Win the Division this week.

[Can_of_corn]

4 minutes ago, dystopia said:

Because I wasn’t trying to appease socially inept clowns like you who take everything 100% literal. 

You are clearly demonstrating a mastery of social decorum.

[Sports Guy]

27 minutes ago, dystopia said:

We won 10-3 yesterday. 
 

When your model is as consistently wrong as Fangraphs’ is, then it’s not worth the bandwidth used to create it. 

Yea but that 10-3 win isn’t the norm for the Os.

Do you evidence that their model is consistently wrong?

[dystopia]

9 minutes ago, Sports Guy said:

Yea but that 10-3 win isn’t the norm for the Os.

Do you evidence that their model is consistently wrong?

Where is the evidence that their model is good? 
 

I think most members of this message board could do as good of a job or better at predicting records than Fangraph’s without a fancy model, especially when they refuse to update their model with data accumulated throughout the year. 

[Hank Scorpio]

I too enjoy giggling at people who think that data can predict the future. But the I don't see any reason to believe there is some inherent bias in their algorithm. 

They're guessing based on data. It's not always going to be right, especially in the process's infancy.

[Moshagge3]

We can argue about run differential and luck and true talent all day, but I can't wrap my head around the idea that the Rays have just a 22% probability of earning a bye vs. the O's 88%, but also the Rays have 50% better odds than the Orioles of winning the World Series. The wild card round is essentially a best-of-three coin flip. If you believe the modern thinking that the postseason is basically all random, there's no way the Rays' talent, even if it is superior to the Orioles', should overcome the penalty of having to play another playoff round. Does the Fangraphs model subscribe to the theory that teams with a bye get rusty?

[Frobby]

1 hour ago, dystopia said:

There is no sane reason why Fangraphs should be biased against the Orioles. But the evidence clearly points to that. This isn’t some new thing either, they were like this in the 2012-2016 years too. 

Did they predict the Orioles to win 47 games in 2018?

There’s no anti-Oriole bias.  There’s very little human involvement in their projections.  My understanding of their process is this:

1.  They use a 50/50 combination of two computer projection systems for individual players, ZiPS and Steamer.   Both those systems just use data about past performance of those players, age, how comparable players have done, etc.  to do their projections.  Nobody is telling the computer “this guy plays for the Orioles.”

2.  Once the player projections are done, they need to be scaled to account for who’s on the team.  That requires a modest amount of human judgment to estimate, for example, how the at bats will get divided between the various overlapping players.  How much will Henderson play 3B vs. SS?   How much will Mateo play?   Etc.  

3.  Once that’s done, the individual numbers are added up to come up with team totals, including runs scored and runs allowed.

4.  Team records are forecasted based on the differential between projected runs scored and projected runs allowed, basically applying the Pythagorean formula.  No effort is made to guess which teams will exceed their projected Pythagorean record or undershoot it.

So, the only real human input is in estimating which players will play how much.  But there’s no different process for the Orioles than for any other team.

 

[Moose Milligan]

I'm not defending @dystopiawhen I say I hate the Fangraphs site.  I just want to make it clear.

[Frobby]

4 minutes ago, Moose Milligan said:

I'm not defending @dystopiawhen I say I hate the Fangraphs site.  I just want to make it clear.

I don’t hate the site.  But their projections have the air of science but have little more accuracy than coin flipping when it cones down to it.  

[Moose Milligan]

Just now, Frobby said:

I don’t hate the site.  But their projections have the air of science but have little more accuracy than coin flipping when it cones down to it.  

Their site hurts my eyes to look at it.  I agree about the accuracy of the projections, too, and I also can't stand the writing.

[Tony-OH]

1 minute ago, Frobby said:

I don’t hate the site.  But their projections have the air of science but have little more accuracy than coin flipping when it cones down to it.  

I like their stat pages especially for minor leaguers where they SW strike percentage and other batted ball data that Baseball Reference does not show for minor leaguers.

[dystopia]

42 minutes ago, Frobby said:

Did they predict the Orioles to win 47 games in 2018?

There’s no anti-Oriole bias.  There’s very little human involvement in their projections.  My understanding of their process is this:

1.  They use a 50/50 combination of two computer projection systems for individual players, ZiPS and Steamer.   Both those systems just use data about past performance of those players, age, how comparable players have done, etc.  to do their projections.  Nobody is telling the computer “this guy plays for the Orioles.”

2.  Once the player projections are done, they need to be scaled to account for who’s on the team.  That requires a modest amount of human judgment to estimate, for example, how the at bats will get divided between the various overlapping players.  How much will Henderson play 3B vs. SS?   How much will Mateo play?   Etc.  

3.  Once that’s done, the individual numbers are added up to come up with team totals, including runs scored and runs allowed.

4.  Team records are forecasted based on the differential between projected runs scored and projected runs allowed, basically applying the Pythagorean formula.  No effort is made to guess which teams will exceed their projected Pythagorean record or undershoot it.

So, the only real human input is in estimating which players will play how much.  But there’s no different process for the Orioles than for any other team.

 

I can’t find their 2018 W/L predictions. I can find other years but not that one. If you could link to it that’d be great. 

I assume it was something like 65-70 wins or whatever. Anyway, predicting what ended up being a historically bad team to have merely a more “normal” bad record seems less egregious than predicting a 95/100 win team to be below .500.

Again, Fangraphs model is either biased or incompetent/useless. Take your pick. Sounds like you’re picking the latter.

[deward]

2 hours ago, Moshagge3 said:

We can argue about run differential and luck and true talent all day, but I can't wrap my head around the idea that the Rays have just a 22% probability of earning a bye vs. the O's 88%, but also the Rays have 50% better odds than the Orioles of winning the World Series. The wild card round is essentially a best-of-three coin flip. If you believe the modern thinking that the postseason is basically all random, there's no way the Rays' talent, even if it is superior to the Orioles', should overcome the penalty of having to play another playoff round. Does the Fangraphs model subscribe to the theory that teams with a bye get rusty?

I suspect it's based entirely on the perceived quality of the rotations, and Tampa's advantage in run differential. The RoS projections for Tampa's starters are significantly better than they are for the O's. I'd be surprised if the extra round is really being factored in, as opposed to a perceived advantage in matchups in any individual series. 

[Sports Guy]

I don’t find any of the projection models to be that valuable, whether it’s stats or records...similar to PECOTA.

That said, saying they are biased is completely wrong. 

[Frobby]

1 hour ago, dystopia said:

 

Again, Fangraphs model is either biased or incompetent/useless. Take your pick. Sounds like you’re picking the latter.

Correct.  I don’t think any of these mathematical models give much bang for the buck considering the work that goes into them.  But they’re still fun to kick around.  

[dystopia]

1 minute ago, Frobby said:

Correct.  I don’t think any of these mathematical models give much bang for the buck considering the work that goes into them.  But they’re still fun to kick around.  

Best thing Szymborski could do for himself is to go touch grass.