Major League Soccer's Matchweek 20 produced a string of stunning results as several of the league's heavyweights stumbled, while our forecasting model endured its toughest round of the season on match results. Only four of the fifteen match result calls landed correctly, though the total goals forecasts provided a bright spot.
How our forecasts did
The model graded 15 calls across every tracked topic. Total goals was the standout success, with eleven of the fifteen forecasts hitting the mark. Both teams to score landed nine times, and win or draw, a market that often leans heavily on the favourite, returned seven correct calls. The most difficult areas were the fine‑margin calls: first team to score connected six times, the exact scoreline landed just three forecasts, and half‑time and full‑time, always the most volatile topic, found only two correct outcomes.
Match result was the round's great disappointment. A series of away wins and unlikely draws meant the model's preference for the home side or the in‑form visitor rarely matched the scoreboard. The four correct match result calls all came in fixtures where the outcome was strongly weighted by the underlying data.
Results
Philadelphia Union 2‑2 Inter Miami
The model saw value in a Philadelphia win but the match ended all square. The calls for more than 2.5 goals and both teams to score landed comfortably, while the forecast that Inter Miami would score first and lead at the break missed after the Herons were pegged back.
FC Cincinnati 2‑1 New York City FC
This was the model's cleanest performance of the night. Every call bar the half‑time and full‑time forecast landed, including a perfect exact scoreline of 2‑1. A high‑confidence home win, over 2.5 goals and both teams to score all obliged as Cincinnati's explosive attack made the difference.
Columbus Crew 1‑2 CF Montreal
The biggest upset of the round. The model had a high‑confidence Columbus win, with the Crew tipped to score first and lead at the break. Instead Montreal, winless in five and with a terrible away record, stunned Lower.com Field. The over 2.5 goals and both teams to score calls still landed.
New York Red Bulls 0‑1 Nashville SC
A correct match result call at medium confidence, as the Eastern Conference leaders edged a tight contest. The model also correctly predicted Nashville would score first and pitched the exact 1‑0 scoreline, but its forecast of over 2.5 goals missed on a night when the defences held firm.
Orlando City SC 1‑2 Chicago Fire
A rare clean sweep for the forecasts that counted. The model correctly called a Chicago win, with both teams scoring and more than 2.5 goals, and even nailed the exact 1‑2 scoreline and the half‑time and full‑time pattern. Chicago's in‑form attack against Orlando's injury‑hit defence played out exactly as the analysis suggested.
Toronto FC 3‑3 Charlotte
The low‑confidence Charlotte win call was undone by a wild draw. The forecast of at least a draw for Charlotte landed, however, and the medium‑confidence calls for over 2.5 goals and both teams to score came in. The exact scoreline and half‑time and full‑time calls both missed as Toronto found a late equaliser.
DC United 0‑3 New England Revolution
A low‑confidence DC United win was emphatically wrong. The model saw an unbeaten but draw‑prone DC side facing a struggling Revs outfit, but New England ran out comfortable winners. Over 2.5 goals hit, yet the forecasts for both teams to score and DC to score first both failed.
Sporting Kansas City 1‑1 St. Louis City
The model expected a St. Louis win and a decisive result, but the draw exposed the fragility of that call. St. Louis did score first, as forecast with high confidence, yet the match result, over 2.5 goals and the half‑time and full‑time calls all went the wrong way.
Minnesota United FC 1‑2 Atlanta United FC
Minnesota's draw‑heavy home form was the basis for a low‑confidence win call, but Atlanta snatched the points. Over 2.5 goals and both teams to score landed, but the forecasts for Minnesota to score first and a share of the points at the break were off target.
Colorado Rapids 1‑0 Los Angeles FC
A medium‑confidence LAFC win was undone by a resilient Colorado performance. The model saw LAFC's superior record and Colorado's strong home form and expected goals on both sides, but the Rapids kept a clean sheet and nicked the win, making the both teams to score and match result calls incorrect.
Seattle Sounders 1‑2 Austin
Two out‑of‑form sides produced a contest that went against the model's low‑confidence Seattle win call. The forecast of more than 2.5 goals landed, but the call that at least one side would be kept off the scoresheet was wrong, as were the predictions of Seattle scoring first and Austin leading at the break.
Real Salt Lake 3‑4 FC Dallas
A seven‑goal thriller made over 2.5 goals and both teams to score obvious winners, but the low‑confidence Real Salt Lake win call missed as Dallas edged it. The model's half‑time and full‑time call of Dallas ahead at the break and winning was spot on, though the exact scoreline was not.
Los Angeles Galaxy 1‑0 San Jose Earthquakes
The model's medium‑confidence San Jose win call was one of the round's highest‑profile misses. San Jose's excellent away form was expected to overwhelm an injury‑hit Galaxy, but LA's narrow win meant the win or draw, both teams to score, and over 2.5 goals calls all failed.
Vancouver Whitecaps 0‑1 Houston Dynamo
The model backed a medium‑confidence Vancouver win, trusting their imperious home record, but Houston's stout defence and a late goal turned the match. The low‑confidence call for fewer than 2.5 goals proved correct, yet the both teams to score and first team to score forecasts were wrong.
Portland Timbers 3‑1 San Diego
Portland's relentless home scoring was correctly identified, with the match result, win or draw, over 2.5 goals, and both teams to score all landing. The exact scoreline missed by a single goal, and the half‑time and full‑time call was off, but the medium‑confidence home win was well judged.
What we got wrong
The round's signature miss was the high‑confidence Columbus Crew victory forecast. The analysis read a strong home side against a Montreal team winless in five with a dreadful 1‑1‑7 away record. Both lineups were missing key midfielders, and the model expected the Crew's home advantage to overwhelm an open game. Instead Montreal capitalised on Columbus's failures to finish and walked away with a 2‑1 win.
Another notable group of misses centred on the model's read of away form. San Jose arrived at LA Galaxy as strong travellers, but the Galaxy's threadbare squad held firm for a 1‑0 win. The model gave a medium‑confidence call to San Jose, expecting their scoring threat to overcome LA's absentees; the Galaxy's clean sheet and solitary goal flipped every attacking forecast on its head. Similarly, LAFC were tipped to beat Colorado, but the Rapids' strong home stand prevailed, and Vancouver's solid home record was no match for Houston's defensive solidity. In each case, the model leaned on a data‑driven strength that simply did not materialise on the pitch.
The low‑confidence DC United win was a miscalculation of momentum. DC were unbeaten in five, albeit with four draws, while New England were winless in three and owned the second‑worst away record in the Eastern Conference. The model expected a tight DC victory, but the Revs produced a commanding 3‑0 away performance, rendering the match result, win or draw and both teams to score calls incorrect.
What the round changed
The avalanche of away wins and unexpected draws tightened the playoff races in both conferences and underlined the capricious nature of MLS. For the model, Matchweek 20 was a humbling reminder that strong underlying numbers cannot always account for the chaos of a league where form and fitness can flip a fixture in an instant.
