Did the tasks get better as the years went on?

They got more chaotic, and it's the strongest trend in this data. The early series wrote pure skill tests half the time; the modern era writes them a third of the time. Series 21 is the most chaotic season ever assembled – half its tasks were levellers, and twenty-one actively rewarded being bad at Taskmaster.

Does that show up on the scoreboard?

Yes. Chaotic task diets keep title races open, and the chaos peak of Series 20–21 produced a three-way tiebreak and a four-point margin. The modern era's famously close finishes are, in part, written.

Did Covid hurt the tasks?

No. The pandemic seasons (10–11) are statistically indistinguishable from the eras around them. Whatever lockdown did to the production, it left no fingerprint on the writing.

Did they run out of ideas?

Doesn't survive contact with Series 12 – the sharpest skill-testing season in show history, written nine years in, and promptly won by the model's biggest upset. One number for the road: a third of all tasks ever set (304 of 897) have negative discrimination. Being good at Taskmaster actively hurt.

Is this measuring Alex or Greg?

Alex. Both lines describe how the tasks are built, before anyone judges them. Greg's half of the equation – the scoring – is measured separately in the State of Greg.

So are modern tasks worse?

No – different. A leveller is a competitive nothing and often the best television of the episode. This data measures the exam, not the comedy. how?