own numbers.

The Ravens and Cowboys traded punches in Rio until the last seven seconds. Lamar hit Zay Flowers for 27 yards, called timeout with one second left, and Tyler Loop drilled a 56-yarder to win it 34-31. In Pittsburgh, Joe Burrow fumbled on the Bengals’ final drive and the comeback died right there. The Jets erased a 14-point hole to tie the Lions with four minutes left before Jahmyr Gibbs closed it out. The Bills turned it over five times and still needed two fourth-quarter touchdowns to put away the Chargers.

Three of those finishes turned on late defensive flags that erased a stop or set up a score. Denver won by four. Cleveland won by three.

That is how the NFL actually works. Most games are decided by a handful of plays and calls in the last few minutes. It is where reputations get made and narratives get written, and it is also where we most reliably forget that football is the ultimate team sport.

All these close games got me asking which quarterbacks actually show up when it matters, and whether “not clutch” is just one more label people use to minimize Lamar Jackson while ignoring what the numbers say about everyone else.

The stat that started it came from Scott Kacsmar after the Bengals loss: Burrow is 1-34 in games where he trailed in the last eight minutes.

My thesis, stated plainly so you can hold me to it:

1. Lamar is better than average in these moments, and probably great.
2. He gets measured differently than the other superstars.
3. He gets less help from the officials late in games.
4. He has lost games where he did his job and the kick or the defense did not hold.

Some of it held up. One part of it did not, and that is the part I have been chewing on.

Start where everyone starts: the scoreboard. When you are behind with eight minutes to go, do you find a way to win?

FIG. 1 · WIN RATE WHEN TRAILING LATE

Lamar wins these games more often than all but five quarterbacks in the league.

Lamar wins 23.9% of the time against a 16.5% league average, sixth among forty active quarterbacks with at least fifteen of these games. Not bad for a guy who supposedly cannot close.

Mahomes is the only true outlier. Allen and Lamar are nearly identical, and only one of them gets called not clutch.

Burrow is not just below average, he is the worst of anyone with a real sample, by a distance. His team puts him in these spots as often as Lamar’s does, so this is not about opportunity.

One caution that governs everything below: a win-loss record is a team statistic. It contains the defense, the kicker and the play caller.

Records are fun to argue about. They do not tell you who made the plays. So I went snap by snap, and this is where I was wrong.

FIG. 2 · EPA PER PLAY, HIGH STRESS

And his play-by-play in those same moments is ordinary. This is the thesis point that failed.

Lamar wins these games, but his individual plays in them are only average for this group. That is my second thesis point, the one where I said he was probably great, coming apart in my hands.

Nobody here is padding. Garbage time is 14% to 20% of everyone’s plays, and those snaps are excluded from the first two columns entirely. Lamar’s garbage-time snaps mostly came while the Ravens were already winning big, not while chasing a lost game.

I counted the four defensive penalties that bail out an offense, per 100 dropbacks, in the last eight minutes when the offense is tied or trailing.

FIG. 3 · BAILOUT FLAGS PER 100 DROPBACKS, LATE AND TIED OR TRAILING

He draws about a third fewer drive-extending flags than the league, and half what Mahomes gets.

Lamar draws about a third fewer late bailout flags than the league average, and roughly half what Mahomes gets. A third of Mahomes’ late comeback wins included one of these flags on the drive.

Give Lamar credit for an average catch on every high-stress drop, and erase the fumbles his receivers lost, and his high-stress EPA nearly doubles, from 0.075 to 0.144 per play. Only Mahomes and Hurts gain more. He still ranks seventh of nine, so it does not flip the verdict, but it closes almost half the gap.

Here is why a drop hurts so much: EPA charges every incomplete pass to the quarterback, even when the ball hits the receiver in the hands. A lost fumble after the catch is worse, because the turnover lands on his play too. Running back fumbles on handoffs, like Derrick Henry’s in Buffalo, do not count against him at all.

Lamar’s pass catchers drop the ball twice as often as Burrow’s when it matters, and his teammates lose fumbles at the second-highest rate here. Ravens fans do not need the table:

• Buffalo, 2025 opener. Up 40-32 with three minutes left, Derrick Henry lost a fumble. The Bills scored twice and won 41-40.
• Buffalo, 2024 playoffs. Mark Andrews lost a fourth-quarter fumble, then dropped the tying two-point conversion with 1:33 left. FTN charted a second drop on him in that game too, in the second quarter with the score tied and the Ravens at the Buffalo 28. Two in one game is tied for his most since charting began in 2022. Ravens lose 27-25.
• Indianapolis, 2023. Isaiah Likely dropped a pass in a tied game with 3:28 left. Ravens lose in overtime.
• Jacksonville, 2022. Gus Edwards lost a fumble, Kenyan Drake dropped a pass on the next drive. Ravens lose 28-27.

Two more that every Ravens fan reaches for, and a caveat about both that I would rather state than have you catch.

• Kansas City, the 2023 AFC Championship. Fourth quarter, Ravens down 17-7, Lamar hits Zay Flowers for eight yards to the Kansas City 1. Flowers reaches for the goal line, the ball comes out, and Kansas City recovers in the end zone. Touchback. The play cost 5.23 expected points, a hair behind Henry’s fumble in Buffalo for the most damaging single play in this section.
• Cincinnati, Week 13 of 2025. Lamar finds Isaiah Likely for what the officials signal as a 44-yard touchdown. Replay reverses it: Likely had fumbled at the 1 and the ball rolled through the end zone. Touchback, no points, and 3.20 expected points off Lamar’s ledger.

Here is the caveat. Neither of those two is in any number above this line. The Flowers fumble came with the Ravens down ten, outside the one-score window that high_stress@v1 measures, and the Likely reversal was the second quarter of a one-point game rather than the fourth. They are the two plays people remember, and the filter this piece runs on excludes both.

That cuts in my favor, which is exactly why it is worth saying. Widen the window to catch them and the teammate-cost case in this section looks stronger than the numbers I showed you, not weaker. I am leaving the window where it is, because moving it after seeing which plays it catches is how you talk yourself into anything.

The playoffs show it hardest. Lamar has only 16 high-stress playoff plays. Two of them are Andrews’ fumble and dropped conversion in Buffalo. Those two drag his high-stress playoff EPA from 0.52 per play, which would lead this group, down to 0.13. A sample that small proves nothing, and it is still the clearest picture of a drive he finished and his teammates did not.

FIG. 4 · LOSSES AFTER LEADING INSIDE TWO MINUTES

Twelve games he had already won, and lost anyway. The most in this group.

Miami 2022, up 21 in the fourth, lost 42-38. Cleveland 2023, a 14-point fourth-quarter lead into a 33-31 loss. Buffalo 2025, up 15, lost 41-40. Pittsburgh 2025, a missed 44-yarder in a 26-24 loss.

To be fair, two of Lamar’s three missed-kick losses were from 61 and 67 yards, and Burrow has it worse with five. The broader point stands: every one of those twelve is a loss on his record, and several came after he had already put his team ahead.

Mahomes starts his final drive with about two timeouts, double almost everyone. That is repeatable, and it belongs to the coach.

The Ravens leave value on the table. They go for it on late short fourth downs less than league average and convert 78% when they do.

Zac Taylor punts the most when trailing late. Put that next to the Bengals’ 6-52 record when trailing after three quarters and the Burrow problem looks less like a quarterback problem every minute.

Here is where I landed. Lamar wins late more than most, with less help from the officials and less support at the finish. The not-clutch label does not survive contact with the numbers. But if you want to call him great in those moments, the play-by-play says there is still room.

Burrow is the flip side. He plays well late and almost never wins. That is not a quarterback problem, it is an organization problem.

And the officiating matters beyond Lamar. Mahomes, Stafford and Herbert each draw about twice his rate of drive-extending flags late. When games are this close this often, those calls decide outcomes. The league should allow replay review of pass interference and roughing the passer in the fourth quarter and overtime, not to re-referee every play, but to fix the obvious misses.

The biggest thing this data reminded me of is simple. Football is the ultimate team sport. We love to boil it down to the quarterback, and I did plenty of that here. The numbers keep pulling it back to everyone else. A receiver has to catch it. A back has to hold it. A defense has to get one more stop. A kicker has to make the kick. Lamar can put his team ahead with two minutes left and it still does not count if the other fifty-two do not finish.

Which Ravens late-game collapse still keeps you up at night, and was it the quarterback, the hands, the defense, or the whistle? Tell me and I will pull the numbers on the best answer.

Definitions used: high_stress@v1 · trailing_late@v1 · bailout_flag@v1 · garbage_time@v1 · Each links to the exact filter, the thresholds, and the argument against them.

Sources: nflverse play-by-play, 2018 through Week 3 of 2026, regular season and playoffs; FTN charting via nflverse (CC BY-SA 4.0) for drop and pressure data. Reporting credited inline. The Bottom Row No. 002 · bottomrowsports.substack.com · Written by a fan, for an argument. Bring your own numbers.

My thesis, stated up front so you can hold me to it:

1. Pass rush closes games more than any other position group. Not blitzes, stunts, or gimmicks, but real penetration from a four-man front.
2. Defenses built on blitzing or coverage eventually get picked apart in the final minutes, when quarterbacks have time and nothing to lose.
3. The Ravens have invested heavily in the secondary and neglected the pass rush for most of a decade, and that lines up with their defense faltering late.
4. Even the Ravens’ good pressure years were driven by scheme, not players.

We all blame the coordinator or the cornerback when a late lead slips away. I suspect the real problem starts up front.

I looked at every dropback since 2022 in the fourth quarter or overtime where the offense trailed by one score or was tied and neither team was better than a 95% bet to win, meaning the defense was trying to close a game still in doubt. Then I split them by whether the defense rushed four or fewer or blitzed with five or more, and whether the quarterback was pressured.

FIG. 1 · EPA ALLOWED, LATE AND ONE-SCORE

Pressure decides the snap. How it arrives barely matters.

EPA allowed measures how much each play helped the offense’s expected points. Lower is better for the defense.

Three takeaways:

• Pressure is the whole ballgame. A pressured dropback is a disaster for the offense, no matter how the pressure arrives. An unpressured one is a win for the offense.
• A blitz that doesn’t get home is the worst outcome on the board. That part of my thesis held up. When the extra rushers don’t get there, the quarterback picks apart the thinner coverage behind them.
• But on average, blitzing late wasn’t worse. Blitzes got home more often, and that roughly paid for the times they didn’t. This is where the data pushed back on me.

One limit worth knowing: pressure here only counts sacks and hits. Hurries that force an early throw aren’t in the public data, so real pressure rates are higher for everyone.

Okay, so pressure wins the snap. But does a great pass rush win you the season’s close games more than great coverage does? This is the heart of my thesis, so I was nervous about this one. Turns out it’s closer to a tie than I wanted. Teams in the bottom quarter of either one hold about 65% of their late leads. Teams in the top quarter hold about 77%.

FIG. 2 · EVERY DEFENSE, 2018-2025

The best defenses sit up and to the right. The Ravens are scattered.

I measured every defense from 2018 through 2025 two ways: how often it sacked or hit the quarterback, and how well it covered on dropbacks where it didn’t get pressure. Then I checked which one better predicted holding a one-score lead with eight minutes left.

When I weighed both together, pass rush had a slightly bigger effect on holding late leads. Coverage had a slightly bigger effect on how many yards and points a defense gave up late. The fair read is a tie, with a small edge to the rush where it matters most: the final score.

There’s also a catch. Good coverage makes pressure easier, because quarterbacks hold the ball longer. And a good rush makes coverage look better, because throws come out early. The two feed each other, which is why the best defenses usually have both.

Most contenders sit above or to the right of the medians, and the best ones sit in both. The Ravens’ seasons are scattered. Their 2023 defense sat near the top of the league in coverage. Their 2021 and 2025 defenses fell into the weak-rush, weak-coverage corner, and those two seasons ranked 30th and 32nd in late-game defense.

Then I looked at the teams we all want to be. I figured every champion would have a monster pass rush. Some did. Some won with elite coverage and a pretty average rush.

Ranks are out of 32 teams that season. Four-man pressure rank is only available from 2022 on.

What stands out:

• The 2022 Eagles and 2025 Broncos fit my thesis perfectly. Both led the league in four-man pressure without leaning on the blitz.
• The 2024 Eagles broke it. They ranked 21st in pressure and won a title with the league’s best coverage, while blitzing less than almost anyone.
• Almost none of these teams blitzed heavily. Only Denver ranked in the top 10. That part of my thesis held up: contenders get pressure with their front, or they cover, but they don’t manufacture it.
• The 2021 Rams surprised me. With Aaron Donald and Von Miller, I expected them near the top. By sacks and hits they ranked 24th, a reminder that this measure misses hurries and double teams.

Now the part I really wanted to get to. I’ve been saying for years that the Ravens pour everything into the secondary and leave the pass rush to chance. So I checked the draft history. I was half right. They drafted plenty of pass rushers since 2016, but the premium picks went to the secondary.

I counted every top-100 pick from 2016 through 2026 and weighted each one by its trade-chart value, so a No. 14 pick counts far more than a No. 93 pick.

The pattern is clear once you list the names. The secondary got four first-rounders: Marlon Humphrey (No. 16), Kyle Hamilton (No. 14), Nate Wiggins (No. 30), and Malaki Starks (No. 27). The pass rush got volume on Day 2 and 3, plus one late first-rounder in Odafe Oweh (No. 31).

The results show the difference. Humphrey and Hamilton became All-Pro caliber players. On the front, only Nnamdi Madubuike, a third-rounder, became a true star. Kamalei Correa, Tim Williams, Jaylon Ferguson, and David Ojabo all delivered little for the Ravens.

So it isn’t quite true that the Ravens ignored the pass rush. They kept drafting it, just cheaply, and those picks mostly missed. Two things this doesn’t cover: free-agent spending, where the Ravens rented veterans like Jadeveon Clowney and Kyle Van Noy, and cap dollars, which I didn’t measure here.

Last piece: I wanted to see how the Ravens’ rush actually looked, year by year, in the moments that matter. Remember the Wink Martindale years, when it felt like we blitzed on every snap? It turns out we basically did, and it caught up with us late. But the 2023 sack record wasn’t what I thought it was.

Late defense rank is EPA allowed per play in the fourth quarter and overtime while protecting a one-score lead or a tie.

What stands out:

• 2019 and 2020 were scheme-driven pressure. The Ravens blitzed on about half of all dropbacks, more than any team in football, and 57% to 71% of their pressure came from blitzes. In 2020 they ranked third in pressure and dead last in late defense. That’s my thesis in one season: manufactured pressure that got picked apart when it mattered.
• 2023 wasn’t what I thought. The 60-sack season came mostly from four-man rushes, with fewer blitzes than average. One caveat: Mike Macdonald loved simulated pressures, where a linebacker rushes and a lineman drops. Those look like four-man rushes in the data. So some of that pressure was still scheme, just a kind the numbers can’t separate.
• 2025 was the bottom. 22nd in pressure, last in late defense, and zero late leads held. The Ravens followed it by drafting another second-round edge rusher, Zion Young, in 2026.

Over eight seasons, the Ravens’ late defense ranked in the bottom third four times. In most of those years, the rush was either average or leaning on the blitz.

Pressure wins the final minutes. But I overstated how much the pass rush matters compared to coverage, and I had parts of the Ravens story wrong.

Here’s where I landed. The best defenses in close games get pressure without selling out for it, and they cover behind it. You can win with an elite version of either, but a defense that has neither, or fakes the rush with blitzes, is the one that gives up the late drive.

For the Ravens, the problem wasn’t a lack of effort on the pass rush. It was a lack of premium investment and poor hit rates. Spending first-round picks on Humphrey and Hamilton worked. Hoping mid-round rushers would develop mostly didn’t.

Now I want to hear from you. Which Ravens late-game collapse still keeps you up at night, and was it the rush, the coverage, or something else? And if you think I’m underrating the pass rush, make the case in the comments. I’ll pull the numbers on the best arguments.

Definitions used: late_one_score_defending@v2 · dropback@v1 · pressure@v1 · blitz@v1 · coverage_epa@v1 · Each links to the exact filter, the thresholds, and the argument against them.

Sources: nflverse play-by-play 2018-2025; FTN charting via nflverse (CC BY-SA 4.0) for pass-rusher counts, 2022 on. Through 9/28/26. The Bottom Row No. 003 · bottomrowsports.substack.com · Written by a fan, for an argument. Bring your own numbers.

I feel bad for this team. Baseball is the hardest sport in the world because more effort does not produce more results, and you can see the effort. The pitching has been terrific for the last few weeks. The problem is that the Orioles cannot get offense, defense, and pitching to show up on the same night, and at 69-75 the 2026 season is over. Ownership spent this winter. Pete Alonso has been everything they paid for, with 33 home runs and a 140 wRC+, and if he had any protection behind him he would be better still. Outside of him, the outfield produces nothing, third base and shortstop are wildly inconsistent, the bullpen is hurt, and the best player on the field most nights is a utility man. That is not a managing problem or an effort problem. It is a talent problem, and talent problems belong to the front office.

FIG. 1 · THE SHAPE OF THE REBUILD

Winning percentage peaked in 2023 and has declined every season since.

2020 was a 60-game season. 2026 through September 6. Source: Baseball-Reference.

When other teams sat at the top of the draft, they came away with franchise players. Baltimore came away with regulars. Adley Rutschman at #1 in 2019 was the consensus pick, but Bobby Witt Jr. went one pick later and Corbin Carroll went 16th, and Rutschman has now been traded to Boston for prospects without ever becoming the player a #1 pick is supposed to be. Heston Kjerstad at #2 in 2020 is indefensible: a signability reach who has been worth negative WAR while Garrett Crochet went 11th and Pete Crow-Armstrong went 19th. Colton Cowser at #5 in 2021 had one good year and has regressed since; Jackson Merrill went 27th. Jackson Holliday at #1 in 2022 has 1.5 career WAR while Zach Neto, a college shortstop taken 13th, is pushing 16. Enrique Bradfield at #17 in 2023 is a glove without a bat, still in Norfolk and still not on the 40-man roster, while Kevin McGonigle went 20 picks later and Nolan McLean went in the third round. Vance Honeycutt at #22 in 2024 went two picks after Trey Yesavage. You cannot have that much horsepower at the top of the draft and come away with nothing better than a solid regular.

FIG. 2 · WHAT THEY TOOK VS. WHAT WAS ON THE BOARD

Career bWAR: the Orioles' first-round pick against the best player still available.

Career Baseball-Reference WAR through September 6, 2026. Also still available: Carroll 20.5 (2019, #16) · Crochet 11.4 (2020, #11) · Misiorowski 5.7 (2022, #63).

Seven first-round picks with elite draft capital have produced, in total, roughly what one second-round pick has produced by himself.

Every homegrown hitter peaks in his first or second full season and declines from there. Rutschman put up a 131 OPS+ as a rookie and never matched it. Cowser was a 120 OPS+ hitter in 2024 and has been below average both years since. Holliday has never been an average bat. And Gunnar Henderson, who we still talk about as a superstar, has gone from a 155 OPS+ and fourth in the MVP voting in 2024, to 122 last year, to 99 this year, at exactly the age when he should be entering his prime. At this point 2024 looks like the outlier, not the baseline. He has led the league in errors at shortstop twice in the last three seasons and the bat is headed the wrong way. If this were a business chart, the trend lines on our most valuable assets are all pointed down, and the one hitter who left, Kyle Stowers, became an All-Star somewhere else.

FIG. 3 · OPS+ BY SEASON, HOMEGROWN HITTERS

Every line peaks early and comes down.

Source: Baseball-Reference. 2026 through September 6. Rutschman's 2026 combines Baltimore and Boston. Westburg missed 2026 (Tommy John).

Home runs do not make a hitter, and I think this organization has become so obsessed with launch angle and three true outcomes that it has forgotten what a good hitter looks like. Ichiro and Tony Gwynn were great hitters. The Orioles are a below-average three-true-outcomes team without the outcomes. And here is the thing: the numbers agree with the eye test. Henderson is not striking out more than he used to; his strikeout rate has been flat around 21 to 22 percent for three years. What has collapsed is the quality of his contact. His hard-hit rate has fallen from 54 percent in 2024 to 49 percent to 43 percent this year, and his average exit velocity is down more than three miles an hour. He is not swinging and missing more. He is hitting the ball softer. When something like that happens to almost every hitter who comes through the system, it is not seven separate personal problems. It is coaching and development. Elias provides the talent, and then the talent gets worse.

FIG. 4 · THE HENDERSON DIAGNOSTIC

He is not whiffing more. He is hitting it softer.

Source: Baseball-Reference advanced batting. Average exit velocity fell from 92.8 mph (2024) to 89.4 mph (2026).

Then there is pitching. Every impact arm this team has had came from somewhere else. Kyle Bradish was a trade. Trevor Rogers was a trade. Shane Baz was a trade. The pitchers Baltimore could have had after its own first-round pick over these seven drafts include Crochet, George Kirby, Hunter Brown, Andrew Painter, Spencer Schwellenbach, Bubba Chandler, Cade Horton, Jacob Misiorowski, Nolan McLean, Trey Yesavage, and Payton Tolle. That is a rotation and a half. The free-agent record is no better: Charlie Morton, Tomoyuki Sugano, and Tyler O'Neill at three years and $49.5 million to hit .216. Elias did offer Ranger Suarez five years and $125 million and Boston matched it, so some of this is players saying no. But outside of Alonso, this front office has not made a free-agent addition that mattered in years. I'll come back to pitching below, because the numbers are worse than I thought.

Everyone points to Houston because that is where Elias came from and the Astros ran the same tank-to-draft model. Look at what they got out of it: Carlos Correa, Alex Bregman, Kyle Tucker, George Springer. Here is the part that stings. Elias was the Astros' scouting director for those drafts; Correa, Bregman, and Tucker were his picks. The same evaluator found stars in Houston and has found regulars in Baltimore. What he did not bring with him was the rest of the machine: a development staff that turned a seventh-round pick named Dallas Keuchel into a Cy Young winner, an international pipeline that produced Framber Valdez and Cristian Javier, and an owner willing to trade for Verlander and Cole. (Altuve, for the record, was a $15,000 international signing before the tank, not a draft pick.) Baltimore copied the draft model and left the other three parts in Houston.

I started out blaming the drafts, and then I started to wonder whether the real problem is development, because the picks keep arriving and then getting worse. So I tested it, and the answer is more specific than either. I looked at every hitter in baseball who was 26 or younger and had 200 plate appearances in back-to-back seasons for the same club, 2023 through 2026, and measured how much his wRC+ moved. The Orioles' young hitters changed by an average of minus 1.9 points. The league average was minus 1.4. That is the middle of the pack. Milwaukee was plus 0.9; Tampa Bay, for what it is worth, was minus 3.0. Baltimore's number includes Henderson's two big drops, but it also includes Holliday going from 63 to 96, Westburg from 97 to 125, Mayo from 95 to 104, and Basallo from 56 to 99. The trend line that scares me is really three players: Henderson, Cowser after 2024, and Rutschman until this year.

Best in the test: Milwaukee, +0.9. Tampa Bay: −3.0. This is not a development collapse, it is three players: Henderson, Cowser after 2024, and Rutschman until this year.

The reclamation projects tell the same story. Ryan O'Hearn was a waiver claim who hit at 118 for two straight years here and 128 last year. Ramon Laureano was a 90 wRC+ hitter the year before Baltimore got him, a 138 in his Baltimore season, and is an 86 in San Diego this year. Blaze Alexander is at 121. Cam Sanders had given up nine earned runs in nine outings for Pittsburgh before the Orioles bought him for cash in July; here he has a 2.81 ERA and a 31 percent strikeout rate. Whatever the big league coaches are doing with veterans, it works. And the theory that our guys get better when they leave is really just Kyle Stowers. Joey Ortiz went from 105 to 67 to 96 in Milwaukee. Connor Norby has been a 90 and an 89 in Miami.

The veteran fix works. The homegrown decline is a different machine.

Where Baltimore is unusual is at the top of the prospect list and at Triple-A. Three different Orioles entered a season as the number one prospect in baseball: Rutschman in 2022, Henderson in 2023, Holliday in 2024. Add Basallo and Grayson Rodriguez in the top ten, and Mayo and Kjerstad in the top 40. No organization has had that many elite prospects arrive in that short a window and produced zero stars from them. And the bat-first guys, Kjerstad, Mayo, Norby, Stowers, Beavers, all destroyed Norfolk and arrived as something less. The International League inflates numbers, and this hitting model (patient, pull-side, lift) produces hitters who look finished at Triple-A and get exposed by big league velocity and spin. That is the real minor league development problem: our Triple-A numbers have been a bad forecast for years, and the organization keeps believing them.

Put it together and the picture is a median outcome from top-of-the-draft inputs. Baltimore ranks 14th of 30 in 2026 WAR from young players, which is fine for a normal team and a failure for one that spent five years at the top of the draft. And look at where the good teams get it. Milwaukee is second in baseball, and 14.8 of its 24.2 young-player WAR comes from pitchers 27 and under. Pittsburgh gets 14.1 from young pitchers. Cleveland 13.6. Baltimore gets 5.0, and 3.5 of that is Shane Baz, who was drafted by Pittsburgh and developed by Tampa Bay. Which brings me to pitching.

FIG. 5 · 2026 WAR FROM YOUNG PLAYERS STILL WITH THE CLUB THAT PRODUCED THEM

The contenders get theirs from young pitching. Baltimore gets almost none.

FanGraphs, through September 6. Players still with the team that produced them. 3.5 of Baltimore's 5.0 pitching WAR is Shane Baz, acquired by trade.

We take pitchers late and assume we will develop them. The record says we do neither. From 2019 through 2024 the Orioles drafted 60 pitchers who signed. Four have reached the majors. Combined, they have produced 0.3 wins above replacement. The best arm Elias ever drafted, Nolan McLean, did not sign; he went back to Oklahoma State, went to the Mets a year later, and has 4.4 WAR. Over the same six drafts, Seattle's pitchers have produced roughly 34 WAR (Kirby, Woo, Bryce Miller), Cleveland's about 33 (Bibee, Gavin Williams, Cade Smith, Parker Messick), and Pittsburgh's 26 (Skenes, Jared Jones, Chandler). Boston and the Yankees, complete counts, are at 9 and 10.

FIG. 6 · CAREER MLB WAR FROM PITCHERS DRAFTED 2019–2024

What six drafts of pitching produced, team by team.

BAL, TBR, BOS, NYY: every signed pitcher, B-Ref WAR through Sept 6, 2026. SEA, CLE, PIT, DET, MIL: notable draftees only (FanGraphs), those bars are floors, not totals.

It is not that the late picks fail. It is that the picks never come. Elias has never used a first-round pick on a pitcher. The first pitcher he took in 2019 came in the 8th round, 228th overall. In 2020 and 2021 it was the 5th round. The earliest he has ever gone is the second round, twice, and the 2025 pick, Joseph Dzierwa at 58, was the first real pitching investment in seven years. Thirteen of the 22 picks in 2023 were pitchers, twelve of them in the third round or later. This is a strategy: take bats early because they are safer, fill the back of the draft with arms, and let the development system sort it out. The development system has sorted out four of sixty.

FIG. 7 · THE FIRST PITCHER TAKEN IN EACH ELIAS DRAFT

Never the first round. Never close.

Bar height = round of the first pitcher taken. *2022's first pitcher, Nolan McLean (rd 3, #81), did not sign, he has 4.4 WAR for the Mets; the first signed pitcher that year came in round 5. Source: Baseball-Reference draft records.

So the staff gets built the other way. Of the roughly 1,280 innings the Orioles have thrown this season, 46 percent came from pitchers acquired by trade: Baz, Bradish, Rogers, Cano, Kremer, Povich, Sanders. Another 15 percent came from minor league free agents and waiver claims, 13 percent from major league free agents, and 11 percent from Brandon Young, an undrafted free agent. Pitchers Elias actually drafted threw 34 innings, less than 3 percent, and those belonged to Trey Gibson, a 7th-rounder with a 7.36 ERA, and Cameron Weston, one inning. The four-man core of the rotation, Baz, Bradish, Rogers, and Young, was drafted by Pittsburgh, the Angels, the Marlins, and nobody.

FIG. 8 · WHERE THE 2026 ORIOLES' INNINGS CAME FROM

A pitching staff built almost entirely from other organizations' arms.

FanGraphs, through September 6. The Elias draft picks: Trey Gibson (2023, 7th round, 7.36 ERA) and Cameron Weston (2022, 8th round, one inning).

Here is the part that should sting, because it is the thing the Orioles are actually good at. Bradish was an Angels draft pick who has been worth nearly 9 WAR since arriving. Kremer came in the Machado trade and has 8. Cano was a throw-in from Minnesota. Tyler Wells was a Rule 5 pick. Rogers was written off in Miami and has been worth 6.6 WAR in a season and a half here. Baz is the newest version. This organization can take other teams' arms and make them better. That is the Rays' model, and for what it is worth, the Rays' own drafted pitchers from these same years have produced 0.7 WAR, almost as little as ours. The difference is volume. Tampa Bay and Milwaukee acquire young arms in bulk, every year, on purpose. Baltimore did it two or three times in seven years and only went big this winter, after the window had closed. You can build pitching by drafting it or by trading for it and fixing it. We chose neither, and then hoped.

Sparklines: OPS+ by season, 2023–26, against the league-average line (100). Not one of the eight is a top-five player at his position today.

So where does that leave us? Today the Orioles do not employ a single above-average position player that Mike Elias drafted. Not one homegrown player is a top-five player at his position. Jordan Westburg, the guy I love, has now missed most of two seasons. Blaze Alexander, a reactionary pickup made because the infield got hurt and because Coby Mayo cannot play third base, was one of the few bright spots until he broke his hand in July, and when a stopgap is your best move of the year, the stopgap is the story. The problem is not that Elias cannot find talent, and it is not that the coaches cannot make players better; the veterans prove they can. It is that the model picks hitters with a low ceiling, trusts Triple-A numbers that do not translate, and treats pitching as something that will show up on its own. Seven drafts with top-five picks, zero stars, zero pitchers, and a core whose trend lines all point down. We can argue about the manager all we want. It comes back to Elias not being very good at his job, and it is surprising that he still has it. It sucks to watch it play out.

Sources: Baseball-Reference (OPS+, WAR, fielding, advanced batting) through games of September 6, 2026; wRC+ via FanGraphs and Camden Chat; draft positions and outcomes from Baseball-Reference draft records; young-player and acquisition data from FanGraphs; rotation notes via The Baltimore Sun and Camden Chat. The Bottom Row No. 001 · bottomrowsports.substack.com · Written by a fan, for an argument. Bring your own numbers.