Home Underdogs Are 14-8-1. Stop Reading Anything Into 22 Games.

Through Week 4, NFL home underdogs are 14-8-1 against the spread. BettingPros counted 23 games, 11 outright wins, and a 64% cover rate on decided bets. Your feed will push the number all week. Do not bet on the number.

This piece does three things. First, you see the 14-8-1 record scored against a coin flip and against the -110 break-even. Second, you see what the research says about the home-underdog edge. Third, you get a Week 6 rule built on price, not on a label.

The record, scored

BettingPros reports home underdogs went 14-8-1 ATS through Week 4 of 2026. The sample excludes the three neutral-site games in Melbourne, Brazil and London. Drop the push and you have 22 decided bets.

A -110 bet needs a 52.38% win rate to break even. The formula is 110 divided by 210. Here is where 14 wins in 22 lands.

Both figures come from the binomial distribution. You recompute them with the sum of C(22,k) times p^k times (1-p)^(22-k) for k from 14 to 22. Neither passes the usual 5% bar. To clear 1%, home dogs needed 17 wins in 22. They finished three short.

The 22 games are a rounding error against the history. Pool the 2026 record with the long-run record below and the combined cover rate moves from 49.1% to 49.4%. The gain is 0.3 points.

Why 22 games fool smart bettors

Start with the spread of outcomes. Twenty-two coin flips have an expected 11 wins and a standard deviation of 2.35 wins. Fourteen wins sits 1.28 standard deviations above the mean. Roughly one 22-game stretch in seven reaches 14 with no edge in the underlying market.

Now count the places to look. The 2026 slate carries dozens of trend splits: home dogs, road dogs, division dogs, dogs of a touchdown or more, dogs after a bye, dogs on short rest. Each split has its own small sample. Check enough splits and one of them runs hot by chance. The home-dog split is the one the headline writers picked, which makes the selection the story, not the result.

The weekly cycle makes the problem worse. A trend looking good in Week 4 draws bets in Week 5. Books adjust, or the sample regresses, and the trend cools. You pay the vig during the cooling. A 64% headline cover rate is a poor reason to pay the vig.

The long-run record is below break-even

Action Network ran the question through Bet Labs. Since 2003, NFL home underdogs went 644-667-38 ATS, a 49.1% cover rate. The 5- and 10-year windows tell the same story. A Las Vegas Review-Journal summary of TeamRankings data listed 211-211-12 over five years and 413-434-25 over ten.

Take the 1,311 decided bets since 2003. A 52.38% rate needs 687 wins. Home dogs won 644. They fell 43 wins short, a 2.4 standard error miss on the wrong side of break-even. A blind bettor lost about 6.3 cents per dollar staked at -110.

The 2007 and 2008 rows carry the lesson. Home dogs went 44-45-1, then 32-45-2. Steven Levitt wrote both seasons up on Freakonomics. He blamed luck, not a market fix. Seasons swing this far in both directions with no change in the underlying edge. The 2026 start is the good version of the same swing.

What the research says

The home-underdog bias is real in the literature. The size and the lifespan are the problem.

• Golec and Tamarkin (Journal of Financial Economics, 1991) tested fifteen years of football spreads. They found two NFL biases. The bias against home teams had nearly vanished, and the bias against underdogs had grown. They added a limit: profit from either bias depends on transaction costs. At -110, the vig is the transaction cost.
• Gray and Gray (Journal of Finance, 1997) used a probit model on NFL spreads from 1976 to 1994. Strategies earned significant profits in sample. Out of sample, some held and others did not.
• Levitt's 2006 Freakonomics post gave the mechanism. Bettors pile onto visiting favorites, so books shade lines toward home dogs and still collect the volume. He entered an 800-person handicapping contest and mostly picked home dogs. Home dogs went 47-30 ATS in the contest season, and Levitt finished near the top.
• Humphreys, Paul and Weinbach (International Journal of Sport Finance, 2013) used a more recent sample. They contradicted the Dare and Dennis explanation for the bias. They also found bettors lean toward the strongest teams and books do not balance their action.

Read these in order and a pattern shows. The edge was visible when the data ended in the early 1990s. The sample after 2003 sits under break-even once you pay the vig. A 2013 inpredictable post carried the verdict in its headline: home underdogs are no longer a good bet.

The mechanism is running in reverse this season

Levitt's edge needs a public who overload visiting favorites. Check what the public did before Week 5. Covers reported underdogs drawing heavy action. At DraftKings, the Giants as underdogs took 75% of bets and 94% of the money.

One game is not a market. The point stands anyway. If the public already leans on dogs, the books do not need to shade home dogs to attract the other side. Do not assume the historical mechanism sits in the 2026 price. You have no evidence of this.

How big a real edge looks

Suppose a home-dog cover rate of 52.5% is the true number. At -110 you profit 0.25 cents per dollar staked: 0.525 times 0.9091 minus 0.475 equals 0.0025. Detecting a 2.5 point gap above 50% at 95% confidence takes about 1,537 decided bets. The formula is (1.96 times 0.5 divided by 0.025) squared.

Count what you have: 22. The 2026 season gives you a few hundred home-dog bets at most, so one season never proves a 2.5 point edge. The edge itself pays a quarter of a cent per dollar. A 22-game sample answers nothing.

The best objection

The strongest case against this piece runs as follows. The market changed in 2026. Books now shade favorites harder, home dogs carry real value, and the early record shows the shift. The case is possible. Three facts weigh against the case.

• The record is 22 decided games. Your test has almost no power to tell a shifted market from a lucky month.
• The shift would need a reason. The only mechanism in the literature is public money on visiting favorites. The pre-Week 5 reporting points the other way.
• Prior bursts faded. Levitt saw a strong contest year and then 44-45-1 and 32-45-2 in the next two seasons. Golec and Tamarkin found one bias nearly gone by 1991.

Treat the shifted-market idea as a hypothesis. Log your own home-dog bets with the opening price, the closing price and the result. If your closing prices beat your entry prices over a few hundred bets, you have evidence. A coin-flip record is not evidence.

What to do in Week 6

1. Give the home-dog label zero weight. Price each game from your own number. If you have no number, pass.
2. Never size a bet off a team-type trend with fewer than 1,500 games behind the claim. The Covers underdog guide warns single-season splits look good on paper because the samples are thin.
3. Shop the half point. A home dog at +3 and a home dog at +2.5 are different bets, and the extra half point costs nothing if you have accounts at several books. See the key number guide for the break-even math.
4. Grade your bets against the closing line, not the final score. Closing line value separates a skill from a hot month. The CLV guide shows the calculation.
5. Add Week 5 and Week 6 to the record before you quote the number. I have no verified Week 5 results at publish time. Six home dogs played Week 5: the Packers, Chargers, Cardinals, Saints, Dolphins and Titans. Whatever they did adds six games to the record and changes none of the above.

The position

Home underdogs are 14-8-1 and a run this good happens about one time in seven with no edge. The long record sits at 49.1%, short of break-even by 43 wins. The research shows a bias shrinking as the data got newer.

Bet a number you believe, at the best price you find. Skip any trend needing 1,500 games to prove when you hold 22.