[ThrowMeDown01]

As the blog post rightly said, they present 538 present the forecast in a way that they are never wrong. I can see the point. I also fond the comparison of NFL vs. Senate forecasts meaningful. This kind of "forecasting" seems to be the current version of crystal balling and astrology. They (models) don't really know anything (no causation, and definitely nothing close to an actual model of the real world like in physics). It's a bit like high-performing fund manager sand CEOs who fail to replicate their successes after they were crowned "person of the year", depending on what exactly they forecast a bit better than that because even without having any models that model actual causation, since the world is pretty stable overall (compared to what the universe could throw at us) even correlations may hold for a while.

As for the "50/50" here, I don't think this is meant as being exact numbers (after all, the whole point of the issue is that those exact numbers don't really tell you anything, if anything still is possible and any outcome can be justified later), but simply as the common usage of a phrase in the vernacular for "we don't know either way".

[yorwba]

> As the blog post rightly said, they present 538 present the forecast in a way that they are never wrong.

The whole point of probabilistic modeling is to replace absolute decisions like "right or wrong" by continuous weights on the possibilities. If you absolutely need a definite decision, you can sample a prediction according to the probability assigned by the model. If the true outcome is x and the model assigned it probability p, then that procedure is going to be wrong (1-p) of the time. You could define that number as the "wrongness" of the probabilistic model, as a continuous analog of the definite case.

The advantage of probabilistic modeling is that you can also ask how wrong the model expects to be and get a meaningful answer. If there are many possible outcomes and none of them very likely, any choice is going to be wrong a lot. But you should expect a good model to have a small difference between its expected and actual wrongness. One might call that value "honesty".

[ThrowMeDown01]

> The whole point of probabilistic modeling

The whole point about the current topic as well as of my post: You missed by about a thousand miles. Please read it again. It's really pointless to argue about a strawman created by you. Your model is useless, that's the point! It makes no real(!) predictions - not usable for anything apart from blowing ever more hot air, and if it doesn't come to pass, you are never wrong because you left the door open by not actually saying anything in the first place.

Do you not understand that the guy/his company did nothing at all? And that giving some arbitrary probability was/is utterly devoid of any meaning (especially if you can't be wrong whatever the actual outcome)? They could have made any prediction at all, what difference would it have made? That is the value of that "work".

However, I realize there's people who like such meaningless drivel. It is a version of appearing to actually do something while not actually doing anything. You make it into the news but you can never be held accountable because whatever happens happens, you just helped create a few more entirely useless headlines (apart from helping with page views and ad impressions of course). It's actually quite ingenious to misuse actually useful tools like statistics.

[scott_s]

Nate Silver analyzes their recent predictions against actual outcomes here: How FiveThirtyEight’s 2018 Midterm Forecasts Did, https://fivethirtyeight.com/features/how-fivethirtyeights-20...

[ThrowMeDown01]

So, and now go back and read what I wrote until you understand it. There is no point to this whole thing (apart from creating clicks for the attention industry). It has no impact (again apart from creating clicks). There is no outcome that hinges on anything. They can claim whatever, whether they are right or wrong matters for absolutely nothing. The creation of attention and clicks is completely independent - it happens before the "forecast" event. Great business where outcomes don't matter.

[dllthomas]

https://en.wikipedia.org/wiki/Scoring_rule#Proper_scoring_ru...

[ThrowMeDown01]

What on earth does that have to do with anything being discussed? Why do you think posting a random link contributes anything?

[dllthomas]

It's a link about how we can assess how good a job someone is doing at making probabilistic predictions. I had (maybe wrongly) assumed the relevance was obvious.