Team Match Insight: connecting running data, xG and results

Breiðablik's coaching staff asked a simple but deep question: is there a difference in running data between wins and losses? And if so, does it connect to anything that matters in the game itself, like chances created and expected goals (xG)? The catch is that the answer is not in any single match. One match is noise: a team can run a lot and lose, or run little and win. Team Match Insight gathers every match of the season, reads each match's GPS and IMA movement against the result and Wyscout stats, and surfaces the patterns that only appear once enough matches stack up. This is a descriptive layer — it helps you understand how the team plays, without ever touching the daily decision of who is green, amber or red.

Team Match Insight joins two data worlds that normally live in separate systems: physical load (GPS/IMA from Catapult) and match performance (results + Wyscout stats). It shows four layers: First half vs the norm. Last match's running data compared to the team's normal baseline — was there an unusual drop or spike in the first half? Wins vs losses. Movement metrics grouped by result (W/D/L) with an effect size (Cohen's d) per metric, so you see which factors are genuinely different rather than only slightly apart. Correlation with results. Each movement metric against a result score (win = 1, draw = 0.5, loss = 0), ranked by strength. xG × movement. The team's expected goals (xG for and xG against) against the match's movement — this layer opened up once we added downloaded per-match Wyscout stats, with no API.

Over this season's 20 matches (11 wins, 5 draws, 4 losses) the difference between wins and losses is clear — and it sits in defence. In wins Breiðablik conceded 1.69 xG on average; in losses, 2.84. At the same time the team's own attacking xG did not drop in the losses — it was in fact slightly higher (2.15 in losses vs 1.79 in wins). In other words: the team created enough in the games it lost; what changed was what it gave up. This is a defensive story, and it is exactly the kind of insight neither the scoreline nor running data alone will give you. But the sample is small — only four losses this season — so this is a signal, not a verdict. The screen always flags when a number rests on few matches. The screen also reminds you that results and performance are not the same thing. Breiðablik lost 4:3 to Fram despite creating more in xG (3.01 vs 2.77), and beat KR 6:3 while being out-created on xG (2.83 vs 3.59). One match can lie; a pattern across many matches is far less likely to.

We ran a Pearson correlation of each metric against a result score (win = 1, draw = 0.5, loss = 0) across this season's matches. Here are the strongest relationships. Match stats (20 matches): • Goals conceded — r = −0.60 (p < 0.01) • xG against — r = −0.47 (p < 0.05) • Shots against — r = −0.39 • xG difference (for − against) — r = +0.28 • xG (for), shots, possession, passes — r ≈ 0 GPS / IMA (19 matches with data): • IMA deceleration per minute — r = −0.55 (p < 0.05) • IMA acceleration per minute — r = −0.42 • Player Load per minute — r = −0.23 • Total distance per minute — r = −0.18 • High-speed running, sprints and top speed — r ≈ 0 Two things stand out. First, defence: goals against and xG against are the strongest relationships with results — the team loses when it gives up chances, not when it creates few. Second, braking: more IMA deceleration (and acceleration) per minute tracks with worse results — most likely because a team that is behind brakes and chases more, which is a consequence of losing rather than a cause. What is striking is what does NOT track: high-speed running, sprints, attacking xG and possession have almost no correlation with results. Running faster or holding the ball more says nothing about whether the match was won. Caveat: the sample is small — only three to four losses with data — so these are signals, not verdicts, and correlation is not causation. None of these metrics causes results; they merely coincide.

When movement metrics are read against results, relationships appear that make sense but must be read carefully because of the small sample. In Breiðablik's data, total distance per minute and high-speed running per minute lean toward better results, while some change-of-direction components lean the other way. The system computes Pearson correlations directly from the data and ranks the metrics by strength, but it always flags when the sample is too small to trust the number. Importantly: the correlation with season xG (accumulated per player) is weak — around 0.3 — which is exactly what you would expect. Running more does not make you a better chance-creator; the two things simply coincide sometimes. Team Match Insight shows that weak correlation honestly rather than hiding it.

This is the heart of how the screen is designed. Team Match Insight never says "run more and you will win." It shows association, never causation and never prediction. There are three reasons. First, a third factor can explain both: a team that is behind in a game often runs more chasing the ball, so high running can be a consequence of losing rather than a cause of winning. Second, the sample is small — around twenty matches — so individual games pull hard on the numbers. Third, xG itself is a model with uncertainty, not truth. That is why the screen shows effect sizes and correlations with a clear small-sample caveat, and keeps everything in the descriptive layer. It helps a coach ask better questions — "why did we give up so much in those losses?" — not receive an automatic answer.

Team Match Insight follows the same rule as everything else in MicroPulse: one system, one verdict, visible everywhere. This screen is a context layer — it never touches the readiness colour, the daily decision, or the training recommendation. Green, amber and red still come from the personal norm (how today compares to each player's usual), not from whether the team won its last match. The data also carries its own provenance: movement from Catapult GPS/IMA, results from the fixture list, and xG from Wyscout stats downloaded per match. You can see where every number comes from and how fresh it is. That is the whole point — a coach reads the verdict in seconds, sees the reason underneath, and drills into the numbers and the research whenever they want.

Impellizzeri, F. M., Marcora, S. M., & Coutts, A. J. (2019). Internal and external training load: 15 years on. International Journal of Sports Physiology and Performance, 14(2), 270–273. https://doi.org/10.1123/ijspp.2018-0935 Bradley, P. S., & Ade, J. D. (2018). Are current physical match performance metrics in elite soccer fit for purpose or is the adoption of an integrated approach needed? International Journal of Sports Physiology and Performance, 13(5), 656–664. https://doi.org/10.1123/ijspp.2017-0433 Gabbett, T. J. (2016). The training—injury prevention paradox: should athletes be training smarter and harder? British Journal of Sports Medicine, 50(5), 273–280. https://doi.org/10.1136/bjsports-2015-095788 Rathke, A. (2017). An examination of expected goals and shot efficiency in soccer. Journal of Human Sport and Exercise, 12(2proc), S514–S529. https://doi.org/10.14198/jhse.2017.12.Proc2.05 Note: Breiðablik figures are computed from per-match Wyscout team stats (this season's 20 matches, 2026) and the system's internal GPS/IMA data. References are formatted in APA 7th edition style and were verified against publisher records.