Baseball Prospectus

Attie · Post by **Attie** » Tue Mar 08, 2011 8:32 am

Just saw the new Baseball Prospectus on the shelves today. As usual, the PECOTA projections made me laugh out loud. They predict THREE pitchers only to have more than 14 wins this year........Felix, Haren and CC with 15........I don't even know if this is mathematically possible. Their projections are so conservative that they are totally useless........Roy Halladay projected to get 14 wins? Really? Does anybody know what PECOTA stands for anyways?

P rojecting

E xcremental

C rap

O ut

T he

A ss

Hells Satans · Post by **Hells Satans** » Tue Mar 08, 2011 8:39 am

The core rate stats are very good, but BP isn't meant for fantasy.

Navel Lint · Post by **Navel Lint** » Tue Mar 08, 2011 8:47 am

Projecting win totals for individual pitchers is not easy. But to put win totals in general into perspective; over the past 28 non-strike years, no season has had fewer than 17 pitchers win 15+ games and the most was 33 pitchers in 1985 and 1989.

That doesn’t mean that BP’s projections are wrong…….. but there are other people here that can speak to this issue better than I.

Attie · Post by **Attie** » Tue Mar 08, 2011 8:52 am

Russel, you hit the nail on the head! How can Baseball Prospectus get away with publishing such garbage? There's never been a season with fewer than 17 pitchers with at least 15 wins, and they project only 3.......and they projected only 1 last year if memory serves.......if you read that book, and all their regression to the mean projections, you'd think all 30 teams will go 81-81 this year.............

Layin'theSmackDown · Post by **Layin'theSmackDown** » Tue Mar 08, 2011 8:52 am

since every hitter in the book is good for about 15 bombs the starting pitcher win totals sound about right.

the above statement is Dead On: 7% Somewhat Right: 23% Clueless: 45%

Wont be read: 78% on BP.com: 1%

Navel Lint · Post by **Navel Lint** » Tue Mar 08, 2011 9:04 am

Originally posted by Attie:

Russel, you hit the nail on the head! How can Baseball Prospectus get away with publishing such garbage? There's never been a season with fewer than 17 pitchers with at least 15 wins, and they project only 3.......and they projected only 1 last year if memory serves.......if you read that book, and all their regression to the mean projections, you'd think all 30 teams will go 81-81 this year............. Just to be clear…..I’m not making a statement one way or the other about BP. I don’t read the site, I’m not a member and I don’t get the book (although I did once in the past). I’m just laying out the facts about the win totals. Nothing more or less.

bjoak · Post by **bjoak** » Tue Mar 08, 2011 5:31 pm

Originally posted by Attie:

Russel, you hit the nail on the head! How can Baseball Prospectus get away with publishing such garbage? There's never been a season with fewer than 17 pitchers with at least 15 wins, and they project only 3.......and they projected only 1 last year if memory serves.......if you read that book, and all their regression to the mean projections, you'd think all 30 teams will go 81-81 this year............. To be clear, as someone who has a few articles on their site I don't feel obligated to come to BP's defense and I do agree that PECOTA uses too much regression to the league mean, but I think you are a tad confused on how projections work.

They are supposed to be "weighted means" rather than "true outcomes." That means that they are giving you a player's average performance if you replayed the season maybe a million times.

Let's look at a hypothetical: pretend for a minute that every SP had the same exact skills, health, etc. and played with a league average team, ballpark, etc. And let's say they all project for 10 wins. At the end of the season, do you think every one would have 10 wins? No. You might have 1 with 15, 1 with 5 and a whole lot in between, but you'd probably have less with 10 than all the other numbers combined.

Does that mean a projection system like PECOTA should arbitrarily choose some of those identical pitchers for 15 wins or 9 or whatever just so that it has the same number as the real world? No. It's going to project them all for 15, come what may.

Back in the real world, baseball isn't exactly like that, but you must realize that luck plays a huge part in stats and especially wins. That is why we might have a 20 game winner, but no one knows who it will be and he's not a perennial 20 game winner. You need some luck to do that well. Players tend to be a lot closer to the middle when you take out the luck. And that's what any good projection system does, it takes out the luck.

Doug Atkins · Post by **Doug Atkins** » Tue Mar 08, 2011 7:27 pm

bjoak, thanks for the explanation. Although I hear what you're saying, I can't fully buy it.........if you played the season out a million times, Roy Halladay would exceed 14 wins at least 950,000 times.......but if you don't like my wins analysis, try this one.......Baseball Prospectus also predicts only three players to score 100 or more runs this year.......THREE!.......last year, in a low scoring year of the pitcher, there were 17.......their projections just simply don't compute.

BigHurt · Post by **BigHurt** » Tue Mar 08, 2011 8:19 pm

Wait a second...are you telling me PECOTA isn't "deadly accurate"? There goes my season.

whipsaw · Post by **whipsaw** » Wed Mar 09, 2011 1:29 am

Originally posted by Doug Atkins:

bjoak, thanks for the explanation. Although I hear what you're saying, I can't fully buy it.........if you played the season out a million times, Roy Halladay would exceed 14 wins at least 950,000 times.......but if you don't like my wins analysis, try this one.......Baseball Prospectus also predicts only three players to score 100 or more runs this year.......THREE!.......last year, in a low scoring year of the pitcher, there were 17.......their projections just simply don't compute. How much money can I put on Halladay under 14.5 wins at 20 to 1? I'll take as much action as you can handle.

While I agree that their projections are a bit conservative and overly regressed, they do have value. If you're just looking at the raw number, you're missing the point.

Thunder · Post by **Thunder** » Wed Mar 09, 2011 3:06 am

Originally posted by whipsaw:

quote:Originally posted by Doug Atkins:

bjoak, thanks for the explanation. Although I hear what you're saying, I can't fully buy it.........if you played the season out a million times, Roy Halladay would exceed 14 wins at least 950,000 times.......but if you don't like my wins analysis, try this one.......Baseball Prospectus also predicts only three players to score 100 or more runs this year.......THREE!.......last year, in a low scoring year of the pitcher, there were 17.......their projections just simply don't compute. How much money can I put on Halladay under 14.5 wins at 20 to 1? I'll take as much action as you can handle.

[/QUOTE]Yep, and a month ago, Adam Wainwright was an NL CY young candidate for 2011..

bjoak · Post by **bjoak** » Wed Mar 09, 2011 6:47 am

Originally posted by Doug Atkins:

bjoak, thanks for the explanation. Although I hear what you're saying, I can't fully buy it.........if you played the season out a million times, Roy Halladay would exceed 14 wins at least 950,000 times.......but if you don't like my wins analysis, try this one.......Baseball Prospectus also predicts only three players to score 100 or more runs this year.......THREE!.......last year, in a low scoring year of the pitcher, there were 17.......their projections just simply don't compute. As I said, they do have too much regression so I agree on that point. I don't think PECOTA has the same value to a fantasy player that it used to. I'm just arguing that the number of real players who have such and such a stat in a year is irrelevant to the projections for individuals.

And I also agree with the Wainwright comment. If you replayed the season a million times Halladay would sometimes have injuries and off years and all of the other issues that many of the best pitchers often have. There are also scenarios where the Phillies offense falls apart. And if you don't buy that, look at what has already gone down this spring.

Beyond all that, it is hard to project any pitcher above about 16 wins. Do I think that luck can drop him down to 14 even in a good year. Yeah, I buy that. I'm going to take the under on the 950,000.

[ March 09, 2011, 12:48 PM: Message edited by: bjoak ]

Hells Satans · Post by **Hells Satans** » Wed Mar 09, 2011 7:50 am

For the most part, I agree with the overregression point, but in my experience, virtually all other projections (other than something like Marcel) pretty dramatically overproject for batting average.

Attie · Post by **Attie** » Wed Mar 09, 2011 8:15 am

Guys, it looks like I'm going to be a rich man at the end of the season.......sign up here for your 20/1 odds........but the over/under for Halladay is 14 wins, as stated in Baseball Prospectus, not 14.5........so 14 is a push.........while we're at it, I'll offer the same 20/1 odds that there are more than 3 players who score over 100 runs......

bjoak · Post by **bjoak** » Wed Mar 09, 2011 9:28 am

Originally posted by Attie:

Guys, it looks like I'm going to be a rich man at the end of the season.......sign up here for your 20/1 odds........but the over/under for Halladay is 14 wins, as stated in Baseball Prospectus, not 14.5........so 14 is a push.........while we're at it, I'll offer the same 20/1 odds that there are more than 3 players who score over 100 runs...... Most people won't take that bet. I have him at 16 so I'm going to take the over on14. I was just saying that if you redid the season 20 times Halladay would have more that one where he doesn't reach 15. He'll probably also have more than one where he reaches 18. See how this works?

[ March 09, 2011, 03:30 PM: Message edited by: bjoak ]

whipsaw · Post by **whipsaw** » Wed Mar 09, 2011 9:50 am

Originally posted by Attie:

Guys, it looks like I'm going to be a rich man at the end of the season.......sign up here for your 20/1 odds........but the over/under for Halladay is 14 wins, as stated in Baseball Prospectus, not 14.5........so 14 is a push.........while we're at it, I'll offer the same 20/1 odds that there are more than 3 players who score over 100 runs...... The original post said WOULD EXCEED 950,000 times, not WOULD GET AT LEAST. I don't really care what BP said.

bjoak · Post by **bjoak** » Wed Mar 09, 2011 9:53 am

I'll offer the same 20/1 odds that there are more than 3 players who score over 100 runs At the risk of banging my head against the wall, I'll repeat that PECOTA doesn't think that only three players will reach the 100 run mark; it thinks there are only three guys you can project that high.

If its goal was to tell you how many players do it, it would just give you the average of the last three years.

Try the converse of your bet. See how many takers you get that say one of those exact three players will not reach 100 runs. Lots of people will bet that especially if you give them the 20 and they give you the one.

It is all about what you can expect individuals to do--not whole groups.

leonem4444 · Post by **leonem4444** » Wed Mar 09, 2011 12:52 pm

In my opinion, the best way to use a projection system (like BP) that uses lots of regression is to think of it more as a rankings system than to use the specific projections.

TickyBurden · Post by **TickyBurden** » Wed Mar 09, 2011 6:03 pm

the BP is garbage? really? please list the superior guides.

Sebadiah23 · Post by **Sebadiah23** » Wed Mar 09, 2011 7:25 pm

Originally posted by leonem4444:

In my opinion, the best way to use a projection system (like BP) that uses lots of regression is to think of it more as a rankings system than to use the specific projections. Listen to the mafia.

Use it to compare players to each other , not to add up your projected team stats to see if they hit past target benchmarks.

ESPN8 THE OCHO · Post by **ESPN8 THE OCHO** » Thu Mar 10, 2011 2:10 am

Attie,

I will put 100 on the under for Halladay 14 wins at 20-1. Let me know if you are really putting your money where your mouth is on that one.

Thanks,

HumenFund

bjoak · Post by **bjoak** » Thu Mar 10, 2011 6:59 am

Originally posted by ESPN8 THE OCHO:

Attie,

I will put 100 on the under for Halladay 14 wins at 20-1. Let me know if you are really putting your money where your mouth is on that one.

Thanks,

HumenFund Make sure you are getting the 20, but I assume that will be the end of this thread. He's not going to put $100 up against $5. If he does, you can sign me up too as long as there is someone to hold the cash.

[ March 10, 2011, 01:01 PM: Message edited by: bjoak ]

Layin'theSmackDown · Post by **Layin'theSmackDown** » Thu Mar 10, 2011 8:44 am

original odds listed were 19-1. dont overpay!

Layin'theSmackDown · Post by **Layin'theSmackDown** » Thu Mar 10, 2011 8:57 am

and to make the bet accurate halladay should have to pitch at or near his bp projected totals..32 gs and 239 ip. no free injury cashes

headhunters · Post by **headhunters** » Thu Mar 10, 2011 9:23 am

kinda think if you did the above it would sort of negate a big reason for regression. inuries are cooked in. sort of like the "over" on 20 homers by utely