A new way to rate a team's pitching staff

[The Wolf]

Originally Posted by YankeePride

That was my point. It's what confuses me the most because skip will be the first to say that you should use multiple stats instead of one. Yet, in a group of stats he picks on one? Out of the ordinary for him.

And how are you going to go from Point A, every pitcher's previous WHIP, to Point B, a quantification of the team pitching staff? Are you going to MLE the WHIP's of the minor leaguers who are being promoted? Are you going to park-adjust? How are you going to decide whose WHIP gets counted how much?

I'm not seeing any results here.

[Skipaway]

Originally Posted by magnet

The love? It was in a group of 7 different stats. You don't use multiple stats, skip?

Actually, that's the point. When you use multiple stats, WHIP definitely shouldn't be among one of them.


You know, raw data are the more the better, but composite stats aren't. It's counterproductive to use WHIP when you have more advanced ones available, and WHIP isn't even as complete as ERA.

[Skipaway]

Originally Posted by YankeePride

That was my point. It's what confuses me the most because skip will be the first to say that you should use multiple stats instead of one. Yet, in a group of stats he picks on one? Out of the ordinary for him.

If you review the stats he listed, you'd see three basically raw stats in HR/9, K/9 , and BB/9. The others are all composites of multiple raw stats. WHIP is the least useful among those.

If he had listed advanced pitching stats, I would have doubted his inclusion of ERA.


To make the idea easier to understand: when you have both the first version of Run Created and the latest version of Run Created, you'll be counterproductive to consider both. You should just use the latest version. Something like 30%first version+70%latest version would not be as good as 100%latest version.

[ctorg]

Originally Posted by The Wolf

I didn't write "ease of use," I wrote "Anybody got a suggestion for a team pitching staff rating method that gives results as easy to compare from team to team as his does?" You do realize that numbers are valueless in and of themselves unless they are information? Infinitely accurate numbers will do you no good at all unless they are telling you something that you can readily use.

Until someone presents another method of predictive team pitching staff quantification, instead of just sniping, Jon Weisman's method remains the best one I've seen to date.

You think you have a better method? Post it. If you do, I'll use it.

Personally, I think just using the staff ERA from the previous season (or major league equivalent ERA for rookies) would be a more accurate predictor than the method given. Not that I think staff ERA from the previous year is a great measure, but just about anything is better than the arbitrary one used.

[The Wolf]

Originally Posted by ctorg

Personally, I think just using the staff ERA from the previous season (or major league equivalent ERA for rookies) would be a more accurate predictor than the method given. Not that I think staff ERA from the previous year is a great measure, but just about anything is better than the arbitrary one used.

I'm going to run an experiment.

Like most of us, I have a league running. So I'm going to quantify all pitching staffs using both the SI method and staff ERA and see which one ends up as a better predictor of overall pitching staff performance. Should be interesting.

Two things. First, everyone seems to have a visceral reaction to anything subjective. That's very good at one level, but when we have to rank things that can't be counted or calculated, that's bad. Second, we need better predictive tools than the ones we have.

[QuestGAV]

Originally Posted by The Wolf

Until someone presents another method of predictive team pitching staff quantification, instead of just sniping, Jon Weisman's method remains the best one I've seen to date.

You think you have a better method? Post it. If you do, I'll use it.

The problem is that it's not a really a method without some definition of what he means with regard to the categories. Without that, it's just a cheesy way of summing up your own subjective opinions about the individual pitchers.

If he wants to specify that a 1 point pitcher had a VORP less than 10 last year, 2 points less than 15, etc, that'd be fine - but why not just add up their VORPs (or your own favorite stat) to start with and be done with it?

[QuestGAV]

Originally Posted by The Wolf

I'm going to run an experiment.

Like most of us, I have a league running. So I'm going to quantify all pitching staffs using both the SI method and staff ERA and see which one ends up as a better predictor of overall pitching staff performance. Should be interesting.

What does any of this actually mean? How are you going to quantify "pitching performance" if you feel it can't be counted or calculated? How are you going to rate the pitchers with the SI method?

[CMH]

Originally Posted by The Wolf

And how are you going to go from Point A, every pitcher's previous WHIP, to Point B, a quantification of the team pitching staff? Are you going to MLE the WHIP's of the minor leaguers who are being promoted? Are you going to park-adjust? How are you going to decide whose WHIP gets counted how much?

I'm not seeing any results here.

What? I was never even considering using it to rate a pitching staff. I was just stating that it should be considered when evaluating talent.

I see what skip is saying but I still say that you consider everything if it's given to you.

Why not consider WHIP if it's thrown out there? Even if it may not fit your raw data stats, it's a stat that shows something.

[genusthemenace]

It's too arbitrary. Without having an objective way to rate the pitchers, the whole system is just dopey.

[ctorg]

Originally Posted by YankeePride

What? I was never even considering using it to rate a pitching staff. I was just stating that it should be considered when evaluating talent.

I see what skip is saying but I still say that you consider everything if it's given to you.

Why not consider WHIP if it's thrown out there? Even if it may not fit your raw data stats, it's a stat that shows something.

I tend to think of WHIP as a less useful version of Opponents' OBA. WHIP is easier to get from standard stats, and you can get a reasonable estimate of OOBA from it (AB is approximately 2.9 times IP plus hits).

[The Wolf]

Originally Posted by QuestGAV

How are you going to quantify "pitching performance" if you feel it can't be counted or calculated? How are you going to rate the pitchers with the SI method?

Have you been paying attention at all?

I'm trying to figure out the best way to predictively rank a team's pitching staff, or at least a better method than the one in the SI article. So far we've seen one - count 'em, one - other idea, team ERA (I was hoping for more, but no one seem willing to offer any more), so I'm going to use the SI and team ERA methods and run an experiment to see which works better as a predictive model over a few seasons.

What part of that didn't you understand?

[Skipaway]

Originally Posted by The Wolf

Have you been paying attention at all?

I'm trying to figure out the best way to predictively rank a team's pitching staff, or at least a better method than the one in the SI article. So far we've seen one - count 'em, one - other idea, team ERA (I was hoping for more, but no one seem willing to offer any more), so I'm going to use the SI and team ERA methods and run an experiment to see which works better as a predictive model over a few seasons.

What part of that didn't you understand?

You can just go out and find all those prediction models for individual players, and combine them with the playing time prediction to come up with a good prediction. I'm sure those would be much better.

There are a lot of those things out there, like Baseball Prospectus'.

[silvam14]

Originally Posted by The Wolf

Jon Weisman at SI (with, it seems, some help) has created a new way to rate team pitching. The article is at

Tisk, at first glance I thought he worked for Sports Interactive.

[magnet]

Originally Posted by silvam14

Tisk, at first glance I thought he worked for Sports Interactive.

The difference of course is that Sports Interactive would have his press pass for writing such a crap article.

[QuestGAV]

I don't think Weissman is a bad and/or dumb writer, I just think any attempt to stretch his "method" beyond the scope of a pair of throwaway articles is doomed to failure. FWIW, I don't think Weisman is seriously proposing this as a method of analyzing rotations.

[QuestGAV]

Dola,

He sure did get some attention for himself with it. In that respect, I'm sure he's happy with his invention.

[The Wolf]

Originally Posted by Skipaway

There are a lot of those things out there, like Baseball Prospectus'.

Got a link to their method? I get BP via RSS feed and have never seen it.

[ctorg]

Originally Posted by The Wolf

Got a link to their method? I get BP via RSS feed and have never seen it.

Don't they publish player predictions every year? Or did I dream that?

[The Wolf]

Originally Posted by ctorg

Don't they publish player predictions every year? Or did I dream that?

I'm not looking for player predictions. I'm looking for a way to predictively rate team pitching staffs that yields an easy-to-compare result.

[ctorg]

Originally Posted by The Wolf

I'm not looking for player predictions. I'm looking for a way to predictively rate team pitching staffs that yields an easy-to-compare result.

You can just multiply the various players' expected ERAs by their expected number of innings, sum it up, and the team with the lowest has the best staff. That's a nearly random thing that is better than the devised system in the article. Really, the system described in the article isn't really anything. It's assigning numbers that aren't based on anything to players.

My previously stated "previous year's ERA or ML equivalent for entire staff" suggestion beats that, and even that isn't too great. You take the ERAs of the players on the team, weight them by innings pitched (or, to be fancier, expected innings pitched), use equivalences for guys who pitched in the minors, and get the team ERA. This isn't that good, but it's a lot better than the system proposed.