538 does sports not just politics forecasts. They gave the 49ers a 93.8% chance of a Super Bowl win.

[highplainsdem]

Link to tweet

[squirecam]

1. ESPN does that in game

Nt

[scheming daemons]

2. That was when there was 7 minutes left in the game and SF was up by 10.

It's an accurate prediction.


The Chiefs comeback was a 1-in-16 event.

Unlikely events happen sometimes.

Please understand probabilities better. You only have a 1-in-36 chance of rolling snake-eyes with a pair of dice. But it obviously happens.

[Gravitycollapse]

6. Thank you.

[RockRaven]

9. This. In-game probabilities which are calculated based on score and time remaining are

reflections of the universe of prior games -- i.e. what each of their scores were at each time point compared to the ultimate winner. This is not a "gotcha" moment which disproves the model in question.

OP is either misinformed or something worse.

[HowardZinnfandel]

14. Really don't think it was

Funny how a similar thing has happened twice in 4 years.

It's like good teams have an ability to do things most teams can't.

1 in 16 for most teams isn't 1 in 16 for teams good enough to make the Super Bowl.

Between this and IDP/Cohn/538/Wasserman all looking like fools, I'm beginning to wonder if anyone knows what to actually make of numbers and how to understand what a number means and know from looking at it when things are off.

[scheming daemons]

15. The team they were playing against was ALSO good enough to make the Super Bowl

.... so the percentage was accurate. SF hadn't blown a 4th quarter lead all season, let alone a 10 point lead.

[HowardZinnfandel]

19. How many times had they played an offense like KC?

A good offense has a lot better chance of scoring enough points to come from behind than a mediocre offense does.

20 was never going to be enough against KC. They failed to score 23+ once this year when Mahomes played.

https://www.foxsports.com/nfl/story/patriots-win-probability-super-bowl-li-99-percent-atlanta-falcons-choke-stats-tom-brady-020617

I'm a data guy, but I believe that article. Because I know how these things are built.

The problem with expected value frameworks is that not everyone is average (average actually almost never happens). Tom Brady isn't average and Patrick Mahomes isn't average. So their ability to engineer a comeback and put up points even on good defenses is way better than average. That's what greatness is, when you can do things the 99% can't do.

So 1 in 100 and 1 in 16 would be 1 in 1600 chance of those two things both happening. So basically we're saying it wouldn't happen again in the life of their careers (especially considering how infrequently their teams would find themselves significantly trailing anyway, bc they're great players at a key position). I'd take that bet.

[Hortensis]

21. Never to forget, though, that although an "accurate prediction"

according to whatever data are available for that specific time, in the actual world, not the statistical one, the prediction was 100% wrong -- not just when it was finally proven to be but at the time it was calculated as well.

These constantly calculated predictions are understood to be what they are by statisticians, but they're not exactly the views into the future we tend to take them as.

[Tiggeroshii]

3. I like how suddenly 538 is awful and cant be trusted by anyone now that they think somebody else

Will win. It's TRUE though. If they were so good they wouldnt have had Biden winning Iowa at any point in this race.

[Gravitycollapse]

5. 538 is a Rorschach test.

What people say about it says more about them then the actual site.

[Tiggeroshii]

7. Hah

Absolutely right

[Awsi Dooger]

17. Exactly

Right wingers in particular love to make fools of themselves via opinions of 538 and Nate Silver. They go nuts when I describe the process of compiling a percentage likelihood at a given point in time.

[Hortensis]

20. Actually, for a long time their text has made me suspect

bias and dishonesty. In an era when even the PBS Evening News is sounding like Fox Lite and the NYT and AP engage in long, sustained hit jobs on the Democratic Party, even publishing major, blatant lies to elect Trump, it's hardly unthinkable.

[Gravitycollapse]

4. That was a live calculation based on game stats.

And it was a good one. The Chiefs made a highly improbable comeback.

We would all benefit from understanding statistics a little better.

[Hermit-The-Prog]

8. I guessed a higher probability ...

The 49ers screwed up 1 possession and the Chiefs made the most of it. That's how it can go in football -- big mo can jump from one side of the field to the other on 1 possession or 1 play.

[BootinUp]

10. LOL. Thats the thing about predicting the future. nt

[carpetbagger]

11. 538 is a sports model anyway

It started as a sports model, and "let's run 100 simulations" is not a bad way to predict a sports match. The 49ers probably would have won 93 of 100 games against the Chiefs started like that. Politics doesn't have the volumes of data to make those predictions with any degree of precision, and so elections can't be preloaded with data in the same way. We'll see how likable the finalists are now that the field is narrowing, and likability is, these days, the only thing that really counts.

[Gravitycollapse]

13. If you knew anything about 538...

You would know they build models that fit whatever it is they are attempting to predict. It’s quite intricate—often overly complicated. And they will sometimes discuss how their models over/undervalue certain data or metrics.

[greyl]

12. Sooo, they said there's a chance Chiefs would win.

Improbable things happen all the time.

It’s Not Actually a Miracle
Five reasons why absurdly improbable things happen all the time.

By David J. Hand Feb 12, 2014

The newspaper notice said that Frank Hughes, of Hilda Street, Darlington, in the United Kingdom, had died on Aug. 1 at the age of 80, and gave the time and place of the funeral ceremony. Several of his friends and workmates turned up to pay their respects, and they joined the family and other mourners afterward.

So imagine the shock Frank’s friends must have felt when, just an hour later, they saw him walking along the street toward them.

It turned out that the newspaper notice announcing 80-year-old Frank’s funeral wasn’t referring to the retired bus driver and D-Day veteran his friends thought it was. It was actually referring to another Frank Hughes of the same age and from the same town. That Frank Hughes was an ex-merchant-seaman, which explained why the family members weren’t surprised to see people they didn’t recognize at the funeral.

More: https://slate.com/technology/2014/02/the-improbability-principle-rare-events-and-coincidences-happen-all-the-time.html

"That a particular specified event or coincidence will occur is very unlikely. That some astonishing unspecified events will occur is certain. That is why remarkable coincidences are noted in hindsight, not predicted with foresight." —David G. Myers

Improbable things happen all the time.

Creationists (e.g., William Lane Craig) and all manner of non-rationalists like to disparage their opponents or bolster their own arguments by pointing out the lack of probability of something happening. Out of all the possibilities, they say, this one is the one that occurred — how fantastically unlikely and amazingly miraculous! It is simply impossible to believe that it just happened by chance!

But improbable things happen all the time because "improbability" is an illusion based on our preconceptions. Often it has nothing to do with statistical truth. The trouble is that we can't grasp the difference between (a) "This particular improbable pattern of lottery numbers came up on this particular day in this particular lottery" and (b) "Some improbable pattern of lottery numbers came up sometime in the last five years somewhere in the world."

In short: "improbability" does not imply "impossibility".

More: https://rationalwiki.org/wiki/Improbable_things_happen

[emmaverybo]

16. Good reads!

[Awsi Dooger]

18. Here's the most interesting application I have ever seen of in-game wagering

There are a handful of guys who follow the PGA Tour week to week. On premises. These guys are primarily from England. I have spoken to them many times. They enjoy my stories of sports betting in Las Vegas decades ago. They have their smartphone opened to the golf wagering sites. They split up amongst the lead groups. Then their goal is to instantly react to a shot in progress. In other words, they are on site so they see the ball in the air and the likely result of that shot, before it is ever finalized and incorporated to the updated odds. So they'll have a wager on-screen and ready to be placed depending on the shot. I've been standing next to them when they see a ball definitely headed to water or out of bounds. They'll register the bet and get it accepted before the shot finishes. Then they'll gleefully contact each other and describe how it happened and what they got.

They do this on win wagers and head-to-head matchup wagers and all types of things. They were so successful with it a few years ago that some sites changed their options. Of course, in doing that type of thing you need to make an enormous amount of wagers. It is volume and an edge. Neither one is any good without the other. Over the same round of golf they might have 8 wagers on one guy and 10 on his opponent in the same matchup. Doesn't matter. In fact, that is ideal, as long as every one of them is more favorable odds than it should have been at the time.

[empedocles]

22. Quants. They do very well in the stock markets, small margins and huge volumes.

[Polybius]

23. At the start of the 4th quarter maybe, before their comeback

They did not give them a 93% chance at the start of the game.

[DeminPennswoods]

24. Nate Silver is a statistician whose real

passion is baseball and the field of analytics pioneered by Bill James.

[jcgoldie]

25. It was based on the game situation so whats the point of this?

Math is witchcraft if it doesn't point to Joe?

[Iliyah]

26. Now if Mayor Pete wins NH the odds will change . .

Then VP Biden wins in SC with a huge bump from AAs, then the odds will change and so on . ..

Looking in the near future, POC should play into the odds.

Just saying . . .

[Loki Liesmith]

27. That doesn't make them wrong

We don’t know what the correct probability was.

[ecstatic]

28. Weren't the final 6 minutes of the game an aberration?

Can any prediction model factor in the probability of a team making 3 touchdowns in the final 6 minutes of the game?

So, in their most dire, most crucial moment yet, the Kansas City Chiefs doubled down again on being down by double digits. Trailing by 10 points, they scored three touchdowns in the final 6 minutes 13 seconds Sunday to beat the San Francisco 49ers 31-20 in Super Bowl LIV at Hard Rock Stadium.