THE BETTING FOOL
By a Man Who Should Know Better
Three and two on the bets, and twelve dollars and seventeen cents lost, which is a week that tells you almost nothing and costs you almost nothing. If there is a story, it is in the leans. Tennessee, Atlanta and the Chargers were the three plays I told you did not quite clear their price, and all three lost, by 16.5, 4.5 and 2.5 points. The two sides I said did clear it, Buffalo and San Francisco, both won: the Bills took the Rams to overtime in Los Angeles and won it 37–34, and the 49ers won outright in Seattle, 32–29. The weather unders went one and two. Chicago at Green Bay finished with 38 points against a total of 44.5 and went under with room to spare. Cincinnati-Miami (55 points) and Las Vegas-New England (62) went over, and the Patriots hung 45 on the Raiders in a game I had flagged as a wind play.
About that last one. A re-pull of the forecast on Wednesday had the wind in Foxborough falling to 6.5 miles an hour, which would have taken the play off the card. I did not touch the published page, because published picks are never rewritten, and the play lost. I tell you so you know the difference between the card you read and the card I would have made on Wednesday. The ledger: the bankroll is $1,007.17, up seven-tenths of one percent since the start of the season, which is to say nowhere. The season is twenty and fourteen on the bets and fourteen and ten with a push on the five-pick pool card. The leans are four and ten with a push for the year. I will not call that bad luck. A play that does not clear its price is a coin flip with a handicap, I size them at half of one percent for exactly that reason, and this year they have behaved like what they are.
A correction, and it concerns the machine, not the picks. While rebuilding the model this week I found a bug in my own scoring data: points scored on the last play of a game, a field goal as time expires or the touchdown that ends it, were missing from the training set in some games. Every game's score now comes from the final box score, and I re-ran the whole model race from scratch. The order did not change. Ridge regression still finishes first among the models I can build, and the closing line from Las Vegas still beats all of them. The team scoring figures printed in last week's column were taken from the actual scores and were right. It is the model's margin estimates behind the Week 5 card that were slightly off, and a dated note is posted on that page.
Here is what the week of work was really about. I was handed four fresh files, play-by-play, snap counts, injuries and the roster, and asked to find what is not priced into the line. So I asked the hardest version of the question I could: if I take every measure I can build from those files, yards before contact, pressure and sack rates, offensive line availability, quarterback efficiency, and let a model learn how each one moves the final margin, does the leftover disagreement with the closing line predict who covers? I ran it forward through every season from 2018 through last year, each year trained only on the years before. It did not. The side the model liked covered 50.8 percent of the time across 2,175 games, and the correlation between the size of the disagreement and the size of the miss was 0.001, which is zero for any practical purpose. The one thing that survived, after I corrected for how many things I had tried, is the thing already on this site: wind. Outdoor games with 10 miles an hour of wind or more went under 57.7 percent of the time across 603 games from 2016 through 2025.
So, a straight answer to anyone who asks me to get to a hundred percent: the ceiling is not a hundred. The best you can do with public information is stand a hair on the right side of a number that is already very good, and be honest about the size of the hair. The football in the paragraphs below is real, and I think it is useful for understanding a game. It is not a claim that the line is wrong. I use it to choose among close calls, to decide where a quarterback question could swing a number, and to tell you what I would be watching on Sunday. The money rides on the Board.
Four changes in method, all of them disclosed on the How We Bet page. First, a play is a bet only if it clears the price by at least a full percentage point. Clearing by less is a lean, however big the disagreement looks in points. Second, where a quarterback is genuinely in doubt, I price every scenario and weight it. Chicago is a coin flip between Caleb Williams, who was listed questionable with a hamstring on the last report, and a backup. Atlanta is seventy-thirty for Michael Penix over Tua Tagovailoa, because Penix took 35 of the Falcons' 38 dropbacks last week and I do not have this week's report. Tampa Bay is seventy-thirty for Jalon Daniels over Baker Mayfield. Third, cover probability now comes from a symmetric curve fitted to the corrected data: each point of disagreement with the line is worth about half a percentage point, capped at 62 percent. Fourth, a thing I almost got wrong. Houston-Jacksonville is played at Wembley, and in London nobody is at home. My first pass gave Jacksonville its home-field edge and found a lean, the fourth of the five sides. With the edge removed, there is no play on either team, and Carolina took the fifth spot.
Ten positions, $171.24, which is 17 percent of the bankroll: two sides that clear their price at fifteen dollars and eleven cents, three that fall short at five-oh-four, and five weather unders at twenty-five eighteen. Nothing on Thursday, where my number for Seattle at Denver is within a point of the market. No props, because I have no posted lines to check them against. Every ranking below is among the 32 teams through Week 5, from the play-by-play file. Injury notes come from the Week 5 report and this week's roster file, the latest I have; the first report of Week 6 comes out Wednesday, and anything that changes goes at the bottom of this page as a dated update. The weather was pulled Tuesday and will move; I will re-pull it Saturday. Lines are DraftKings' posted numbers on Tuesday, October 13.
Five sides, every week, whether or not any of them is a bet worth making — because a pool makes you pick five. New this week: a side is a bet only if it clears its price by a full percentage point. Two do. The other three are leans, and the card says so on its face.
Atlanta plus three at home against Chicago, at even money, and this is the side where the quarterback question is the whole story, and the bet survives it. My number has the Falcons favored by 1.9, so I differ from Las Vegas by 4.9 points. Much of that gap rides on who plays quarterback for the Bears. Caleb Williams was listed questionable with a hamstring on the last report, and Chicago split Week 5 between Case Keenum (21 dropbacks) and Tyson Bagent (23) in a 31–7 loss at Green Bay. I price Williams at a coin flip. If he plays and Penix does too, my edge shrinks to 2.0 points and the cover probability to 51.1 percent. If he does not, the edge is more than a touchdown. Even the worst case clears even money by 1.1 points, which is the reason this is a bet and not a lean. Now the football. Atlanta is the most run-first offense in the league, throwing on 49.5 percent of its plays, the lowest rate anywhere, and it is good at it: 5.1 yards a carry, second-best, with the fourth-best points added per rush. Chicago gives up 4.2 yards a carry (22nd) and a big play on 7.6 percent of snaps, 28th. A run-heavy team facing a defense that leaks yards on the ground keeps the clock moving and keeps its own quarterback out of trouble. Both teams are minus-five in turnover margin, tied for 28th; turnovers are mostly noise from week to week, so I leave them out of the case. The caution is Chicago's defense overall: 19.2 points allowed a game, fifth-fewest in the league. And Atlanta lost at home to Baltimore last week, 30–23. Three is the most common margin in the NFL, and at plus-three and even money a Chicago field goal returns the stake instead of costing it. Needs 50.0, cover probability 52.7. Fifteen dollars.
Arizona plus twelve and a half at Los Angeles, at minus-105, and nobody has ever enjoyed writing that sentence. Start with how bad it is. The Cardinals have allowed 32.4 points a game, the most in the league, they give up 0.35 points per pass play, 31st, and they were beaten 45–21 by Detroit last week. On offense they have the fewest big plays of any team, 2.4 percent of snaps. Nobody has a good reason to back this team. So here is the arithmetic. Twelve and a half asks the Rams to win by 13, and I make them a 7.5-point favorite, five points less than the market. The Rams are a team that throws on two plays out of three, 66 percent, 7th in the league, and a team that throws that much is not built to bleed the clock when it is ahead. Its last three games were decided by four, four and three points, including a 37–34 overtime loss to Buffalo on Sunday. Arizona's injuries are real (cornerback Will Johnson and lineman Isaiah Adams are on injured reserve, safety Andrew Wingard was out), and my model takes them off the Cardinals before it says a word, about a point between offense and defense, not enough to change the pick. Here is the honest case against. Favorites of 12 to 14 points covered 53.8 percent of 117 regular-season games from 2016 through 2025. That is not different from a coin flip in a sample that size, but it is not on my side either, so it is a reason to keep this at the stake I promised rather than a reason to press. The number also moved a point against us since spring, from 13.5. Needs 51.2, cover probability 52.7, and it clears by 1.5 points. Fifteen dollars.
Kansas City minus eight and a half against the Chargers, minus-105, a lean, and the matchup is one weakness against another. The Chargers have the worst offense in the league by points added per play (minus 0.207, 32nd), they score 16.8 a game (30th), they have given the ball away 12 times, more than anyone, and their line has let the quarterback be sacked on 9.7 percent of his dropbacks, 31st. Their injury list and roster show four linemen out (Rashawn Slater and Trevor Penning on injured reserve, Joe Alt and Kayode Awosika out), plus receiver Quentin Johnston and safety Derwin James. Kansas City's offense is the opposite: 29.5 points a game, 5th, and the league's best running game at 5.7 yards a carry, with the best points added per rush. Here is the other side of the ledger. The Chiefs have the weakest pass rush in the league, sacking the quarterback on 2.5 percent of dropbacks, 32nd, so the one line that could hurt the Chargers is the one Kansas City does not have. I have the Chiefs by 12.1, 3.6 points more than the market, but the market has already moved six points since the spring, from 2.5, and the Chargers have lost five straight by 12, 12, 8, 7 and 6, narrowing as they go. Eight and a half needs the Chiefs to win by nine, and the Chargers' last three losses were by eight, seven and six. The cover probability is 51.9 against 51.2 needed. It clears its price by 0.7 points, which is three-tenths short of the one-point bar, so it is a lean. Five dollars.
Las Vegas plus five and a half at home against Buffalo, minus-108, a lean, and I want to be careful, because this is a pick based on a single week's disaster. Las Vegas lost 45–17 at New England on Sunday. The market has remembered it. Buffalo is the best offense in the league by points added per play (0.211), scores 32.8 points a game, first, and leads the league in big plays, 9.6 percent of snaps. But this Bills team gives up 28.2 points a game (28th) and 0.134 points per play (30th), and it just survived an overtime shootout with the Rams; defensive lineman Ed Oliver and receiver Josh Palmer were both out on the last report. Las Vegas is tied for the league's best turnover margin, plus-five, with ten takeaways, the most anywhere, while Buffalo is minus-four and has given the ball away nine times. I do not put much weight on turnovers, but a team that takes the ball away is a team that keeps a game close. The cost is what I owe you in honesty. The Raiders allow 5.0 yards a carry, the worst in the league, and Buffalo runs for 4.8, fifth-best. If Buffalo gets what it wants on the ground, this is a Bills win by ten. I make it a game Buffalo wins by 1.2, so my number is Las Vegas plus 1.2 against the market's plus 5.5, a difference of 4.3 points. A Buffalo win by exactly six, one of the most common margins in the sport, would beat this ticket by the hook. Needs 51.9, cover probability 52.3. Five dollars.
Carolina plus one and a half in Philadelphia, minus-115, a lean, and I find it more interesting than the number suggests. The line opened with Philadelphia a 6.5-point favorite. Five points of that have disappeared, and the moneyline is now close to a pick'em, Philadelphia at minus-115 and Carolina at minus-105. I think the market has not come far enough: my number is Carolina by 4.0. The reason is that Philadelphia has one of the weakest offenses in the league, and not many people are saying so. The Eagles score 18.4 points a game, 29th. They are 30th in points added per play, 29th in big plays, and they have taken the ball away three times, 29th. They sack the quarterback on 3.1 percent of dropbacks, 30th, which is the number I would look at if I were Bryce Young. The last report listed DeVonta Smith, center Drew Kendall and Saquon Barkley out, and guard Landon Dickerson on injured reserve. Carolina, meanwhile, scores 30.2 points a game, 4th, throws on two plays in three (5th) and has given the ball away just three times, 5th-fewest. It also allows 27.2 a game (26th) and 5.0 yards a carry (30th), the best argument for Philadelphia, though the Eagles' own running game is 27th at 3.4 yards a carry. The price is the issue: minus-115 on a point and a half asks for 53.5 percent, and I make it 53.0. A lean, short of its price by half a point. Five dollars.
Weather unders: an outdoor game with a forecast wind of 10 mph or more. Seven qualify this week and five are on the Board. Forecast pulled Tuesday, October 13; the Saturday re-pull will move some of these.
Under fifty-two and a half at Lambeau Field, minus-105, with 17.5 miles an hour of wind and gusts near 38 forecast across the kickoff window. This is the highest total on the Board and the one the market priced for a shootout, so let me say where the shootout comes from. Dallas has the worst pass defense in the league, 0.40 points per pass play allowed, 32nd, and Green Bay and Dallas are the two most pass-heavy offenses in the league, 68 percent and 69 percent of plays. That is the argument for the over, and it is a good one on a calm day. The wind does not end the passing game. It makes the deep ball a lottery, makes the long field goal a decision, and makes the punt a different game, and it does it to teams that both live on the throw. Green Bay cannot make up the difference on the ground: 3.3 yards a carry (28th) and the second-worst rushing points added per play in the league, 31st. The same stadium produced 38 points last Sunday against a total of 44.5, with Chicago a running team. My own model has this at 50.7, 1.8 under the number. Caesars has 53.5 at minus-113, which is a point better on the number for a worse price; I use DraftKings' number everywhere on this page. The risk is plain: Dallas scored 34 at Houston two weeks ago. Needs 51.2, rule probability 54.5, clears by 3.3 points. Twenty-five dollars.
Under forty and a half in Tampa, minus-115, and I am putting this on the Board because 18.9 miles an hour is the strongest wind forecast on the slate, a notch above Green Bay's 17.5 and nearly double the 10-mile-an-hour line the rule uses. I will tell you what is wrong with it too. The price is the worst on the Board and it clears the rule by exactly one point, the thinnest cushion I allow, and my own model has the game at 41.2, 0.7 over. The total has already fallen five points since the spring, from 45.5, so the market has priced something in; I cannot tell how much of it is the forecast. The football is better than the price. Pittsburgh throws on 67.5 percent of its plays (3rd) but its defense gives up big plays on 8.5 percent of snaps, the worst in the league, which is exactly the thing a gale reduces, because it is the deep pass that stops working. Tampa Bay scores 20.0 a game (25th), its line lets the quarterback be sacked on 9.0 percent of dropbacks (29th), and the Buccaneers' last report had safety Antoine Winfield Jr. and cornerback Benjamin Morrison out, while Pittsburgh was without Michael Pittman (injured reserve) and cornerback Jamel Dean. Rain is a minor factor, a 15 percent chance, and it is 79 degrees. Needs 53.5, rule probability 54.5. Twenty-five dollars.
Under forty-two and a half at Wembley, minus-102, the best price on the Board, with 11.5 miles an hour of wind and gusts near 28 forecast for the 2:30 p.m. London kickoff. Here the football and the weather do not agree, and I would rather tell you than hide it: my model has this game at 45.6, three points above the number. The rule says under; the model says over. I follow the rule, because the rule has 603 games behind it and the model has just spent a week failing to beat a closing line. I think the football supports the rule anyway. Jacksonville's defense has allowed 14.0 points a game, the second-fewest in the league, and it allows 2.8 yards a carry, the best in the league; Houston is second, at 3.3. Houston's own rushing offense, at 3.1 yards a carry, is dead last. A game between the league's best two run defenses, with Houston unable to run, becomes a game of throws in a crosswind, and Jacksonville throws less than almost anyone (55 percent, 29th). The risk: C.J. Stroud's team scored 26 last week, and Jacksonville's passing offense is fourth in points added per pass play. A London forecast five days out is also the least dependable thing on this page, so I trust it a little less than the Tampa one. It needs 50.5, the rule says 54.5, and it clears by 4.0 points. Twenty-five dollars.
Under forty-three and a half in Foxborough, minus-105, with 10.7 miles an hour of wind and gusts of 21.5. The claim here is smaller than it was last week. Two of the five stingiest defenses in the league are on this field: New England has allowed 18.8 points a game (4th) and the Jets 19.2 (5th). The Jets' offense is one of the worst in the league, scoring 18.6 (27th) and letting the quarterback be sacked on 9.2 percent of dropbacks (30th). The Patriots turn the ball over, 11 giveaways, 31st, and gusts of 21 miles an hour do not help a team that is careless with the ball. The model says 43.0, half a point under the number, so the two methods agree for once. The total has also moved up a point, from 42.5, and I read that as the market looking at last week's box score rather than this week's forecast. Needs 51.2, rule probability 54.5, clears by 3.3 points. Twenty-five dollars.
Under forty-five and a half at MetLife, minus-112, with 11.0 miles an hour of wind across the window. This is one of the two weakest forecasts of the five, so the case has to carry itself. New Orleans throws on 67 percent of its plays, fourth-most in the league, but it cannot run the ball (3.2 yards a carry, 30th), it gave the ball away 10 times in five games (29th), and it scored 13 points in a loss to Minnesota on Sunday. The Giants have fewer big plays than anyone but one team, on 3.7 percent of snaps, and last week they tied Washington 27–27. When a team has to build a drive one short throw at a time, wind and a bad running game turn every drive into a coin flip. My model has the game at 44.2, 1.3 under. The risk is New Orleans' defense, which has allowed 30.2 points a game, 31st, and gave up 45 to Atlanta two weeks ago. Needs 52.8, rule probability 54.5, clears by 1.7 points. 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.