On Models

[EDIT 2025.03.23 - I originally had a terrifically bad estimate of the average blades of grass in a Michigan lawn, giving the incorrect product of two numbers, but have since fixed it.]

[Wine & Longevity, Lawn Care Charged Per Blade of Grass, 1969 Science Still Got Us To The Moon, The Model in the Arena]

2025.03.23

XCII

To help us understand the world, we make “models” of it.

These are simple representations that miss some details but are still useful.

We can’t really ever hope for a “perfect” model, but it doesn’t matter, as long as the model helps us get the results we want.

[Wine & Longevity]

Models for the world help us understand cause and effect while controlling for other variables.

A common model of the world might include the detail that if I take care of myself, then I will live longer. 

I don’t think anyone would disagree on that, but they certainly would disagree on what we mean by take care of ourselves!

Some people might say that one glass of wine every other night is taking care of themselves, while others might say that drinking any alcohol at all will bring you closer to the grave. 

So, two more specific models (at odds with one another) might be:

1. Completely abstaining from alcohol will make you live longer

2. Drinking one serving of wine 3-4 times a week will make you live longer.

Both of these models encapsulate simple cause and effect. Neither says anything about the underlying biological & metabolic processes that occur when you drink, the chemical composition of wine, or the psychological effects.

In reality, both sentences would come with “because” clauses that explained the model more in depth. They’d have some of the why about the human body’s interaction with alcohol, but certainly not all of it.

And that’s okay.

[Lawn Care Charged Per Blade of Grass]

You can have a model that is too complex to be useful. 

As a business, you have a pricing model that tells you & your customer how much they owe you. A lot of the time, this can be very simple–maybe a lawn care company charges for lawn mowing services based on the size of your lawn.

That makes sense–if the lawn is a little bigger, it might take them a little longer to do it. 

Now, imagine a lawn care company that charged customers based on the number of blades of grass cut. 

To charge like that, the lawn care company would have to get a count of the blades of grass. (If they simply trusted their customers, they’d surely get ripped off). That’s not easy–in an average MI lawn, my market sizing says there’s almost 63M blades of grass!*

Since it would cost more to count the grass then the mower would likely ever make cutting the lawn, they might instead estimate the blades of grass based on the square feet of the lawn.

But, at that point, they might as well just charge on square feet, because now the number of blades of grass is just a function of the square feet.

Even though the blades of grass model is more “accurate”, it’s kind of silly to care about. It adds complexity to the model without adding any real value when it comes time to execute. 

*3000 blades square foot x 21,000 square foot avg yard size

[1969 Science Still Got Us To The Moon]

Deriving a perfect model of the Universe is too much of an ask.

On one end, it is very likely impossible to simulate the universe without a computer that can contain as much information as the universe.*

On the other end, there was a comprehensive attempt to derive even just all of mathematical logic from a few atomic principles, the Principia Mathematica. Mr. Gödel came along and proved such a task impossible.

It doesn’t actually matter that you’ll never have a perfect model of the world, though. You can still achieve a lot with the models we do have, or the models you might build yourself. 

Some examples:

• In 1969, our scientific model of the world was missing many details**. Still, the science we had was sufficient to get us to the Moon.

• Jim Simons’ Medallion Fund no doubt fell short of a “perfect” model of financial markets; still, it was able to return an average annual rate of 63.3% over 30 years.

• Humans discovered fire without knowing anything about chemistry

As you can see, you don’t need to know everything to achieve some cool stuff. 

*David Deutsch discusses this in The Fabric of Reality. As a brief point, even if we unlock something like quantum computing, the quantum computers would then have to simulate quantum computing.

**Bose-Einstein Condensate, Higgs Boson, detection of gravitational waves

[The Model in the Arena]

To be useful, models really just need to get you the results that you’re after.

If you are convinced you can predict the future, and you keep telling people about it, but don’t seem to be accruing the benefits of being a future teller (wealth, respect, wisdom), then your model for predicting the future will rightly be called into question. 

A trick many people pull is to confuse others with the complexity of their model so that they forget about what matters, which is really just the results that the model is supposed to get.

At the end of the day, though, no matter how complex the model is, the real test of it’s strength is nothing more than whether or not it does the thing it’s supposed to do. 

Jack & I are convinced that our model of the world that we are selling via BirdDog is accurate enough to get results. Nobody really cares about the model itself, though. The flash of it is enough to get attention and book us some meetings, but if a user or team does not get results with it, eventually they will leave.

Any model will never be perfect and have 100% fidelity to the world, but it doesn’t matter, because that’s not actually the point of the model. 

You know your model is good if you or someone else uses it and it does the thing you said it would. You know your model is bad if it doesn’t do the thing you said it would. 

Live Deeply,