Does travel matter in the NBA? 8 ways of asking, 0 survivors at n=8033

The idea that travel wears NBA teams down is one of the oldest in basketball betting. Road trips grind, legs get heavy, and the fourth quarter of the fourth road game in a row is supposed to look different. We didn't test just one version of that story. We tested the whole family together: 8 separate ways of asking whether travel is mispriced, across a combined sample of n=8033.

The answer is no. Of 8 tests, 0 survived. The weighted ROI across the family is -4.78%.

What we found

Testing a family instead of a single angle is the point. If you run one travel cut and it happens to show a profit, you can't tell whether you found an edge or got lucky with a slice. When all 8 cuts are graded together and none clears the bar, the travel hypothesis fails as a whole.

The individual results all point the same way:

- Fade teams 4+ games into a road trip: -3.5% ROI at n=542, p_raw 0.1971. Not even close to a signal. - Deep road trips go under: -6.7% ROI at n=1001, p_raw 0.0127. This is the most instructive miss. The fatigue story says tired road teams score less. The market already prices that, so betting the under loses. - Home after a 3+ game road trip: -6.9% ROI at n=564, p_raw 0.0425. The "welcome home" bounce is already in the number. - Road underdogs: -4.0% ROI at n=2280, p_raw 0.0237. This is the largest single cut in the family, and it's still negative.

The low raw p-values here aren't hidden edges. They measure how consistently each travel narrative lost against the closing price. A story everyone knows gets priced, and whoever pays for it after that is paying the spread. At n=8033, the travel family is a clean negative.

The trends

A failed family is only useful if it points toward what works. The same framework runs on today's MLB board, where 4 games are on the slate and the reads look very different.

Money is concentrated, not spread out. The top campaign is $139,438 on 200CHCSD SD ML at 96% one-sided. $97,962 on 000BOSNYY NYY ML sits at 92% one-sided, and $49,065 on 400PHIATL PHI ML sits at 93% one-sided. Total tracked whale money is $445,239 across 8 campaigns, and the biggest single ticket on the board is $41,065.

Lines are following the money. Over the last 20h, 200CHCSD-SD moved 52.0c -> 56.5c (+4.5), 000BOSNYY-NYY moved 53.5c -> 55.5c (+2.0), and 400PHIATL-PHI moved 49.5c -> 51.5c (+2.0). The other side of each of those lines moved away: 200CHCSD-CHC 48.0->43.5c (-4.5), 400PHIATL-ATL 50.5c -> 49.0c (-1.5), and 700CWSHOU-HOU 60.5->58.5c (-2.0).

Totals lean over. All 4 of 4 flagged total edges point over. The largest is CHC @ SD, with a model fair of 8.21 against a line of 7.5 (+0.71 runs).

Pitcher form is split. Matthew Boyd (CHC) is trending the wrong way: a 3.7 ERA over his last 5 and 3.16 over his last 3, against a season mark of 3.91. On the other side, Hagen Smith (CWS) carries a 0.0 ERA over his last 5 with 13.5 K/9. All of these are pregame reads, not live in-game prices.

The proven signals are holding out of sample. TRAILING_LATE is at 10.7% ROI over 341 graded bets, and VELO_DROP is at 10.5% ROI over 200 graded bets.

What it means

The contrast is the finding. Travel is a *narrative* signal. Anyone can see a schedule, so the fatigue story is public and already priced in. TRAILING_LATE and VELO_DROP are *measurement* signals. They respond to something that changes inside the game or inside the arm, and the market has to catch up to it. Across 8 travel tests at n=8033, the schedule-based idea found nothing. The measurement-based signals are the ones holding up.

That shapes what we test next. Instead of asking whether travel matters in the abstract, the next run looks for measurable effects of fatigue in the NBA, the way VELO_DROP measures a pitcher's arm directly instead of guessing from his workload. If travel costs a team anything, it should show up in-game before it shows up in the final score. That's the version worth grading.

The travel family is closed for now. The CHC @ SD total gap (fair 8.21 vs 7.5) and the one-sided money on SD are what we're watching today. These are probabilities with a track record behind them, not guarantees. 21+ where legal.

*What would change our mind: a travel cut that clears the bar out of sample, rather than a new slice of the same 8033.*