- Umpires are human too (really…) – why their preferences must be taken into account
- The extra level of detail that Total Brownlow possesses
Welcome to Part 4 of Brett’s Brownlow Breakdown. In the lead-up to the Brownlow Medal 2020, Brett – the man behind our new Total Brownlow package – is sharing his key thoughts on betting on the Brownlow Medal 2020.
Looking at Brownlow betting models, I would say the x-factor in our solution is simple:
I don’t mean on a global scale, like you hear some people speak about players and umpire biases overall.
For example, “we’ll re-rate Marcus Bontempelli for a factor of “X” or “Y” amount, because generally he’s been awarded more votes than we expected in previous games or seasons”.
That approach is okay if you’re not too fussed about precise accuracy, but it doesn’t have the granularity on umpires that I wanted to be confident in our model and our prices.
Detailed umpire modelling
What we’ve done is tried to model the individual preference of each and every AFL umpire.
So we now know, for each individual umpire, what their preferences are based on how they’ve awarded Brownlow votes in the past.
That’s combined for the crew of three field umpires that officiate on each game during the AFL season.
We don’t know exactly what they’ve done in 2020, of course, but we know what they’ve done in prior seasons and hence we can model what their biases are when it comes to awarding Brownlow votes.
This changed the whole ballgame for us, because when you think about the whole Brownlow voting process, it becomes obvious how important it is.
Votes aren’t awarded objectively. A player might play a great game… but how was it from the umpires’ view, and do those umpires rate or like him?
We now have an idea of when the planets align for that to occur – a player playing well, and the right umpiring crew – and we can factor that into the probability of players receiving votes.
Brownlow Medal 2020: The impact of umpires
It’s definitely impacted our final probabilities. There’s players that have shortened considerably, in terms of the probability of them getting votes in a certain game. And, of course, there’s players that have blown out. The impact on some players’ probabilities is subtle, but it’s more substantial for others.
The connection of biases between every player and every umpire has ramped up the complexity of things markedly… but we’re looking for an edge here, and we’re confident this can help with that. It was difficult work, but we’re confident it was worth it.
We think our solution has nailed umpires better than anybody else ever has, and that’s our “X factor” when it comes to this model.