Switzerland 0–0 Colombia: how the Jarvis supercomputer's prediction held up

The run of play: ninety minutes without the net rippling once

The ninety minutes between Switzerland and Colombia ended 0–0, a result I suspect made more than a few people glance back at the scoreboard just to be sure. At a World Cup, where you turn up expecting goals and moments that lift you out of your seat, these two instead walked one another into an almost perfect stalemate, with neither side prising open a gap wide enough to turn a chance into a goal.

The 0–0 alone tells a story of its own. This was an evening when both defences did their jobs and the forwards fell silent. Not one name made it onto the scoresheet, and the scoreboard did not move once in 90 minutes of normal time. The stalemate held through both periods of extra time, and Switzerland went through after winning the penalty shootout 4–3. Davinson Sánchez and Cucho Hernández missed for Colombia, Manuel Akanji was the only Swiss player who failed to score, and Rubén Vargas struck the decisive kick.

The talking point: the gap between expectation and a blank scoresheet

The biggest talking point, ironically, was the thing that never happened: a goal. Colombia went in as the side the model rated more highly, and yet across the whole match they could not turn that paper advantage into a single breach of the Swiss defence. Switzerland, for their part, dug in and kept a clean sheet, even if the price was that they did not score either.

If I had to keep one image from the night, I would choose the gap between expectation and reality. The Jarvis supercomputer model pictured this as a match worth roughly 2.28 goals in total, an evening with goals at both ends. In the end, both nets stayed spotless. The 0–0 is therefore not just a result, it is a reminder that football always reserves the right to overrule the maths.

Jarvis compared: the pre-match forecast versus the actual result

This is the part I find most worth pulling apart. Before kick-off, the Jarvis supercomputer model weighed three scenarios: Colombia to win at 41%, a draw at 33%, and Switzerland to win at 26%. In other words, the draw was the second most likely outcome, sitting just behind a Colombia victory. And it was precisely that second scenario that came true, at 0–0.

So was Jarvis right or wrong? The honest answer is that it was half right. Right in that the draw sat squarely among the outcomes the model considered likely, with a 33% probability that is far from small. Wrong in that the side rated higher, Colombia, did not win the way most of the weight suggested. The predicted score the model pointed towards was 0–2 to Colombia, and the actual 0–0 matches on only one side of that: Switzerland not scoring. The other half, the Colombia attack, never delivered the 2 the model expected.

One detail I want to underline: the confidence attached to this forecast was labelled low from the outset. That is not an excuse, it is an honest signal that even the machine saw this match as hard to call. The 0–0 is a vivid illustration of that word, low.

Tactical angle: reading the match through the probability frame and the goal markers

I read this match through the probability frame and the goal markers the model left behind. Jarvis expected a total of around 2.28 goals and put the chance of three or more goals at 40%. In reality, neither threshold was reached: the match fell towards the fewest goals possible, zero. The match data shows just how tight it was: Colombia had 15 shots, more than twice as many as Switzerland's 7, but only 3 on target, while Switzerland edged possession at 53.4% and also put just 2 efforts on target.

Read the other way, the 0–2 predicted score shows the model imagined a Switzerland that would struggle to score, and that half proved correct. What the model did not foresee was that Colombia would go just as quiet. Put simply, both attacking edges were neutralised, and once both front lines fall silent, every calculation tilted towards a higher goal count collapses by itself. This is the kind of match where the low-scoring tail of the distribution, which is always there, speaks up for itself.

What the probability means: 41% is not a promise

I want to pause on that 41% for Colombia, because this is where it is easiest to misread. 41% is not a promise that Colombia will win. It means that if this match were replayed a great many times, Colombia would win in around four cases out of ten, while the other six would belong to a draw or to Switzerland. Tonight simply landed in the 33% draw group, a group that was never small to begin with.

In the same way, a high probability in any given match is not a guaranteed ticket either, only a way of saying that the balance tips firmly one way while the chance of a surprise remains fully intact. With the low confidence label already attached here, Jarvis had in effect warned in advance that the gap between the three scenarios was not large. The 0–0 therefore does not break the model, it sits exactly in the zone of uncertainty the model had honestly acknowledged.

Sources and notes

All the forecast figures in this piece, including the three-way probabilities, the predicted score and the expected total goals, come from the Jarvis supercomputer model, recorded before the match so they could be checked against the actual 0–0. There are no goalscorers, simply because neither side scored.

I share these numbers so you can see how a model views a match, and to show that a good forecast can still miss the result without being wrong about its method. This is statistical analysis, not betting advice.