off-by-one
Longform thoughts of @_off_by_one

Bets as a thinking tool

About 2 min reading time

Frequently, when someone asks me for a prediction, I'm not totally sure.

The questions I'm talking about sound like this:

Will doing w make x better?

Is z going to be better next year?

Are we confident we've actually fixed ?

Shrug emoji

The answer is often the same: I don't know. It's easy for me to come up with reasons for why z is going to get better, and reasons why z is going to stay the same. I can also come up with reasons why z is going to get worse. I can get so overwhelmed by my brain's ability to come up with explanations, that I can't figure out what I think the future holds.

When I do know what my gut is saying, I have a hard time expressing my degree of confidence. I hedge. After all, I'm really not sure! This can sometimes be confusing for coworkers, especially those who are used to speaking with and hearing confident statements.

I've learned a technique to figure out what I think is likely to happen, and express my degree of confidence all at once.

Want to make this interesting?

Let's bet on it

No, I don't think you should make casual gambling part of your workplace. But, as a thinking and communication tool, the concept of a "bet" is useful.

Do I think the most recent released caused the production bug.... I can see both sides of it. But--you know what--I do. I'm leaning in that direction. I'd bet $20 that it was the cause. Would I bet $200? Are you crazy? No way.

There, I just figured out what I think, and how much I think it.

There are also times, where it shows me that I don't think something is true/will happen.

Do I think we'll finish the bugs we marked for this slightly overloaded sprint? Well, we could. I really believe it's possible, but I wouldn't bet on it. I wouldn't bet even $1 that we will. Now, that's not the same as saying that I think the opposite is true. I also wouldn't bet even $1 that we won't finish the bugs. I just don't know.

Risk built in

Bets contain a little more information than bayesian priors alone. They force you to imagine what it could be like to be wrong, especially if you imagine you only have $100 in your pocket to bet with. Sure I think it's true? But... how much are we willing to lose? We're on a tight risk budget, after all.

This may steer the ship in a conservative direction, which isn't always good. Loss aversion is powerful and not at all rational. But, it's better than being guided by only boolean predictions.


I don't think I came up with this, but I'm not sure where I picked it up--could have been a book or old coworker. If you find this useful, let me know. Have another way to conceptualize and communicate uncertainty? Let me know that too!


👉   Like this post? Follow on twitter and tell me what you think!