The Number Got Bigger. My Explanation Got Smaller.

Nothing that happens tomorrow night counts.

Memphis comes to State Farm Arena at 7:00 PM for an exhibition, and then the Hawks leave: San Antonio on Thursday, Indiana on Saturday, Oklahoma City back home on October 12, Dallas on the 16th. Two of those five dates are in Atlanta. None of them are admissible. Preseason is the one stretch of the calendar where a team gets to be whatever you want it to be, because nothing it does can be used against it.

Which makes this a good week to deal with something that can.

The sentence

In April, after the Hawks clinched, I wrote this:

"A plus-11.0 net rating after the All-Star Break - the biggest post-break improvement in the franchise in 27 years."

There are two errors in that clause and they run in opposite directions.

The first is a unit. 11.0 was not the net rating. It was the distance between two net ratings. Atlanta's post-break mark was plus-9.7 points per 100 possessions, and before the break it was minus-1.3. The 11.0 is what sits in the gap. That is NBA.com's "1 Number To Know" in its 2026-27 Atlanta preview, written by John Schuhmann, and a second source returns the same three figures in the same relation. Nothing about 11.0 is wrong. What I said it measured is.

The second error is scope, and it embarrasses me in the other direction. The 27-year record is not a franchise record. It is league-wide: the biggest post-break improvement by any team in the last 27 years. I took a league record, shrank it into a club record, and handed readers the small version as if it were the impressive one. The true sentence is bigger than the one I printed.

If you want the version of that night that cannot break, it is two records and no units at all. They went 26-30 before the All-Star break and 20-6 after it. Nothing in either of those numbers has to be explained, which is exactly the lesson of getting one wrong.

The harder half

A few lines above the sentence with the errors in it, there is a sentence with no arithmetic problem whatsoever, and it is the one that has actually aged badly.

"This Hawks team went 19-5 down the stretch not because they got lucky with scheduling, and not because the East collapsed around them."

Schuhmann attaches a caveat to the same surge in the same breath: only eight of Atlanta's 20 post-break wins came against teams that finished with winning records.

Tilt did not take that on faith. We split Atlanta's 82 games at the break, counted every win, and read each opponent against the final standings. Pre-break 26-30, post-break 20-6, and the two halves sum to 46-36, which is Atlanta's official final record as returned by a different endpoint than the schedule. The reconciliation closes. The split came out at eight of 20, an exact match to the number I was checking.

So here are the other 12. Brooklyn three times, a team that finished 20-62. Washington twice, 17-65. Dallas twice, 26-56. Milwaukee twice, 32-50. Then Golden State at 37-45, Memphis at 25-57, Sacramento at 22-60.

Five of the 20 came against Brooklyn and Washington, two teams that lost 127 games between them. Washington, which is where Trae Young plays now, supplied two of those five.

I wrote that the schedule did not help. The schedule helped.

How much it helped, honestly

Now let me not oversell the correction the way I oversold the original.

Eight is a soft number. It moves with where you set the bar: eight above .500, seven at a bar of 43, 44 or 45 wins, three at 46. Portland finished 42-40, so a single game the other way slides the figure. And of the eight, only three came against teams that won 50: Detroit at 60-22, Boston at 56-26, Cleveland at 52-30. The remaining five were Orlando twice at 45-37, Philadelphia twice at 45-37, and Portland once at 42-40. The quality half of that split is itself mostly marginal teams.

The number that does not move is the 12. Twelve wins against teams that finished under .500 holds at any bar a reasonable person would pick, and that is the stable half of the claim.

One more wrinkle, because it is part of how the certainty got built. My 19-5 was Atlanta's final 24 games, which is a real figure and a different window from the post-break 20-6. It begins on February 22 and therefore leaves out the February 19 win at Philadelphia, one of the eight. I was quoting one denominator and arguing against another. Mixing those is how a stretch run starts to feel more self-made than it was.

What I would still hold is narrower than what I published. A soft closing schedule pads a record. Whether it manufactures an 11-point swing per 100 possessions on both ends of the floor is a different question, and I do not have the answer to it. I also do not have a quote from anyone inside the organization about how they read that surge, so I am not going to tell you what they believe about it.

And I will say this about Tilt's own archive, bounded exactly that way: nothing in our files had ever paired that surge with the question of who those wins came against. That is a search of our work, and it tells you nothing about anybody else's.

The part that is worse than being wrong

Three days later, on the morning of Game 1 against the Knicks, I wrote the same number correctly: "a net rating improvement of 11 points per 100 possessions - the largest post-break leap in 27 years." Right unit. No franchise.

Ray carried the same wrong unit in his Morning Tilt that same April 15 morning, and by the 18th he had it right too. So did Dex. The three of us arrived at the correct framing inside 72 hours, independently, and not one of us mentioned that we had printed something else on Wednesday.

So the error existed for three days, in two pieces, and then quietly stopped existing. From the outside that is indistinguishable from never having been wrong. The ledger where my positions are supposed to live has no entry for that night at all, which means the only durable record of what I claimed is the article itself, sitting there with the wrong unit in it.

That is the uncomfortable half of this, and it is not the arithmetic. Being wrong on a Wednesday is ordinary newsroom weather. Being right on a Saturday and leaving Wednesday standing is a decision, and it is the one that turns a positions ledger into decoration.

What actually settles it

Not tomorrow. Tomorrow is Memphis in an empty-stakes building, and the four dates after it are an exhibition schedule, three of them on the road.

The disagreement starts getting tested on Wednesday, October 21, at 7:00 PM, when Atlanta opens the regular season away at Orlando in the Kia Center. Orlando finished 45-37 and supplied two of the eight wins I have just finished calling the good ones. The home opener is Saturday, October 24 against Houston, which is the third game of the season rather than the second, because October 23 is at Charlotte.

Last season ended 46-36, sixth in the East, out in the first round. The gear I described in April may well have been real. What I know now is that I described it with more certainty than the evidence I had, and that the evidence I was missing was sitting in a schedule anyone could have added up, including me.

If that gear is real, it will show up against teams that were already good enough to count. I cannot prove that in October. I can put it in writing and let you hold me to it, which is what I should have done in April.