Ben Clemens explains why Fangraphs rates the O’s playoff odds so low

[Frobby]

Good article by Ben Clemens explaining why Fangraphs rates the Orioles’ playoff chances lower than all the other sites that calculate playoff odds.   The bottom line is here:

“So far this year, the O’s have scored 4.21 runs per game while allowing 4.13. We think their offense will actually get better down the stretch, scoring 4.33 runs per game. The pitching fares far worse, though: we think they’ll allow 4.91 runs per game, which places them squarely below .500 in terms of expected record — .437, to be precise.

* * *

“Mostly, that’s because they’d put up poor performances before this year in the majors. For rookies, their minor league numbers didn’t herald a breakout. If you believe in the projection systems, Orioles pitching has been playing over its head. That’s not to say it can’t continue to do so, but I think systems are right to be skeptical that an entire unit can continue to show a heretofore unseen level of performance.”

Lots of other stuff in this long article, but that’s the gist of it.

https://blogs.fangraphs.com/why-are-the-orioles-playoff-odds-so-low/

 

 

[RZNJ]

Fair enough but what they seem to be saying is that "we didn't predict this and we were wrong about them all year so we aren't about to change now".    I know it's all about statistical probabilities and all but, who knows.

[Can_of_corn]

1 minute ago, RZNJ said:

Fair enough but what they seem to be saying is that "we didn't predict this and we were wrong about them all year so we aren't about to change now".    I know it's all about statistical probabilities and all but, who knows.

I don't get that at all.

They are saying we reran the numbers using the most current data and this is what out data tells us is likely to happen.

And then they state that they might be wrong.

I think anyone having fault in their statement is being overly defensive.

If this was the Tigers you'd probably be nodding along with their explanation.

[baltimoreO's22]

3 minutes ago, Can_of_corn said:

I don't get that at all.

They are saying we reran the numbers using the most current data and this is what out data tells us is likely to happen.

And then they state that they might be wrong.

I think anyone having fault in their statement is being overly defensive.

If this was the Tigers you'd probably be nodding along with their explanation.

They projected an entire pitching staff wrong, not a couple of players. Seems like their model just is wrong on the Orioles. Again though, all those models hated the Orioles from 12-16 when they were the winningest team in baseball 

[NCRaven]

7 minutes ago, baltimoreO's22 said:

They projected an entire pitching staff wrong, not a couple of players. Seems like their model just is wrong on the Orioles. Again though, all those models hated the Orioles from 12-16 when they were the winningest team in baseball 

The model is predicting that the Orioles won't make the playoffs.  You can't say that it's wrong until the Orioles do make the playoffs.  And, to be clear, none of the models are projecting that the Orioles have a 50% or greater chance of making the playoffs, so they're all predicting that the Orioles won't make it, just with varying degrees of certainty.

Edited September 1, 2022 by NCRaven

[DrungoHazewood]

14 minutes ago, RZNJ said:

Fair enough but what they seem to be saying is that "we didn't predict this and we were wrong about them all year so we aren't about to change now".    I know it's all about statistical probabilities and all but, who knows.

Not to be too flip about this, but everyone thinks their favorite team is the one who won't regress to career numbers.  And most people are wrong.

But any projection system can't just ignore past performance or it's going to be wrong more than it's right.  A typical system will do like Marcels, which is something like (3 x this year) + (2 x last year) + (the year prior) all divided by six.  Although they probably tweak that because this year is really only 5/6ths of this year so they'd do well to not rate it quite as highly.

[DrungoHazewood]

11 minutes ago, baltimoreO's22 said:

They projected an entire pitching staff wrong, not a couple of players. Seems like their model just is wrong on the Orioles. Again though, all those models hated the Orioles from 12-16 when they were the winningest team in baseball 

That's misrepresenting what happened.

[Sports Guy]

11 minutes ago, baltimoreO's22 said:

They projected an entire pitching staff wrong, not a couple of players. Seems like their model just is wrong on the Orioles. Again though, all those models hated the Orioles from 12-16 when they were the winningest team in baseball 

As did 100% of the people out there.  

[Tony-OH]

Models can't take into consideration the coaching and the integrations of advanced analytics they are getting vs what other may be not getting. The one constant in all of the Orioles pitchers "playing over their heads" is perhaps the system they belong to. 

I've seen pitchers with new and improved pitches, better use of their repertoires, and he ability to attack hitters in the most conducive way of having success. 

No algorithm is going to be able to access that because they work of using mass amounts of data.

I'll admit, I'm never cared much for any of the prediction algorithms like STEAMER or ZIPS or any other ones because there are too many human factors involved that can only be generalized across players not the individual.  

Now if Fangraphs had to give out betting odds on their low percentage, I bet we'd see a little more realistic odds for a team a few games out. This reeks of them trying to cover their butts for a system that is really unable to accurately access surprise breakout teams like the 2022 Orioles.

[Frobby]

5 minutes ago, Can_of_corn said:

I don't get that at all.

They are saying we reran the numbers using the most current data and this is what out data tells us is likely to happen.

And then they state that they might be wrong.

I think anyone having fault in their statement is being overly defensive.

If this was the Tigers you'd probably be nodding along with their explanation.

I don’t have a problem with their explanation, and I’m happy they gave one.  I do think their formula may weigh performance in past seasons too heavily, once you get to this time of year.  But, at the end of the day, these odds are just for fun.  We’ll see what happens on the field.  

[btdart20]

There has to be a catalyst to change.  To put it another way, there has to be an underlying reason why today's stats are sustainable compared to the player's/unit's history.  To put it in business/investment terms:  a black swan.

Models like FG's programming have a hard time incorporating 'out of the box thinking' and the vast majority of weighting will/should come from history.  Voth introducing a sweeping slider would be a missing variable that COULD cause the models to not reflect reality.  

All that said, models and inputs are getting much more detailed and respond faster to variables.  The quants are truly amazing!  But GIGO still rules the day.

[OriolesMagic83]

15 minutes ago, baltimoreO's22 said:

They projected an entire pitching staff wrong, not a couple of players. Seems like their model just is wrong on the Orioles. Again though, all those models hated the Orioles from 12-16 when they were the winningest team in baseball 

I don't think anyone, including OH members, could have predicted the O's pitching would be this good.  When Means went down in April, it looked like the season was headed for disaster (even if it wasn't different from the disasters of recent years).  I have to give massive credit to Elias for identifying and acquiring guys like Voth, and to Holt, Sig and whoever else contributed, to developing the talent of the pitchers on this team. 

[Tony-OH]

5 minutes ago, DrungoHazewood said:

Not to be too flip about this, but everyone thinks their favorite team is the one who won't regress to career numbers.  And most people are wrong.

But any projection system can't just ignore past performance or it's going to be wrong more than it's right.  A typical system will do like Marcels, which is something like (3 x this year) + (2 x last year) + (the year prior) all divided by six.  Although they probably tweak that because this year is really only 5/6ths of this year so they'd do well to not rate it quite as highly.

Which is why I don't really pay attention to any of these "predictive systems". Nothing that happened last year has any affect on how a player is going to perform over the next month. Pitchers may have different pitchers, hitter may have new information and you know what, somebody just might get hot because they happened to get hot.

The best predictor for how a player will perform over the month is how they've performed this year, especially of the last 80 games or so. But even that has it's flaws. 

[Can_of_corn]

7 minutes ago, Sports Guy said:

As did 100% of the people out there.  

To include folks here.

Would it at all be surprising if Voth does for the O's what Bundy did for the Angels when they traded for him?

[StottyByNature]

For once it appears that the O's are leading the charge in something, as obviously they have figured out a way to increase pitchers' effectiveness.  I don't think they have simply lucked into it given how many pitchers they have been able to do this with.  It also seems to match our draft philosophy - hitters are easier to project and pitchers are easier to fix.  It's nice to feel like an elite organization in this regard.