Beyond the Goal Total: Looking beneath the projections of two MLS forwards
As the 2026 MLS season progresses, we wanted to explore what early-season performance can tell us about what comes next for the league’s forwards. Using Relevance-Based Prediction (RBP), we examined whether player performance in the first half of the season could provide useful information about expected goals per 90 minutes (xG/90) in the second half, and how much confidence we should have in each prediction.
For the analysis, we divided the season-to-date into two periods and used player performance from the first half (February 21 through April 23) to predict xG/90 for the second half (April 25 through August 2). The analysis included 64 MLS forwards who met our playing-time thresholds, with each prediction generated out-of-sample using a leave-one-out approach.
Looking beyond the predictions themselves, we focus on two forwards with particularly strong outlooks: Petar Musa and Prince Owusu. Both had high predicted xG/90 and high fit, RBP’s prediction-specific measure of reliability, but the information underlying their predictions tells two very different stories.
FC Dallas forward Petar Musa has established himself as one of MLS’s leading attacking threats this season. In our analysis, Musa also stood out as one of the forwards with the strongest RBP outlooks, combining a high predicted xG/90 with high reliability.
RBP predicted 0.58 xG/90 for Musa during the second half of the season, placing him around the 95th percentile of predicted performance. His fit also ranked around the 95th percentile, indicating that RBP found particularly strong information to support his individual prediction.
Looking beneath Musa’s prediction reveals a relatively consistent story. Goals per 90 was by far the strongest positive contributor to his predicted xG/90, while pass completion and duel win rate also pushed the prediction higher. His high forecast was therefore driven primarily by his scoring production.
Many of those same signals also supported the reliability of the prediction. Goals per 90 was again the strongest positive contributor to fit, while pass completion and duel win rate strengthened both the predicted value and the evidence behind it. Other characteristics, including ball recoveries and shot accuracy, added further support to fit. For Musa, the information driving the prediction and the information supporting its reliability were largely aligned.
CF Montréal forward Prince Owusu provides an interesting contrast. Like Musa, RBP identified Owusu as having both a strong expected outcome and strong information supporting it, but the characteristics underlying that outlook were different.
RBP predicted 0.55 xG/90 for Owusu during the second half of the season, placing him around the 94th percentile of predicted performance. His fit also ranked around the 94th percentile, indicating that RBP found particularly strong patterns to support that prediction.
While Musa and Owusu received similarly strong outlooks, RBP arrived at them using different information. For Owusu, pass completion had the largest positive impact on his predicted xG/90. Age, assists per 90, and shot accuracy also pushed the prediction higher, while his early goals per 90 and xG per 90 pulled it lower.
The information supporting the reliability of that prediction was different. Minutes per game was the strongest positive contributor to fit, followed by age and pass completion. Ball recoveries and shot accuracy also strengthened the information supporting his outlook, while conversion rate, xG/90, and assists per 90 reduced fit.
This is particularly interesting because Owusu’s strong outlook was not simply an extrapolation of his early scoring or chance-generation numbers. Instead, RBP arrived at the prediction through a broader combination of characteristics, some of which affected the predicted value and reliability in different ways.
Owusu’s case highlights an important feature of RBP: the variables that push a prediction higher or lower are not necessarily the same variables that determine how well supported that prediction is.
Conclusion
Both Musa and Owusu have had strong seasons, and both stood out in our analysis with high predicted xG/90 and high fit. But looking beneath those headlines reveals an important distinction: even when two players receive similarly strong projections, the information supporting and influencing those predictions can vary meaningfully.
For Musa, scoring production played a particularly important role, with goals per 90 strongly influencing the predicted value and also strengthening its reliability. For Owusu, RBP relied on a different combination of factors: pass completion had the largest positive impact on the predicted value, while minutes per game, age, pass completion, ball recoveries, and shot accuracy helped strengthen the reliability of the prediction.
This is where RBP can provide additional context beyond the projected number. Rather than assuming the same information matters in the same way for every player, RBP evaluates each prediction individually. It also shows how well supported it is and which aspects of a player’s profile matter for that specific prediction. For scouts, coaches, and front offices, that added context can help turn a projection from a single number into a more transparent piece of information for decision-making.
Comments