João Pedro: one certain fixture, and a date that can slip

A knee problem, and a line of data that rarely says this much: the injury feed carries an end date. João Pedro is declared unavailable since 13 September, with a return on 17 October. That is a declaration, not a diagnosis. But it is not nothing, and it changes the question worth asking.

Because if that date holds, the absence covers a single Premier League fixture.

What the date actually declares

The end-date field is rarely filled in for an injury — it is filled in mostly for suspensions, whose length is known in advance. When it is filled in anyway, you have to look at what it is worth.

As of 25 September, Chelsea declare five unavailabilities. Four of them carry exactly the same end date, 10 October, even though the injuries differ and began on four different dates, the oldest going back to August. 10 October is the date of the club's next fixture. So this is not a medical prognosis: it is a marker that says "not before the next match", and nothing more.

The fifth is João Pedro's, and it is the only one to depart from that: 17 October, one matchday further out. That is where the information sits. The feed does not say the striker will play on the 17th; it says Chelsea are not counting on him for the 10th, and that he is the only injured player at the club in that position.

One certain fixture, two hypothetical ones

Chelsea have already missed one fixture without him. On 25 September we replayed Chelsea's next three outings twice each: once with João Pedro in the eleven, once without him, Danny Welbeck stepping into the same detailed position. Same opponent, same venue, same parameters — everything that moves between the two columns comes from the absence.

Only one line of that table is settled. On 10 October Chelsea host Bournemouth without their striker, and it costs them 1.72 percentage points of win probability on that fixture. The other two lines are not measurements of an absence: they measure what an absence would cost there, if it happened.

What a slip would cost

The second line is the uncomfortable one, because the declared date lands on the fixture itself. Returning "on 17 October" and holding a place in the eleven for Everton – Chelsea on the morning of 17 October are not the same thing. That fixture is a coin flip on the declaration, not on the injury.

If the striker misses that trip too, it costs 1.46 percentage points on that fixture. If the return slips one matchday further and he misses the Tottenham home game on 24 October, it costs 1.79 points on that one.

These three figures do not add up. Each is a gap between two probabilities on one specific fixture; stacking them would produce a total that measures nothing at all. What they do allow is ranking the possible delays by what they cost — and the ranking is not the one you would expect.

Because the fixture where the absence would weigh most is the last of the three, precisely the one the declared date puts out of reach. If 17 October roughly holds, Chelsea will have paid the smaller of their two home costs and dodged the larger.

Why the figure changes from fixture to fixture

The objection writes itself: same player, same deputy, so why 1.46 at Everton and 1.79 against Tottenham?

Because an absence cost is a gap between two probabilities, and a gap compresses as soon as you approach 0 and 1. Chelsea travel to Everton at only 29.30% to win: it is the most lopsided of the three fixtures, and on the wrong side. From a side already badly placed, there is very little left to take away. The two home games, by contrast, sit near the coin flip, where a player moves the most probability.

That leaves the order of magnitude, and it is the figure that puts everything back in place. From one fixture to the next, the cost of the absence varies by 0.33 points. Across those same three games, Chelsea's win probability varies by 17.27 points. The opponent and the venue weigh more than fifty times what the identity of the missing man weighs.

AFC Bournemouth		1.72
Everton		1.46
Tottenham Hotspur		1.79

What would prove these figures wrong

The date first. It is declared, recorded on 25 September, and nothing binds it. We cannot even tell you how often such dates slip: the feed returns only current unavailabilities, never completed ones. So we hold no history of dates declared and then met against which to measure their reliability. That is a gap in what we know, and saying so is better than treating 17 October as settled.

The eleven next. It is reconstructed, not announced. If Chelsea field something else on 10 October, both probability columns move, and the gap with them.

The scope last, three fixtures and not one more. Beyond 24 October, the strength estimate for both sides will have been revised by the results in between: a cost computed today for a December fixture would no longer be measuring the same Chelsea, nor the same opponent.

What to take away

In the table, the club sits 10^th in the Premier League as calculated on 21 September, with 7 points taken, and the simulation sees it finishing in the same place — projected rank 10.0.

A return date is not a recovery, and 1.72 points of win probability on one fixture does not make a season. What is really at stake in October is whether the 17th was a prognosis or a hope.