THE BOOK--Playing The Percentages In Baseball

I filled out the Yankees ballot. I took the approximate midpoint between the IP ranges for the pitchers (i.e. 30-59 became 45) and added them up, and it came out to a little over 1600 innings, which is about 200 innings too high (I think). Most of the 200 was probably in the bullpen, though. The way you break down the ranges makes it difficult to pick one for a reliever.

For example, I figured all of the following guys would throw somewhere between 50 and 70 innings this year: Jose Veras, Mariano Rivera, Phil Coke, Edwar Ramirez, Brian Bruney, Dave Robertson Damaso Marte. So almost every member of the bullpen. I put the first 5 for 60-99, but I don’t think any of them will throw more than 75 innings. It might even out once more people vote, but this seems like somewhat of a problem to me. A better range, IMO, of “regular reliever” would be 50-80 IP. Only 10 relievers threw 80 IP or more in 2008, and none threw more than 90.

AWESOME! BRILLIANT!

This is the sort of thing on which I bet the wisdom of the crowds beats the very best experts. I’m betting it, and I’m so committed to it that I’m willing to volunteer my services at the end of the season to run correlations and see what the RMSE is for the crowd and compare it with Baseball Prospectus’s forecasts. The only constraint is that you have to remind me to do it. Then I will. There’s a 99% chance that I have the April 5th play time projections from BP sitting in my fantasy baseball folder when this season starts, so having the necessary data from them shouldn’t be a problem.

Quick question: what happens if I fail to rank a player? Say for example I know a lot about the Braves rotation but I don’t know a think about their offense. Can I just leave the offense blank and have it remain unweighted in the final accounting?

HAH!

I got so excited about this project that I didn’t even read the whole post. Turns out your already planing a project to compete with “the big boys” which I assume will include Baseball Prospectus (the biggest boy in the projections play ground.)

Anyway, if for some reason calculating the RMSE for the two projections isn’t in your plans, my offer still stands. (And I’d volunteer to do other work that can be accomplished with a simple algorithm or excel sheet.)

Tango,

Are you going to post these on all the SB Nation sites? I could help you out if need be.

Since I’m the first to vote for the Rangers, I should point out that you missed a player:  Elvis Andrus.  You might have heard of him, he’s been in all the papers.

It also appears you split Chris Davis off as the DH, when he will be the 1st baseman and Hank Blalock will be DH.

Finally, what should be a strong pointer to how the Rangers’ 2009 season will go, I could only manage to find three guys who might pitch more than 100 innings.  And two of those (Millwood and Padilla) will be gone next year.  But wait until 2011.  Or 12.  Or ...

Tango, you should add Brett Cecil and Brad Mills to the Toronto Blue Jays.

Since the Giants’ situation is such a mess right now (in terms of who is and isn’t going to make the Active Roster), and with the likelihood that so-called “proven vets” on minor league deals will get a decent amount of playing time, I’d suggest adding the following players:

Non-pitchers:
Juan Uribe
Rich Aurilia
Jesus Guzman
Andres Torres
Josh Phelps

Pitchers:
Justin Miller
Brandon Medders

The list was based on the 40-man players for each team as seen on MLB.com.

It would make my life easier if all those missing players (Andrus, etc) would already be placed on the 40man…

Lucky for me, since I’m doing this database driven, it’s very easy to just add missing players.

***

If you don’t vote for a player, it simply won’t count in the final tally.

***

Don’t worry about things not adding up to 162*9 innings.  When it comes time to it, I’ll make the adjustments…

And whoever wants to post Fanshots at SBNation, please do so with my appreciation!

I updated the main page so you can see how many ballots were cast for each team.

Filled in the Braves’ form, but looking at some of the others, I think a write-in or three would be in order. Especially when the options for Orioles catchers consists of Gregg Zaun.

I’ll try to fill in a few more later in the day.

You’ve obviously got a post about this at BtB, and I’ll send an email to the rest of the baseball writers, although some tend to ignore stuff like this.  Crazy, I know.

Added my take on the Mets, but every indication is that Livan “Ugh” Hernandez will be the 5th starter, so he should probably be added to the list.

I’ll add mine to the mets one too.

I updated each team page so you can see the distribution of votes:
http://www.tangotiger.net/survey/index4.php?teamid=136&team=Seattle%20Mariners

Obviously, we have some a-holes already voting for Ichiro, Beltre, et al with 0 games played in 2009.

Similar situation with Jeter and others playing 0 games:
http://www.tangotiger.net/survey/index4.php?teamid=147&team=New%20York%20Yankees

Don’t worry.  When it comes time to at the end, I will remove obvious junk ballots.

When do the 40-man rosters have to be set by?

I added a note in the front-page, so that people who want to do write-ins can send me an email.

I’m no expert, but I don’t believe there is a strict deadline for the 40-man in the spring.  Players on the 25-man have to be set by opening day, so they have to be on the 40-man then as well.  The only 40-man deadline I’ve ever heard of is for the rule-5 in November.

Tango:
I think you should add Zimmermann, Bard, Colome, and Ledezma to the Nats’ list. There are others (Valentin, Tavarez maybe) who might end up with some playing time, but those initial four should definitely be there. I’ll check back in a couple days to do a ballot with those guys included. Great project.

Looks like choosing the 40-man list wasn’t good enough!  I should have used the spring training stats as well to get the missing players…

Somebody hates the Dodgers...what a shame.

Don’t worry… a-hole picks are easy to pick out.  I’ll be massaging the data when I produce the final results.

All non-40man players that were requested have now been added.

Please add in your votes for those extra players.

SBNAtion-ers, and other team blogs.  I am woefully under-represented by these teams:

5 Marlins
6 Angels
6 Rockies

These could use for a few more votes:
9 WhiteSox
9 Padres
10 Pirates
11 Tigers
11 Royals
13 D’Backs
13 Rangers
15 Nationals

Everyone else is at 20 ballots and higher.

At the bottom of the home page, I have a link to view all players on one page.

I’m overwhelmed by the speed of the responses.  Nearly 1200 ballots in less than 2 days.

I can’t wait to link these playing time forecasts to the Marcels.

Btw, thanks to the Pirates, Whitesox and Nationals who posted so quickly since my last plea.

I added totals to the front page.  You can see that most fans are “high” on the IP.  The total should be around 162*9, or 1458.  There are only 3 teams below that, with the rest much above it.

The one that is very very low is the Padres.  According to their fans, they only have 3 starting pitchers:
http://www.tangotiger.net/survey/index4.php?teamid=135&team=San%20Diego%20Padres

I am 370 innings short on their estimates.  Do they have NRI that expect to make the 25-man roster?

Shawn Hill is a Padres NRI on a minor league contract.

Ok, on each team page, I added the “standard deviation” for each player, so that you can see the spread in one number.

An example is Jason Schmidt, with a mean of 52 and an SD of 47!

http://www.tangotiger.net/survey/index4.php?teamid=119&team=Los%20Angeles%20Dodgers

Kevin Correia’s probably going to make the 25 man roster for the Padres.

Thanks again to all for your ballots.  Close to 1600 already.  And where the heck did all those Pirates fans come from?

I’ll be updating the teams with the extra suggestions here and in emails in a few minutes.

Truly overwhelming the input.  I’m going to let the thing run until Opening Pitch.

***

Any suggestions on how to normalize the innings?  For example, Braves are forecast for 1733 innings (275 too much), and A’s at 1299 (159 too little).  I won’t talk about the Padres.

My first thought is that if there are any undertotals, like the A’s, then I accept as-is, and presume that the A’s will be getting IP from players other than who we see on the roster right now.

Now, for the overtotals, the choices are:
a. remove an equal number from everyone (so if there are 24 pitchers on the Braves ballot, then take out 11 IP from each player)

b. remove 275/1733, or 16% from each pitcher, which means Vazquez and Lowe lose 33 IP

c. remove the bottom pitchers, set them to zero, and do it until all 275 are used up; for the Braves, that means setting to zero everyone from Acosta down

I definitely do not like b, I can kind of get behind c, since no one cares about the bottom pitchers too much, but a seems to have the simple solution that is always appealing.

The other thing I was going to try was a bit more complicated: right now, I use the midpoint of each choice range as the figure I use in the calculation.  So, for the 60-99 range for relievers, I use 80 as the number.  Instead, I was going to let those midpoint numbers float downward as much as I need it so that I get my 162*9 IP.  This way, the workhorses won’t go below 200, in those selections.  And those with 30-59 IP will move down faster than those with 175-199 IP.

It’s a cool programming problem to tackle, but in the end, I have to believe it’ll end up being almost identical to option “a”.

I decided to take a spot a bit below the midpoint.  Works out pretty well.

This is a great idea. I never saw a message thread about the forecasters challenge, but I had a couple of questions.

For the systems that are actively participating, are they allowed to rank their players in any order? Are they expected to somehow follow their own projections?

The question is really about playing time. Since some systems, like Marcel or ZiPS, don’t project playing time, how will they rank their players? Obviously you are using the fans’ input to do playing time, but what about the others?

For the sites like CBS or ESPN, did you think about including an entry just using their draft rankings (cheat sheets), as a separate entry from their projections? That would be interesting.

Matt/34: they can do whatever they want, since I have no way to force them to do otherwise.  Ideally, they stick to a systematic process.

Marcel will use the Crowdsource project.  Otherwise, it’ll end up with an all-injured team (ARod, Mauer, Francis, et al).

For all the inactive participants: I’m going to download their forecasts for the 5 categories, total up their points using the point system of the project, apply some positional adjustment, and boom.

If you go to the home page, and click the “All players” link at the bottom, I made that page sortable.

The cool category to sort on is the “standard deviation” one.  You will instantly see all those players that the Fans are most at odds with.  Among players with a bit of votes (as of the time of this post), that would be:
- Brandon Wood (Angels) among non-pitchers and
- Dan Meyer for pitchers (Marlins).

Okay, I cast my ballot. My alma matter makes it to the final four. I come downstairs to start shouting about it on the internet and I discover that Brandon Morrow has decided that he does not want to be a starting pitcher. Besides seriously killing my final four joy, I wonder what I should do about that community forecast I gave to Tom Tango. What do you do with stuff like this?

Is there anything we can do about Brandon Morrow’s projection? And, of course, what about the pitchers (RRS, for example) who are probably now looking at a starting role?

Had I known this survey would have been as popular as quickly as it was, I would have waited until today or tomorrow before launching (rather than launching a week ago)

As it is, I will just have to live with it.

***

I’m thinking of a way to convert Games for nonpitchers into PA.  I’m thinking of something like this:

PA per G = (2.8 + G/100)

I’ll have to play around with that 2.8 to see what it should be.  That “100” seems to work well-enough.

PA per G
= 2.5 + G/100

The top-end guys with 155 games would come in at 628 PA.  (The top 20 Marcel forecasted players in playing time is around 625 PA.  So, me-likey.)

The total for all the players I have comes in at 181,000 PA, which is pretty close to what you get if you remove pitchers.

I’m pretty happy about how easy and clean it is to convert the G to PA.

I’ll be adding this shortly to the output.

to adjust to the number of innings pitched, maybe you should factor in the standard deviation. The Braves add up to 1611 ip, and the stdev is 503 - to get to 1458, you subtract .30 of the stdev.

The guys you adjust the most are the ones who people are most unsure of. You only take 3 ip off of Vazquez, who is 201 +/- 9, while you take 12 ip off of Tommy Hanson, who has a stdev of 39.

Everyone still is on the list, but your widely accepted workhorses dont get penalized very much for a battle for the 4th and 5th spot in the rotation.

Cast 1st Vote for Marlins - Ross Gload in mix to play?

Oooohhhh… I love the idea of adjusting based on STDEV.  Fantastic!