Fangraphs article on why their projections hate us

[dystopia]

31 minutes ago, Can_of_corn said:

And 2018?  Were they biased in favor of the O's?

Not sure why you think anyone would care that they predicted a team that won 47 games to have won 60 instead. It’s only when they do well that FG starts shitting on them. 

[dystopia]

34 minutes ago, DrungoHazewood said:

I mean, they could just use metadata and/or your IP address to geolocate you and redirect to a page that bumps the team in that area up 4-5 spots in the rankings. Certainly would make them more popular, if also ethically suspect.

 

[Can_of_corn]

25 minutes ago, dystopia said:

Not sure why you think anyone would care that they predicted a team that won 47 games to have won 60 instead. It’s only when they do well that FG starts shitting on them. 

Let me get this straight in my head.

If their projection system overestimates the O's performance no one should care but if it underestimates their performance it shows a bias against the O's and we, as fans, should feel disrespected on behalf of the organization or we aren't real fans.

Yea, that still makes zero sense to me.

[Can_of_corn]

1 hour ago, DrungoHazewood said:

You never know the cryptic motivations of baseball's deep state.

It couldn't be that projections tend to group everyone towards the middle and badly miss outliers at both ends on the curve.

[Can_of_corn]

19 hours ago, Hallas said:

 

I agree that it's not *trying* to poo-poo the O's, or that it doesn't have some sort of ulterior motive, but for whatever reason Fangraphs has had difficulty projecting the Orioles even when they've been good over the past 10-12 years, so its easy to wonder IMO.  The rational response is that their model is having trouble capturing something that the O's are doing well.

 

The bias arguments are pretty laughable since one of the main guys (Szymborski) is a diehard O's fan AFAIK.

I think I may have figured it out.

The Fangraphs team sabotages the O's in their projections because they are funding their website from betting on the O's.

It makes perfect sense.  They lowball the O's  chances and get to make money betting on a team they are secretly rooting for.

It explains everything!

[Frobby]

1 hour ago, dystopia said:

Okay, let’s say they are a in reality a 94 win team. That was their Pythagorean record. The question is still how Arizona has better odds than them, or Minnesota, or Philadelphia. All with worse records and worse Pythagorean records. And none of those teams played in as tough of a division as the O’s. 

I believe that in addition to the Pythagorean record issue, there’s also the issue that our runs scored exceed what we “should have” scored based on our hits, walks, homers, total bases, etc.   That’s because we had an extraordinarily good year hitting with RISP, .837 OPS, best in baseball, as compared to our overall .742 OPS.   The Fangraphs model assumes that such a RISP performance isn’t likely to continue, and thus our ability to produce runs is likely to drop.  The average team hit 21 OPS points higher with RISP than they did overall.  
 

[dystopia]

41 minutes ago, Can_of_corn said:

Let me get this straight in my head.

If their projection system overestimates the O's performance no one should care but if it underestimates their performance it shows a bias against the O's and we, as fans, should feel disrespected on behalf of the organization or we aren't real fans.

Yea, that still makes zero sense to me.

Nobody cares if a projected 60 win team ends up winning 47 instead, no. That’s not hard to understand. 

[Can_of_corn]

5 minutes ago, dystopia said:

Nobody cares if a projected 60 win team ends up winning 47 instead, no. That’s not hard to understand. 

It's incredible hard to understand, because it's the same thing.

A projection system failing to correctly predict something.

It's the same failing.

You can go ahead and act like I bothered to finish the template.

[dystopia]

21 minutes ago, Can_of_corn said:

It's incredible hard to understand, because it's the same thing.

A projection system failing to correctly predict something.

It's the same failing.

You can go ahead and act like I bothered to finish the template.

Winning and losing is literally not the same thing. 

[Can_of_corn]

2 minutes ago, dystopia said:

Winning and losing is literally not the same thing. 

But a failed projection is.

[dystopia]

21 hours ago, Hallas said:

 

I agree that it's not *trying* to poo-poo the O's, or that it doesn't have some sort of ulterior motive, but for whatever reason Fangraphs has had difficulty projecting the Orioles even when they've been good over the past 10-12 years, so its easy to wonder IMO.  The rational response is that their model is having trouble capturing something that the O's are doing well.

 

The bias arguments are pretty laughable since one of the main guys (Szymborski) is a diehard O's fan AFAIK.

Fair post. If nothing else all FG and the Blue Jays debacle has proved is that just because you’re a nerd who can put together fancy equations and type numbers into a computer doesn’t mean you should be anywhere near the sport of baseball. 

[Frobby]

14 minutes ago, dystopia said:

Fair post. If nothing else all FG and the Blue Jays debacle has proved is that just because you’re a nerd who can put together fancy equations and type numbers into a computer doesn’t mean you should be anywhere near the sport of baseball. 

And yet, there are several former Fangraphs analysts employed by teams.  And, the O’s and pretty much every other team have a whole department that crunches numbers and provides data for the team to use to make decisions.  

I think the big issue is that projections can’t provide the degree of certainty we’d like.   If a coin flip is 50/50 and a perfect projection is 100, a baseball projection system is probably something like a 56.   We’d like it to be 80+, but it just isn’t.  But if you were betting coin flips and were right 56% of the time, you’d be ahead of the game. 
 

[Can_of_corn]

23 minutes ago, dystopia said:

Fair post. If nothing else all FG and the Blue Jays debacle has proved is that just because you’re a nerd who can put together fancy equations and type numbers into a computer doesn’t mean you should be anywhere near the sport of baseball. 

Like a biomathematician working for NASA with degrees in mechanical engineering and aeronautical engineering?

[24fps]

23 minutes ago, dystopia said:

Fair post. If nothing else all FG and the Blue Jays debacle has proved is that just because you’re a nerd who can put together fancy equations and type numbers into a computer doesn’t mean you should be anywhere near the sport of baseball. 

There is a whole, albeit small, subset of nerds for whom the sport of baseball is secondary to the statistical data it generates.  They need a place to hang out too, don't they?

[DrungoHazewood]

2 hours ago, Can_of_corn said:

It couldn't be that projections tend to group everyone towards the middle and badly miss outliers at both ends on the curve.

No, no. Because it's not like observed performances always have a much larger spread than predicted because you can't predict random variation, or anything like that.