Two clubs play each other for the sixth time this season and every one of the first five went over the number. Somebody at the bar tells you these two always score. Somebody on the timeline calls it a trend. And a lot of money moves on the idea that whatever has been happening between these two specific teams is going to keep happening.
It is a testable claim, and 2026 has enough completed baseball in it to test it properly. The answer is that a season series streak carries essentially nothing, and that the longest streaks, the ones that feel the most like a pattern, are the ones most likely to break.
2,153 completed regular season games through September 6, 2026. Those games average 8.96 combined runs and 43.24 percent of them finish under eight. Every figure below is measured against those two numbers.
Of those games, 1,741 had at least one prior meeting between the same two clubs earlier in the season. That is the working sample, and each of those games has a history attached to it that a bettor could actually have looked at before first pitch.
Start with the naive version, because it is the one that makes the trend look real. Correlate the mean total of a pair's earlier meetings against the total of their next meeting and you get r equals 0.0537. Positive. Small, but positive, and if you stop there you can tell yourself the series knows something.
The problem is that the correlation is not measuring the series. It is measuring the clubs. Two offenses that score a lot will produce high totals against each other and also against everybody else, so any pair of high scoring teams generates a high series history and a high next game, and the correlation picks that up without any series specific information being involved at all.
So take the clubs out. For every game, subtract what the two clubs' own season scoring already predicts, and correlate what is left of the series history against what is left of the next game. That residual correlation is r equals 0.0168.
A 10,000 shuffle permutation test on that residual returns p equals 0.4828. Not marginal. Not close. Roughly half of random shuffles produce a correlation at least that large, which is the statistical way of saying there is nothing there.
| Measure | Value |
|---|---|
| Games with prior meetings in the same series | 1,741 |
| Raw correlation, prior series mean total vs next total | 0.0537 |
| Residual correlation, after removing both clubs season scoring | 0.0168 |
| Permutation p value on the residual, 10,000 shuffles | 0.4828 |
Correlations are abstract, so here is the version you can bet. Define a high scoring meeting as nine combined runs or more, count how many of those a pair has strung together, and then look only at what happened in the very next meeting.
| Consecutive meetings at 9+ runs | Next meetings | Avg total | Under 8 |
|---|---|---|---|
| 1 | 535 | 8.82 | 43.36% |
| 2 | 212 | 9.48 | 38.21% |
| 3 | 63 | 9.13 | 41.27% |
| 4 or more | 40 | 8.57 | 47.50% |
| League baseline | 2,153 | 8.96 | 43.24% |
Read the bottom row of that table twice. After four or more consecutive high scoring meetings, the next game in the series has averaged 8.57 runs and gone under eight 47.50 percent of the time. That is not a continuation of the streak. It is better than the league baseline for the under, on the games where the over trend looks most obvious.
The two game bucket is the only one that leans the other way, at 38.21 percent under across 212 games, and one bucket out of four pointing somewhere on a sample that size is what noise looks like when you cut it four ways.
If streaks carried, they would carry in both directions. Run the identical test on low scoring series, meetings at seven combined runs or fewer.
| Consecutive meetings at 7 or fewer | Next meetings | Avg total | Under 8 |
|---|---|---|---|
| 1 | 482 | 8.59 | 49.17% |
| 2 | 191 | 8.71 | 45.03% |
| 3 or more | 71 | 8.90 | 38.03% |
| League baseline | 2,153 | 8.96 | 43.24% |
Same shape, mirrored. One quiet meeting is followed by a slightly under leaning game, and by the time a pair has strung three or more low scoring meetings together the next one goes under only 38.03 percent of the time, well below baseline. Both tails decay toward the middle and then overshoot past it.
That is mean reversion, and it is the opposite of the thing the streak is supposed to be telling you.
Three things, none of them exotic.
A series meeting is three or four games out of 162, and a rotation turns over every five days. The pitchers who produced the streak are frequently not the pitchers in the next game. That alone dissolves most of what a series history could carry.
Ballparks change between meetings. A four game set in Minneapolis and a three game set in Detroit are not the same run environment, and the series record blends them into one number that describes neither.
And the extremes are selected. A pair reaches four straight high scoring games partly through talent and largely through luck, and the luck component does not repeat by construction. That is the entire mechanism behind the 47.50 percent figure.
Twelve pairs of clubs have put together five or more consecutive meetings at nine runs or more in 2026. The longest are Colorado and San Francisco at seven, and Atlanta and Miami at seven. Houston and Texas, Detroit and Minnesota, Cincinnati and Colorado, and Baltimore and Toronto have all reached six.
Read that list and the confound announces itself. Colorado appears twice. Both of the sevens involve at least one club whose home park is a run environment on its own. These are not mysterious series, they are mostly parks and offenses, and both of those are already in the price.
One. A season series streak is not information. If your handicap of a game changes because the last four meetings went over, you are being moved by something that measures at p equals 0.4828.
Two. The longer the streak, the less it should move you, which is exactly backwards from how streaks feel. Four or more consecutive high scoring meetings has been a mildly pro under spot, not a pro over one.
Three. Handicap the two clubs, the two starters and the park, and let the series history be what it actually is, a record of games that already happened to different versions of these teams.
Four. Where the streak does have value is in other people. If a number has been pushed up because a series has been loud, the correction is available on the under, and it is available for free.