THE BETTING FOOL
By a Man Who Should Know Better
Five and oh. That is what the sides did last week — Tampa Bay, Miami, San Francisco, Jacksonville, Cleveland — and I would love to tell you it was genius, but the arithmetic will not let me. Those five plays were each about a fifty-two percent proposition. A card of five fifty-two-percent plays goes five and oh about one week in twenty-seven. I will take the money. I will not take the credit.
The ledger. The weather unders went one and three: Chicago and the Jets stayed under, and Detroit-Carolina (58 points), New England-Buffalo (55) and Pittsburgh-Cleveland (51) did not come anywhere close. The week was six and three on the bets, up twenty-one dollars and ninety-one cents, and the bankroll is $1,019.34, up 1.9 percent since the start. The season is seventeen and twelve on the bets, and twelve and seven with a push on the five-pick pool card. The weather rule has now lost a week. It won 57.7 percent over ten years, and four games is a sample the size of a sneeze, but I would be lying if I told you the sneeze did not sting.
Two corrections, and they are bad guesses rather than bad numbers. Last week's column said Washington would probably start Marcus Mariota. He did, for ten snaps. He got hurt in the first quarter and Athan Kaliakmanis played the rest of a thirty-to-thirteen loss to Indianapolis. And it twice named Case Keenum as Chicago's quarterback; it was Tyson Bagent who played the whole game, and Keenum who had started the week before. No bet on the card depended on either name, but I built paragraphs around guesses and the guesses were wrong. The correction is posted at the top of last week's page, where it belongs.
The machine got another going-over this week, and it was already right about the one thing a machine can be right about. I re-ran the race — plain regression, Ridge, Lasso and XGBoost — across every game from 2018 through last year, each model allowed to see only games already played. Ridge won again, missing the final margin by 10.24 points on average; Lasso was a hair behind at 10.24, the plain regression missed by 10.35, and the fancy tree-based XGBoost finished last at 10.40. The closing line from Las Vegas missed by 9.83. That is the whole story. The best thing I can build is still about four-tenths of a point a game worse than the market, and the more sophisticated the tool, the worse it did. What the extra work bought me this week is a tuned penalty (chosen by cross-validation across 5,116 team-games since 2017), a model that knows who is at home, and inputs built the way a coach would build them — yards, sacks, turnovers and points that did not come off touchdowns — instead of raw scoring. The practical result is a card full of small disagreements. On only one side this week, Atlanta, do I differ from the line by more than three and a half points.
The football is in the quarterbacks and the injury lists, and I have to be straight about how little I can see. The first injury report of the week does not come out until Wednesday, so everything below prices availability from the Week 3 and Week 4 reports and from what actually happened on the field. Three things matter. In Atlanta, Michael Penix is back, and the Falcons' season numbers still carry two ugly weeks without him. In Baltimore, Lamar Jackson was sacked with twelve seconds left in the first half last week, was hurt on the play, and never came back; Tyler Huntley played the whole second half, and no report I have says what happens next. In Cincinnati, Ja'Marr Chase played fifteen snaps and left. I price all three in points before the model says a word, and I tell you where each one bites. If anything changes by Wednesday night, it goes at the bottom of this page as a dated update. The card does not change.
Eight positions, $122.32, which is 12 percent of the bankroll: two sides that clear their price at fifteen dollars and twenty-nine cents, three that fall just short at five-ten, and three weather unders at twenty-five forty-eight. Nothing on Thursday — my number for Tampa-Dallas is within a point of the market, and I do not bet a coin flip on a short week. No props either. I do not have posted lines to check against, and the last time I priced them myself I went oh and eight. Ratings below are ESPN's Total QBR, zero to a hundred, fifty is average. Pressure, blitz and yards-before-contact numbers come from Pro Football Reference's charting, which my files carry only through Week 3; everything else covers all four weeks. Lines are DraftKings' most recent posted numbers as of Tuesday morning.
Five sides, every week, whether or not any of them is a bet worth making — because a pool makes you pick five. Each one is tagged with what is actually behind it. This week only two clear their price, and by less than a point.
Buffalo plus two and a half in Los Angeles on Monday night, at minus-102, and this is a pick about styles. Start with what the Rams are doing: Matthew Stafford has dropped back 57 times and then 52 times in his last two games, a workload that would have made a coach in 1985 faint. Heavy throwing means heavy risk, and it shows: six interceptions and two lost fumbles against three takeaways, a minus-five turnover margin that is one of the league's worst. I want to be fair about that number, because turnover margin is mostly noise from week to week and the Bills' own minus-two is noise too. What is not noise is how Buffalo runs the ball. The Bills' ball carriers are getting 3.7 yards before a defender touches them, the best in the league, and the offense is one of the five best in the league in points added per play. The Rams are stingy through the air — 154 passing yards a game allowed, the fewest anywhere — which is exactly the kind of defense a patient, run-first game plan travels well against. And Aaron Donald missed last week with a back injury, along with cornerback Jaylen Watson. Buffalo's problem is the other side of the ball: it misses 12 percent of its tackles, the sixth-highest rate in the league, and Los Angeles' ball carriers are getting 3.3 yards before contact themselves. I make it a pick'em, Rams by a third of a point. The line opened at Rams minus three and has come in half a point toward us. Three is the most common margin in the NFL, and a Rams field goal beats this ticket by exactly the hook. At minus-102 it needs 50.5 percent. Fifteen dollars.
San Francisco plus two and a half in Seattle, even money, and the case is the 49ers' offensive line against the one weakness in their own house. Start with what they do better than anyone. Brock Purdy has not been sacked once in four games, and he has been pressured only 12 times through three weeks of charting — the fewest in the league, with Atlanta next at 17. His QBR has averaged 85, and San Francisco leads the league in points added per play, scoring 30.5 a game while giving up 16. Now the weakness. Nick Bosa, Mykel Williams and James Thompson Jr. were all out on last week's report, and a thin defensive line is how you lose a close game: the way to attack it is to run straight at it and make it play every snap. That is where Seattle's problem shows up. Zach Charbonnet and Jadarian Price were both ruled out on the same report, which left Emanuel Wilson and George Holani to carry it at 36 and 22 snaps last week. And at quarterback, Sam Darnold's QBR has been 38 over his last two games. Seattle's defense is the real thing — 18 points a game allowed and 161 passing yards, second-fewest in the league — so I am not pretending this is easy. My number is Seahawks by 1.2, which makes it a 1.3-point disagreement with a line of 2.5, and the cover probability is 50.6 percent against a break-even of 50.0. It clears its price by six-tenths of a point. Fifteen dollars, and do not bet the mortgage.
Tennessee plus six and a half at home against Houston, a lean at minus-107, and the setup is a trick question: both teams are oh and four. Houston has lost by 5, 14, 2 and 4 — outscored by 6.25 points a game. Tennessee has lost by 13, 4, 5 and 6 — outscored by 7.0. That is a three-quarters-of-a-point gap in how the two teams have actually played, and the market has the Texans favored by nearly a touchdown on the road. Now the football. Houston's trouble is in front of C.J. Stroud: sacked 12 times in four games, tied for the fourth-most in the league, and pressured 30 times through three weeks, and his QBR is 55. The Texans' defense is hurting in the middle — safety M.J. Stewart is out, linebacker Azeez Al-Shaair is out, Jadeveon Clowney is questionable — and a team that allows 27 points a game does not have much of a cushion to lay points with. Here is the honest case against my own pick. Tennessee scores 13.75 points a game, the second-fewest in the league, Cam Ward's QBR is 48, and if Houston gets to 23, which is my number, the Titans have to find 17 to cover. Six and a half is a decent number to own: a Houston win by exactly seven, the second most common margin in the sport (8.5 percent of games since 2016), beats us, but a six does not. I have it Texans 23, Titans 19, and the cover probability comes out at 51.4 percent against a break-even of 51.7. A lean, short of its price by a quarter of a point. Five dollars.
Atlanta plus two and a half at home against Baltimore, at minus-112, and this is the biggest disagreement on the card: I make the Falcons favorites by 1.9, the market has Baltimore by 2.5, a 4.4-point gap. Do not take the 4.4 at face value. About three points of it is one man, Michael Penix. Atlanta's season numbers still include two awful games with Cooper Rush and Jack Strand — 13 points, then 3 — and since Penix came back the Falcons have scored 35 at Green Bay and 45 at New Orleans. My model prices his return at three points. The market has already paid most of that bill: Baltimore opened minus-four-and-a-half and is now minus-two-and-a-half. So what is left? Football. Atlanta is built to run — 181 rushing yards a game, second in the league, with eight broken tackles on runs, fourth-most — and it runs against a Ravens line with a hole in the middle: both of Baltimore's centers, Ethan Pocic and Jovaughn Gwyn, were ruled out on last week's report, and the center is the man who makes the protection calls. Atlanta's defense generates pressure (36 through three weeks of charting, seventh-most in the league) but has turned it into only seven sacks in four games, 22nd in the league. Now the part that makes this a lean and not a bet. Lamar Jackson was hurt on a sack at the end of the first half last week and did not return. The only report I have for him is last week's, which listed a back problem he practiced fully through; there is nothing yet for this week. If he is out, the line moves toward Atlanta and the number on this card gets better. If he plays, I have a pick'em where roughly half the edge has already been paid for, and the cover probability is 52.4 percent against a break-even of 52.8. Five dollars.
Chargers plus three and a half at home against Denver, minus-110, and here is the honest version: this is the ugliest pick on the card, and it is on the card because a pool makes me pick five. Los Angeles is oh and four. It scores 16.75 points a game, fourth-fewest in the league. Justin Herbert's QBR is 29, the lowest of any starting quarterback with four games behind him, and his line has allowed 31 pressures through three weeks, the fourth-most. Now the case. Denver is not much better with the ball: 18.5 points a game, Bo Nix's QBR is 47, and the Broncos rush for only 81 yards a game, fourth-fewest in the league. A team like that cannot put anybody away, which is how a three-and-a-half-point underdog survives into the fourth quarter. And look at Denver's defense for the matchup that matters: it blitzes more than all but two teams in the league (78 times through three weeks), it misses 17 percent of its tackles — the worst rate in the league — and it gives up 143 rushing yards a game, fourth-most anywhere. A defense that sends extra rushers and tackles badly is a gift to a team with screens and draws, and the Chargers have broken 11 tackles on running plays, the most in the league. The risks are real: defensive tackle Dalvin Tomlinson is out, Derwin James did not play last week, and Ladd McConkey played just 24 snaps with a foot problem. I have it Broncos 21, Chargers 20, and a cover probability of 51.4 percent against a break-even of 52.4. It is short of its price by a full point, so it gets five dollars and a label.
Under forty-two and a half in Miami, 12.8 miles an hour of wind with gusts to 15 across the kickoff window, in 90-degree heat. The weather rule is the only thing I bet with ten years of proof behind it, and the history is plain: outdoor games with real wind go under 57.7 percent of the time. Here the football and the weather point the same way. Miami has not scored more than 13 points in any game this season — 13, 13, 10, 10 — and Willis' QBR is 40. Cincinnati's defense has 39 pressures through three weeks, third-most in the league, and 14 sacks in four games, tied for third-most, pointed at a Dolphins line that has allowed 11 sacks. Wind takes the deep ball away from a team that has none to give. My model has this at 42.0, half a point under the number, so for once the machine and the rule agree. I will be straight about the case against: Joe Burrow dropped back 56 times last week, and a team that throws that much can score in short pieces even when the wind takes the deep ball. But short pieces take more plays, and more plays in a 13-mile-an-hour wind means more chances for the drive to die. And Ja'Marr Chase played fifteen snaps last week and left; if he is out again, there is less to scare anybody on one side. This is a forecast five days out, not a fact. The price is minus-115, which is steep. Twenty-five dollars.
Under forty-five and a half in Foxborough, 13.9 miles an hour of wind with gusts near 25, and this is the cleanest weather play on the board because of who is playing. The Raiders cannot run the ball. Their ball carriers get 1.5 yards before contact a carry, near the bottom of the league, which means Las Vegas has to move the ball by throwing it — Kirk Cousins dropped back 50 times in a loss to Kansas City last week — and a team that has to throw is the team the wind hurts. New England has its own troubles: 16.25 points a game, third-fewest in the league, ten giveaways in four games (seven interceptions and three lost fumbles), and Drake Maye's QBR is 58. The model has this total at 41.1, more than four points under the number, so the machine agrees with the rule by a wide margin. One more thing I do not love: the total was first posted at 44.5 in the spring and has gone up a point since. That is the market looking at box scores rather than this week's forecast, and I read it as an extra point on our side. The price is minus-115, which needs 53.5 percent. Twenty-five dollars.
Under forty-four and a half at Lambeau, 15.4 miles an hour of wind with gusts near 25, the strongest wind forecast on the whole slate, and the price is the best one on the board: minus-107, which needs only 51.7 percent against a 54.5 percent rule. Here is the football. Chicago runs it — 196 yards a game, the most in the league — and a team that runs into a wind keeps the clock moving and the score down. Green Bay is the team the wind hurts: the Packers rush for 58 yards a game, the fewest in the league, so they have to throw; through three weeks of charting their quarterback has been hit 20 times, the most in the league, and pressured 37 times, second-most; and their receivers have dropped 11 passes, also the most. Put a gusting wind on that and every incompletion is a punt. The Packers score 18.25 points a game, fifth-fewest in the league. The risk is plain too: Chicago hung 59 on Carolina in Week 1 and 27 on Philadelphia in Week 3, and Caleb Williams' hamstring means the quarterback is probably Tyson Bagent, who played the whole game last week (Case Keenum started the week before) — I do not know which. My model has this at 42.4, two points under the number, so the machine and the rule agree. Twenty-five dollars.
“Bet less than you brag, learn more than you lose, and never trust a tout who smiles too much.”
— The Betting Fool
Disclaimer: The Betting Fool is for entertainment purposes only. Nothing on this site is financial advice, and past performance does not guarantee future results. Must be 21+ to wager where legal. If you or someone you know has a gambling problem, call 1-800-GAMBLER.