Numbers are just numbers. Sometimes the pitch laughs at them.

Benfica 0-0 Lech Poznań

The market gave Benfica an 87% chance. Implied probability of a draw sat at 10%.

Lech Poznań are not a big name. They sit mid-table in Poland. Benfica had won their last five in all competitions. The 10% figure looked like a polite nod to randomness, nothing more.

Then 90 minutes passed. Then 95. The ball never found the net.

This is the biggest miss of the week by margin. An 87% number suggests near-certainty. Near-certainty does not account for a team defending in two compact banks of four, or for Benfica's 28 shots producing zero goals.

What does 87% feel like when you watch it dissolve in real time?

Portland Timbers 1-1 Houston Dynamo

94%. That was the market's number for a home win. The draw sat at 3%.

Houston had lost four of their last five. They were already eliminated from playoff contention. Portland needed the result. Providence Park at night, November, everything on the line.

Houston scored in the 9th minute. Portland equalised in the 77th. The 3% outcome held.

MLS produces these. The league's parity is structural — salary caps, allocation money, the works. The market still priced this like a European giant hosting a relegated side. The gap between 94% and reality is not a rounding error. It is a category mistake.

Does MLS parity break models built on European data?

Anagennisi Karditsas 1-0 Marko

93% for the away win. 4% for the home win. Greece Super League 2, Sunday afternoon.

Marko came in first place. Karditsas sat 11th of 12. The match was in Karditsas, but that rarely matters at this level — small stadiums, no real home advantage in the data.

Karditsas scored in the 22nd minute. Marko never answered.

4% implies once in twenty-five. Super League 2 has played 72 matches this season. If 4% events happened exactly as predicted, we would have seen two or three home wins like this. We have seen more than that. The market consistently compresses these distributions. It sees table position and assumes certainty.

How many 4% wins does a league need before the model adjusts?

Heerenveen 3-3 Feyenoord

82% for Feyenoord. 16% for the draw. Heerenveen trailed 3-1 after 71 minutes.

The comeback completed in the 89th. Sydney van Hooijdonk — yes, his son — scored the equaliser.

Feyenoord are second in the Eredivisie. Heerenveen are mid-table, defensively porous, nothing special. The 16% draw probability assumed the match would follow the form. It did not account for Heerenveen scoring twice in eight minutes, or for Feyenoord's goalkeeper pushing a routine save into his own net.

This miss is different from the others. The result was not a 0-0 grind. It was chaos, visible from the stands, unpredictable by algorithm.

Can any number capture a goalkeeper's hand turning solid into disaster?

What this actually tells us

The market processes information faster than any fan. It does not process what happens after kickoff.

These four matches share nothing in common — different leagues, different stakes, different styles of failure. The common thread is certainty where certainty was not deserved. A number can tell you what has happened. It cannot tell you what will. The gap between 87% and 0% is not a bug. It is the point.

Did you see it coming?

Drop your calls below — the one you got right, the one you talked yourself out of. We read every reply.