Run Differential in a Short Season

[PuffyPig]

For those that speculate that there is little correlation between having a positive run differential and making the playoffs, especially in a 60 game season, there are currently 8 teams in the NL with a positive run differential, and they are currently the 8 teams in a playoff spot.

FWIW, it is generally better to score more accumulative runs than your opponents, and it bodes well for making the playoffs.


Sent from my iPad using Tapatalk

[757690]

I am one that has brought it up.

My point was that you couldn’t look at a team with a good run differential and a losing record and make the same assumption that their record will eventually catch up to their run differential like you could in a 162 game season. I think that is still a solid point. The smaller the sample size, the more volatility.

I think everyone agrees it’s better to score more runs that give up.

[RedsManRick]

So I was looking into this further after the conversation in the other thread.

I was making the assertion that if the question is, "How good is this team?" (e.g. "How many wins should we expect this team to win moving forward"), the answer to that question is generally better informed by run differential (pythag record) than by current actual W/L record. That's generally true. But...

I came across this Fangraphs article from 2015 that looked at 10 years of data to see what best predicts rest of season win percentage as of June 7th in that season. Why is that interesting? Well, just so happens that June 7th is about the time when teams have played ~55-60 games.

So firstly, just read the article: https://blogs.fangraphs.com/the-mean...far-adding-on/

But here's the big takeaway:

Approach		R^2
Early Actual Record	0.13
Early Pythag Record	0.15
Preseason Projection	0.27

R^2 it's a widely use metric for predictiveness that essentially measures how much of the differences among teams actual final records can be explained by the listed approach. In short, the table gives us three big takeaways. Through 55ish games:

• Pythag record is a barely better predictor of rest-of-season record than actual record
• Preseason projected record is a much better predictor of rest-of-season record than either actual or pythag record
• None of them are exceedingly amazing predictors of rest-of-season record (e.g. "It's hard to make predictions, especially about the future.")

Now, if you did that same study, but you changed it to after 10 games predicting a 60 game season, but with a schedule comprised of fewer games against the same teams, I think you'd find the same structure of findings, but with R^2 for the actual and pythag approaching useless while pre-season stayed about the same. Ultimately, rosters change, players improve/decline, luck happens, etc. There's a lot we can't know in advance, so predicting the future is hard.

So while I was "right" that run differential is better than actual wins, I missed the bigger picture that both suck as mid-season predictors of the rest of the season, let alone as very early season predictors of a short, quirky season. There's just WAY WAY WAY too little data and too much noise and that premise is likely to be true throughout the entirety of this season. We should collectively stop trying to read too much in to what has already happened as a signal that changes the way we think about how the rest of the season will play out. The games that happened happened and their outcomes matter a lot in terms of whether we have a successful season. But mostly because they count toward that outcome directly, not because they change what we think about how good the team really is.

I think the more useful framing for lamenting the first 11 games would be this: We're currently 5-6. But the important question If we need to win 30 games to make the playoffs, we need to go 25-24 (83 win pace) from today forward. What are the chances of that happening? Not too bad, I'd say.

But instead of doing all that yourself, I would recommend people look at projected standings & playoff odds at sites like BP Fangraphs:
https://www.fangraphs.com/depthchart...tion=Standings
https://www.fangraphs.com/standings/playoff-odds

They not only do the above logic of considering the games played as real and banked and then factoring in good projections for the rest of the games, but they base their rest-of-season projections on updated versions of pre-season projections that consider the current roster and the actual schedule. So while I'm sure people will want to continue to whine about micro level issues about Bell or the bullpen or runners left on base, if we're going to take the macro view, let's anchor it to good thinking about what the future looks like:

According to Fangraphs, this 5-6 team with a +3 run different projects to 31-29 with a 15% chance of winning the division and a 64% chance of making the playoffs.

[DocRed]

So much for this thread...