Scottish Premiership split: more goals, but almost no pooled home edge

A clean seven-season window

The Scottish Premiership split divides the competition after 33 league rounds. The top six and bottom six then play five further matches within their group, producing 198 regular-phase matches and 30 split-phase matches in a standard 12-club season. The official competition rules describe the same 33-plus-five structure.

The comparison selects 2012–13 through 2018–19 because both phases are consistently labelled and complete in the checked record. Later seasons contain interruptions or different group labels and, in several cases, extra or missing post-split records. They are not silently merged into a clean-looking average.

Home, draw and away percentages describe the final result share within each phase. They do not measure points, expected goals or the number of home fixtures assigned to each club.

The phase-by-phase record

The table keeps every season and phase visible. It shows why the pooled result should be treated as a summary of varied seasons rather than a rule that every split is higher-scoring or more away-friendly.

The seven seasons share the same basic shape, but their absolute levels differ. Regular-phase scoring ranges from 2.540 in 2018–19 to 2.763 in 2015–16. Split-phase scoring ranges from 2.500 in 2018–19 to 3.433 in 2015–16. The post-split sample is therefore both smaller and wider, which makes the pooled result useful only with the seasonal rows beside it.

The home, draw and away columns sum to 100% within rounding. That check matters because a phase can look more “balanced” in one category while the movement is actually being absorbed by another. In 2016–17, for example, the split draw share falls to 6.67% while away wins rise to 53.33%; the two movements explain the visible result mix without requiring a change in the home share alone.

The scoring effect

Pooled split-phase scoring was 604 goals in 210 matches, or 2.876 per match. The regular phase produced 3,682 goals in 1,386 matches, or 2.657. The difference is 0.219 goals per match, an increase of about 8.2% relative to the regular phase.

The result is driven by variation. Split scoring was higher in 2013–14, 2014–15 and 2015–16, with 2015–16 reaching 3.433. It was almost unchanged in 2012–13 and 2014–15, then slightly lower in 2017–18 and 2018–19. The pooled rise is real for this selected window, but it is not universal across its seven rows.

Short samples matter. Each split phase contains 30 matches, compared with 198 before the split. Five or six unusual scorelines can move a split rate materially, so a 0.219 difference should be read as a descriptive contrast, not as a stable law of the format.

Three seasons show a clear split-phase scoring lift: 2013–14 rises from 2.692 to 3.100, 2015–16 from 2.763 to 3.433, and 2016–17 from 2.687 to 3.200. The increases are not small rounding differences. In 2015–16, 103 goals arrive in 30 split matches, compared with 547 in 198 regular matches.

Four seasons move the other way or stay close. 2012–13 falls slightly from 2.737 to 2.700, 2014–15 moves from 2.571 to 2.600, 2017–18 slips from 2.606 to 2.600, and 2018–19 falls from 2.540 to 2.500. The pooled increase is therefore a majority pattern rather than a universal phase property.

Home, draw and away balance

The pooled home-win shares are almost identical: 42.42% in the regular phase and 42.38% in the split phase. The home result did not disappear, but it also did not account for the scoring increase. This is the clearest reason to separate goals from balance.

Draws fell from 24.17% to 21.43%, a decrease of 2.74 percentage points. Away wins rose from 33.41% to 36.19%, an increase of 2.78 points. In the three-way result mix, the post-split movement is therefore away from draws and toward away wins, while the home share stays level in aggregate.

The pooled numbers conceal the extremes. In 2015–16, away wins were 50.00% of split matches and home wins 30.00%. In 2016–17, away wins reached 53.33% and draws fell to 6.67%. In contrast, 2017–18 had a split home-win share of 46.67% and away wins of 26.67%. The split can change the balance in either direction from one season to the next.

The 2.78-point rise in away-win share is also a pooled statement across unequal competitive contexts. It does not mean every split gave away teams an advantage. In 2013–14, home wins were 50.00% and away wins 23.33%; in 2015–16 and 2016–17, the away shares were 50.00% and 53.33%. The sign changes from season to season.

Likewise, the near-identical pooled home shares conceal a five-point spread in the split rows: 30.00% in 2015–16 versus 46.67% in both 2017–18 and 2018–19. Home advantage is therefore persistent only as an aggregate balance across the selected window, not as a guaranteed property of each post-split schedule.

Why seasons disagree

The split is played by groups formed after the first 33 rounds, so its opponents and competitive objectives are not a random sample of the regular season. Teams chase a title, European places or survival against a narrower set of opponents. That context can help explain why the post-split result mix should not be interpreted as a neutral league average, even though the aggregate alone cannot test the explanation.

Fixture balance is another limit. Clubs do not necessarily finish with exactly the same number of home and away matches; the official SPFL guidance notes that reversals and 20/18 home-away totals can occur. Our percentages are based on outcomes, not an adjustment for those schedules.

The cleanest evergreen claim is consequently modest: in the seven consistently covered seasons, split matches were more goal-heavy on average and slightly more away-friendly in results, while the home-win share remained almost unchanged. The year-by-year rows show how much uncertainty sits behind that pooled sentence.

The phase is also strategically different from the first 33 rounds. After the split, each group faces a narrower set of opponents with a shared objective. A high-scoring or low-draw split can reflect that concentrated schedule, but the available aggregate does not identify which mechanism dominates. Treating the split as an experimental intervention would go beyond the data.

Coverage after the selected window reinforces that caution. The 2019–20 record stops before a complete split, 2020–21 through 2023–24 use mixed group labels or non-standard row counts, and 2024–25 is not labelled in the same way as the selected seasons. Excluding those rows reduces the time span but improves comparability. The result is a seven-season comparison with an explicit boundary, not a forced 15-season average.