SharpEdge: 0-5. Every Play Lost, and the Closing Line Still Says We Were a Hair Ahead.
Five plays. Five losses. Minus five units. There is no way to dress that up, so this issue does not try. What it does is show you the number that matters more than the record, explain why that number is small comfort, and admit where the pack ran thin on us this week.
The Ledger
A unit is a fixed bet size so results can be compared. One unit is one unit whether that is $10 or $100. Hit rate is the percentage of bets that won. At standard -110 odds you need 52.4% just to break even. CLV is closing line value: whether the line you bet was better than where the market closed. It is the best long-term sign a bettor knows what they are doing, because closing lines are the sharpest numbers there are.
Here is the honest problem with the table this week. The stat pack graded all five plays and delivered the totals, but the list of individual plays came through empty. The house rule is that no number goes in this letter unless it is in the pack. So the per-game rows are not here, and I am not going to reconstruct them from memory and call that accountability.
| Game | Play | Number taken | Closing number | CLV | Result | Units |
|---|---|---|---|---|---|---|
| 5 plays graded | per-play detail not delivered in this week's pack | avg +0.4 | 0-5-0 | -5.0 |
Running totals, from the pack:
| Record | Units | Hit rate | Avg CLV | Break-even hit rate |
|---|---|---|---|---|
| 0-5-0 | -5.0 | 0.0% | +0.4 | 52.4% |
Five plays published, five graded, zero pending. Nothing is hiding in a "still open" column.
What the Closing Line Said
Average closing line value was +0.4 points across the five plays. That means, on average, the numbers we took were four tenths of a point better than where DraftKings closed. The market moved slightly toward our side after we bet, five times out of five on balance.
Why does a fan care about that when the scoreboard says 0-5? Because a single week of results is mostly coin flips landing on the same side. The closing line is the collective opinion of every sharp bettor and every book's risk desk, hammered into a number right before kickoff. If you consistently get a better number than that, you are ahead of the smartest money in the room, and over hundreds of bets the wins show up. If you consistently get a worse number, you can go 5-0 this week and you are still going to lose money by Thanksgiving. Results tell you what happened. CLV tells you whether you should expect it to keep happening.
Now the honest part. Plus 0.4 is a good sign, not a great one. A half point is worth real money when it sits on a key number, and worth much less when it sits on a number like 5.5 versus 6. The pack does not tell me which of these five had the half point on 3 or 7, so I cannot claim the CLV was high quality. I can say it was positive, and I can say positive CLV with an 0-5 record is the classic signature of variance rather than a broken process. I can also say that five plays is nowhere near enough to prove that. Two weeks from now, if the CLV column is still green and the record has evened out, this paragraph was right. If the CLV drifts to zero, it was not.
The Lean, Revisited
There were no wins, so there is no luck to dismiss and no read to take a bow for. Every play gets the loss paragraph. Since the pack did not deliver the individual plays, I will give you the one paragraph that applies to all five, and then the one thing I can back with a number.
On the losses, the question is always the same: was it the read, the number, or variance. The number was not the problem. The closing line moved our way on average, so the market agreed we were on the right side of the line at the time we bet. That leaves the read and variance. The read is where I owe you the hard truth, and the backtest in the pack says it plainly. Walk-forward from 2023 through 2025, playing only when the model disagreed with the closing spread by two or more points, the model went 51.7% against the spread on 437 plays. Break-even is 52.4%. That is a losing rate over a large sample. The 2 to 3 point disagreements hit 54.7%, which is real. The 3 to 4.5 point disagreements hit 52.1%, which is nothing. The 4.5 point and bigger disagreements hit 47.7%, which means when the model screams loudest, it is usually the model that is wrong.
So here is what the number missed this week, in the only form I can honestly state it: if any of the five plays came from a large model disagreement, the read was the failure, not bad luck, because the model's own history says those are the plays to fade. If they came from the 2 to 3 point band with a good number, then this week is variance and the process holds. The pack does not let me sort the five into those buckets. Next Tuesday it has to, or this section is not doing its job.
One more honest note on the model itself. It builds a power rating for each team, which is a number for how good a team is, in points, against an average team on a neutral field. Subtract two ratings, add home field, and you have a model spread. Those ratings are built on 20% this season and 80% last season regressed. Regression means pulling an extreme number back toward average because small samples lie. A 6-0 team is good; it is not undefeated-good. Two weeks of 2026 data is a very small sample, and the pack is right to lean on last year. But it also means the model's picture of a team that changed a lot over the offseason is mostly a picture of last year's team.
Looking Ahead
Straight talk first: the pack's slate is this week's games, all of them final, with the DraftKings numbers already pulled. It does not contain next week's openers. I am not going to guess what DraftKings will hang on Wednesday morning, and I am not going to guess the matchups.
What the pack does have is where DraftKings opened this week's games on Wednesday at 9:00 AM ET, and that is a useful lesson in reading a board for key numbers. Key numbers means NFL games land on 3 and 7 far more than any other margin, so a half point on either side of those numbers is worth real money. Here is how the week opened and where the model sat, so you can see the pattern to hunt for when the new board drops:
| Game (kickoff ET) | DK opener | Model spread | On or next to a key number |
|---|---|---|---|
| CIN @ HOU (Sun 1:00) | HOU -3 | HOU -5.9 | On 3 |
| PHI @ TEN (Sun 1:00) | PHI -7 | PHI -6.3 | On 7 |
| NYG @ LA (Mon 8:15) | LAR -7 | LAR -8.9 | On 7 |
| SEA @ ARI (Sun 4:25) | SEA -3.5 | SEA -5.7 | Half point off 3 |
| CAR @ ATL (Sun 1:00) | CAR -2.5 | ATL -1.9 | Half point off 3 |
| JAX @ DEN (Sun 4:05) | DEN -2.5 | DEN -0.5 | Half point off 3 |
| NO @ BAL (Sun 1:00) | BAL -7.5 | BAL -3.8 | Half point off 7 |
Three games opened dead on a key number and four more opened a half point off one. That is seven of sixteen games where a half point of movement was the whole bet. The two on 7, Philadelphia at Tennessee and the Giants at the Rams, are the kind you want to be early on if you like the favorite and late on if you like the dog, because -7 becoming -7.5 changes the math for the favorite in a way that -5.5 becoming -6 never does.
Also note the two openers the model disagreed with by the most: Carolina at Atlanta, where the model made Atlanta a 1.9 point home favorite against a 2.5 point Carolina opener, and New Orleans at Baltimore, where the model said 3.8 against an opener of 7.5. Both are disagreements north of 3.5 points. Per the backtest, those are exactly the disagreements that hit 47.7% historically. When the Week 4 board opens, the plays to trust are the ones where the model and DraftKings are 2 to 3 points apart and the line sits a half point on the right side of 3 or 7. The plays to fade are the ones where the model is shouting.
Line shopping means checking the same bet at several sportsbooks and taking the best number. Getting +3.5 instead of +3 is free equity, and it costs nothing but a minute. The pack saw no other books this week, so there was no line shop available. That is the single most repeatable edge in this whole letter, and it was off the table.
Ledger Entries
| Game | Play | Number taken | Closing number | CLV | Result | Units |
|---|---|---|---|---|---|---|
| none |
Plain English
- Units: a fixed bet size so results can be compared. One unit is one unit whether that is $10 or $100.
- Hit rate: the percentage of bets that won. At standard -110 odds you need 52.4% just to break even.
- Closing line value (CLV): whether the line you bet was better than where the market closed. The best long-term sign a bettor knows what they are doing, because closing lines are the sharpest numbers there are.
- Spread: the head start the underdog gets. Seahawks -3 means Seattle has to win by more than 3 for that bet to cash.
- Key numbers: NFL games land on 3 and 7 far more than any other margin, so a half point on either side of those numbers is worth real money.
- Power rating: a number for how good a team is, in points, against an average team on a neutral field. Subtract two ratings, add home field, and you have a model spread.
- Regression: pulling an extreme number back toward average because small samples lie. A 6-0 team is good; it is not undefeated-good.
- Line shopping: checking the same bet at several sportsbooks and taking the best number. Getting +3.5 instead of +3 is free equity, and it costs nothing but a minute.
Sources
Lines and openers: DraftKings via ESPN. Team profiles, EPA, and power ratings: nflverse play-by-play. Injuries: official injury reports (none listed this week). Ledger grading and backtest: SharpEdge model, walk-forward 2023-2025 against closing spreads.
21+. Gambling problem? Call 1-800-GAMBLER.