Models and math geeks can be infuriating, and for some fans of yesterday's Lions-49ers slugfest, that was on full display. Making decisions that just don't "feel right" do not sit well with us. We're emotional beings. And we sure had alotta feels yesterday, especially on 4th down calls.
For football, or as we wager on horse racing, modeling helps us learn something by taking the emotion out of it. I think we all know that. But, according to many that's the problem - models can't account for a lot of things outside the numbers.
As for Dan Campbell's aggression, the pure numbers did say he made the right decisions. The win probability gained definitely shows it.
But what about game state, and momentum and everything else we use as a crux to be anti-model?
That's where we can use some of the qualitative (and for anti-model people, in this instance, this probably isn't going to make you feel any better).
As Bill Barnwell notes, should Dan have been less aggressive or more aggressive based on these (mostly) unmodelable factors?
"Absolutely (more aggressive). There's hardly a question. The strength of the Lions is their offense, especially their offensive line and ability to overpower opposing defenses. Their weakness is their defense, particularly their pass defense. They should be more aggressive than the numbers suggest on fourth down because it aligns with the strength of their team."
So, long story short, yes, we don't have to be married to the numbers. We can pivot and adjust, but if it walks like a duck and quacks like a duck it's a duck. At that point, no emotional factors or stubbornness or feelz should cement us to go against it.
Switching over to horse racing, we hear much the same about the "models" - they don't factor in 'stuff' - and that can be true. This gives us an out to hold on to some old angles, or things we think or does work, but the same prescription should apply.
We may notice, say at the Big A, a 0-11 horse taking money in pick 3's and doubles to 3-1. He's on the board at 3-1, as well. We've watched replays, it ain't our first rodeo and we're convinced this horse simply isn't a "winner".
We may be right and it's an auto-pitch, but the modelers are telling us he has a 25% chance. Here, rather than just relying on the 4th down decision bot, we can go deeper.
Is the horse dropping? Did he lose the races simply because he was against better horses? Will he get an easy lead today, where as we all know that can turn perennial losers into winners?
Rather than sticking what makes us feel good and relying on base cases that may or may not be correct, we keep open minded and can pivot. No, we may not bet the horse, but we might decide to sit the race out, or not go all-in on the 9-5 shot, because we convinced ourselves the 3-1 loser has "no shot".
I've changed my horse racing play dramatically over the last six months or so.
Just recently I loved an animal for a sizeable wager. Just loved the horse. When the board opened and I checked will-pays, a horse who has a high ceiling but has been racing terrible was completely hammered. I watched replays, I looked as hard as I could for something I missed, and nothing. Meanwhile the horse I loved was not being bet.
Five years ago I bet the horse I love with both fists. Two years ago I bet my horse with one fist. Six months ago I might sit the race out.
In this case I didn't just sit the race out, I actually bet on the horse I was confused about. He won by 8 and paid $5.
The bottom line is that models work. No not all the time, but they work.
It's why NFL coaching is getting more efficient, and at least partially why we're seeing more close games. It's why every topic under the sun in racing involves "CAW" teams winning too much.
No, they aren't the be all and end all, and knowing and learning when to add data and pivot is important. But they aren't going anywhere, and if we're sticking with emotion or feels, or any other heuristic that's married to old mindsets, in my view we're not doing ourselves any favors.
Have a great Monday everyone.
No comments:
Post a Comment