2017 MLB Gameday Thread

[redsoxsuck05]

2 minutes ago, ramssuperbowl99 said:

I've tried. With all the money put on daily fantasy, I did the math a while back. No impact. Previous HRs are not a significant covariate compared to pitcher handedness, park effects, etc.

Then why are extreme streaks and slumps so commonplace then?  Am I succumbing to patternicity bias?

 

 

[Utley]

Just now, mse326 said:

It's not ignoring. It's considering context.

What is the context? That it was recent and that the data points happened to occur near each other?

[mse326]

1 minute ago, utley4568 said:

What is the context? That it was recent and that the data points happened to occur near each other?

The context is that it is bucking expected randomness trends.

[Utley]

Gallo had a stretch in early July where he had one hit in 19 PAs. So you're saying that should come out of the sample as well?

[ramssuperbowl99]

Just now, redsoxsuck05 said:

Then why are extreme streaks and slumps so commonplace then?  Am I succumbing to patternicity bias?

They aren't common. At any given time, most players aren't on hot or cold streaks. Think about it, on your 25 man roster how many guys are hot and how many are cold? Only a few right?

That said, I don't dismiss the idea of hot or cold streaks entirely. You just run into the same problem that people who try and time the stock market run into - the hot streak is only a hot streak in hindsight. So how quickly from when it starts do you know it's happening, and when is it going to end?

For an entire population, we can't answer those questions.

[redsoxsuck05]

3 minutes ago, ramssuperbowl99 said:

They aren't common. At any given time, most players aren't on hot or cold streaks. Think about it, on your 25 man roster how many guys are hot and how many are cold? Only a few right?

That said, I don't dismiss the idea of hot or cold streaks entirely. You just run into the same problem that people who try and time the stock market run into - the hot streak is only a hot streak in hindsight. So how quickly from when it starts do you know it's happening, and when is it going to end?

For an entire population, we can't answer those questions.

And I suppose there are qualitative elements at play too (at least with individuals), more in some areas than others.  But that's also hard to prove and usually empty narratives to explain random variation.  

Very informative discussion going on, though.

[ramssuperbowl99]

1 minute ago, redsoxsuck05 said:

And I suppose there are qualitative elements at play too (at least with individuals), more in some areas than others.  But that's also hard to prove and usually empty narratives to explain random variation.  

Very informative discussion going on, though.

The qualitative part doesn't help for sure, but I think there's something to be said for guys who make swing changes. The problem is that there are more than 1,000 baseball players all tweaking stuff at any given time, so if you pay attention at a high level you won't know if a player has made a successful adjustment until the results bare it out. If you pay attention to a player so closely that you can predict it, you aren't able to follow a decent percentage of the population.

Same problem that stock brokers have.

It's much more easy to follow pitcher adjustments because you get these 20 or 100 pitch samples of arm slot, velocity, or pitch usage data. With hitters, we just don't have a way to track the day to day workings of the hundreds of MLB players objectively.

[Forge]

Sam Dyson has become a legit reliever again. Good trade by Evans.

[mse326]

12 minutes ago, utley4568 said:

Gallo had a stretch in early July where he had one hit in 19 PAs. So you're saying that should come out of the sample as well?

Again you mistake bad for outlier. He's had multiple stretches like that. It isn't an outlier

[redsoxsuck05]

3 minutes ago, ramssuperbowl99 said:

The qualitative part doesn't help for sure, but I think there's something to be said for guys who make swing changes. The problem is that there are more than 1,000 baseball players all tweaking stuff at any given time, so if you pay attention at a high level you won't know if a player has made a successful adjustment until the results bare it out. If you pay attention to a player so closely that you can predict it, you aren't able to follow a decent percentage of the population.

Same problem that stock brokers have.

It's much more easy to follow pitcher adjustments because you get these 20 or 100 pitch samples of arm slot, velocity, or pitch usage data. With hitters, we just don't have a way to track the day to day workings of the hundreds of MLB players objectively.

So you can reasonably predict the cumulative effect of an entire league's worth of hitting adjustments in high level studies, but at an individual level the variation isn't necessarily random?  I get what you're saying but I'm so confused... 

[NewAge]

8 minutes ago, ramssuperbowl99 said:

The qualitative part doesn't help for sure, but I think there's something to be said for guys who make swing changes. The problem is that there are more than 1,000 baseball players all tweaking stuff at any given time, so if you pay attention at a high level you won't know if a player has made a successful adjustment until the results bare it out. If you pay attention to a player so closely that you can predict it, you aren't able to follow a decent percentage of the population.

Same problem that stock brokers have.

It's much more easy to follow pitcher adjustments because you get these 20 or 100 pitch samples of arm slot, velocity, or pitch usage data. With hitters, we just don't have a way to track the day to day workings of the hundreds of MLB players objectively.

Exhibit A: Daniel Murphy 

A lot of people figured he was on a hot streak in the 2015 playoffs. It turns out he actually was that good. 

[ramssuperbowl99]

Just now, redsoxsuck05 said:

So you can reasonably predict the cumulative effect of an entire league's worth of hitting adjustments in high level studies, but at an individual level the variation isn't necessarily random?  I get what you're saying but I'm so confused... 

The problem is with the observer.

If you're close enough to a player to see a hot streak coming, you aren't focused on the league. So that situation would be like a google employee putting their retirement in only google stock. Even if they know google, that's not a smart decision because they are excluding 99% of the market.

if you're objective and looking at the whole league, you aren't close enough to any one player to see a streak coming before it's already happened. Then, how do you know when the streak is over?

[redsoxsuck05]

2 minutes ago, ramssuperbowl99 said:

The problem is with the observer.

If you're close enough to a player to see a hot streak coming, you aren't focused on the league. So that situation would be like a google employee putting their retirement in only google stock. Even if they know google, that's not a smart decision because they are excluding 99% of the market.

if you're objective and looking at the whole league, you aren't close enough to any one player to see a streak coming before it's already happened. Then, how do you know when the streak is over?

so this is basically like stats 101 with confidence intervals correct?

I was gonna say that over the short-term, you can have changes that are only partially random yet things stabilize in the long run--but I guess that's incorrect.  Like in a gambler's fallacy kinda way.  It implies that randomness and factors in your control will balance each other out.

Ok I have no idea what I'm saying at this point.  

[Utley]

5 minutes ago, mse326 said:

Again you mistake bad for outlier. He's had multiple stretches like that. It isn't an outlier

That's his only stretch like that this season.

[ramssuperbowl99]

Just now, redsoxsuck05 said:

so this is basically like stats 101 with confidence intervals correct?

If it is I missed that class. I think it's equal parts stats/data analysis and psychology. 

The best analogy I can think of is that observers of large samples have bifocals on. If they use the near-sighted lens, they lose the big picture. If they use the far-sighted lens, they can't see until it's too late.