Why I Track Every Forecast

One thing I've noticed while looking around at football prediction websites is that most of them are very happy to show you the wins.

Which makes sense from a marketing perspective.

From the beginning I wanted to do it differently. Given the domain, I think that's vital.

The output is for informational purposes only...but let's not pretend people won't sometimes act on it.

I do! 🙌

The problem with cherry-picked output is that you never get the full picture.

What happened to everything else?

Did those forecasts lose?

Were they deleted?

Were they just quietly forgotten about?

You can never know.

So when I started building StatStrike, GoalLab and PopGoals I made a decision very early on.

Track everything.

Not just the forecasts users see in the app.

Everything. Including the forecasts that didn't pass the barriers to upload.

The Archive Isn't Just A History Book

The archive was never just somewhere to store results.

It's central to making decisions.

It is the golden egg.

Okay...occasionally it's a mouldy egg. 😂

But mostly it's shiny.

As months of forecasts accumulated I stopped manually deciding which leagues should appear after manually studying the archive.

I took the mental system, coded it in and let the archive do it for me in every daily forecast cycle.

If a league consistently performed well, it stayed in the upload.

If performance dropped, it lost qualification.

Sometimes a league recovers and earns its place back.

That's exactly how a system built around performance should behave.

The Premier League doesn't get special treatment.

Neither does Queensland NPL.

The archive doesn't care.

It just looks at the results.

Every Forecast Matters

I also learned quite quickly that losing forecasts are every bit as useful as winning ones.

A cluster of losses might reveal that a league has changed.

Or that a particular scoring environment isn't behaving the way it used to.

If I'd simply deleted those forecasts, I'd never have seen it.

Why I Think Transparency Matters

Football is unpredictable.

Every forecasting model will get things wrong.

Mine certainly does.

I'm much more interested in building something where users can see the complete picture than something that only ever celebrates the wins.

That's why every forecast is tracked. And every uploaded forecast is tracked in the app.

Not because it makes the app look better.

Because over time it actually makes the forecasting system better.

And perhaps more importantly...

It lets users decide for themselves whether it's any good. It shows users a losing streak which is deviant from what the system should be achieving...and for the canny user that can land you an accumulator!

It also lets users make their own decisions.

Perhaps a qualified league is going through an unusually poor run.

Perhaps you think that's meaningful.

Perhaps you think it's simply variance.

The important thing is that you can see it.

I think that's also much more useful than trying to convince people the system is faultless.

And I don't think pretending otherwise is either honest or credible.