TotalMinors Ratings Discussion

[JaFFiTh]

Hmm..ok how do you do the pitchers control and movement ratings?

I've done batters stuff, it can be pretty easy to figure out but pitchers..how the heck can you rate their control and movement, gotta be confusing 8O

[GForce22]

Well you're right about one thing...hitters are MUCH easier than pitchers to get right.

It got revised/global edited when it became clear the lack of walks issue was so prevalent, but the original formulas for pitching went this way...

Stamina: Easy if I had Pitches/GS. This was just that number - 12.

What if I didn't have that figure? Bear with me...this gets a little out there:

(((((IP/G)/5.75)*0.*80)+(((1-(walks/IP ratio relative to league adjusted for age and level)*0.1)*80)+((Ks/IP ratio relative to league adjusted for age and level*0.1)*80),0))

IP/G is an obvious component, but the question is why walks and Ks? Simple...the more walks and Ks a pitcher racks up, the more pitches he's throwing. So these need to be factored along with the innings total to really see how much this guy is racking up on his arm. A guy like Al Leiter for example, doesn't throw many innings, but it's not because of endurance, it's because he walks so many guys, and he strikes out a good amount too. So those things needed to be factored in to the stamina equation.

As for control, I think of this more in terms of "command," so it involves more than just walks.

The original control formula last year (I concede I don't remember if I used this as average control rating or peak control rating...believe it was average though):

(((R1*80)*0.5)+((T1*80)*0.35)+((U1*80)*0.15),0)

Where R1=weighted 3-year BB/IP ratio adjusted for league, age and level, T1=weighted 3-year K/BB ratio adjusted for league, age and level, and U1=weighted 3-year H/IP ratio adjusted for league, age and level.

Walks/IP is the strongest weighted, followed by K/BB and then hits/IP.

Now, movement, similar, just a couple of different variables, focusing more on Ks than walks, as you would expect:

(((S1*80)*0.7)+((U1*80)*0.2)+((V1*80)*0.1),0)

Where S1=weighted 3-year K/IP ratio adjusted for league, age and level, T1=weighted 3-year H/IP ratio adjusted for league, age and level, and V1=weighted 3-year HR/IP ratio adjusted for league, age and level.

By the way, these variables being multiplied by 80 is what I worked off as the base for ratings. An average major league pitcher would be 80s across the board. So everything worked off being above or below that figure.

Obviously, you don't have my weightings for age and league, but this should as least give an idea how I went about it.

Questions or comments, fire away.

GH

[JaFFiTh]

Oh boy, so confusing :| I think I'm just gonna stick with the batting stuff

Only thing i pay attention to the pitchers are their pitches, but movement and control..i'd just leave it at default

[GForce22]

Believe me, it sounds more complicated than it really is. I guess it also just comes easier to me because a) I created the formulas and I've been using them for 5+ years.

Seriously, though, all it is is math. By the way, those numbers WERE averages...I built off those averages in assigning ratings to the individual pitches. Secondary pitches lower, top pitches better, obviously, giving way to scouting report evaluations of pitches in ranking them.

GH

[llcmac]

Hey GH,

I am a roster maker and I was just wondering, is there any easy way to rate minor leaguers? Like if I have all their stats, should I just weight them differently? like take 20 percent off for AAA. Any help would be appreiated.

[GForce22]

Hey GH,

I am a roster maker and I was just wondering, is there any easy way to rate minor leaguers? Like if I have all their stats, should I just weight them differently? like take 20 percent off for AAA. Any help would be appreiated.

I would say, if you're going to do it quickly, dig up the ZIPS projections and work off of those. You can find them as BaseballPrimer.com. They are a good projection tool for both pitchers and hitters.

GH

[livnlegn]

I don't know. I have always been of the mindset that talent exists independently of stats. Therefore, rating a guy based on statistical measure is somewhat flawed. I do not contend that stats are completely irrelevant (in fact, they are more accurate representations of talent when sustained over a long period) but they are often misleading and, therefore, only a small fraction of talent evaluation. Since ratings are a measure of talent I tend to rely on scouting reports and observation.

[GForce22]

I don't know. I have always been of the mindset that talent exists independently of stats. Therefore, rating a guy based on statistical measure is somewhat flawed. I do not contend that stats are completely irrelevant (in fact, they are more accurate representations of talent when sustained over a long period) but they are often misleading and, therefore, only a small fraction of talent evaluation. Since ratings are a measure of talent I tend to rely on scouting reports and observation.

Ratings aren't a measure of talent, they are a measure of performance. This isn't OOTP, where you rate a guy's current and potential...you're just rating current status here, how he's likely to perform now. The star rating, via MVPEdit, is the only place you can adjust for talent/potential.

GH

[GForce22]

Bunting was brought up earlier...I've never really done formulas for that, but let's see if we can come up with something that works.

First thing is hitters and pitchers need to be evaluated separately, given that pitchers are called on to bunt significantly more often.

So let's look at position guys first:

Sac Hits leaders (Sac Hits/AB)

Royce Clayton - 24 (4.2%)

Adam Everett - 22 (5.7%)

Omar Vizquel - 20 (3.5%)

Derek Jeter - 16 (2.5%)

This isn't a stat where I necessarily think the difference between an 85 bunter and a 75 bunter matters a whole hell of a lot. So what about a scale of this sort:

> 2.5% : 95

> 2% : 75

> 1.5% : 55

> 1% : 35

> .5% : 20

< .5% : 10

Thoughts, comments? I'll look at pitchers later.

GH

[ZMan]

The only problem is that sac hits/AB can be tremendously influenced by managerial decisions. Some managers like playing small-ball, while others don't. You can see this difference markedly as players switch teams. Opportunity does not completely equate with "chance of successful bunt."

Though, in any formula, sac hits/AB should probably play some role. It is evidence of performance. For non-pitchers, I would suggest a formula that takes into account SH/AB as well as batter speed and perhaps contact rating. Speedier players tend to have a quicker reaction time and facility in squaring up to bunt (Juan Pierre, Luis Castillo, Ichiro, etc.), and having good contact ratings certainly benefits these types of guys as well.

[GForce22]

The only problem is that sac hits/AB can be tremendously influenced by managerial decisions. Some managers like playing small-ball, while others don't. You can see this difference markedly as players switch teams. Opportunity does not completely equate with "chance of successful bunt."

Though, in any formula, sac hits/AB should probably play some role. It is evidence of performance. For non-pitchers, I would suggest a formula that takes into account SH/AB as well as batter speed and perhaps contact rating. Speedier players tend to have a quicker reaction time and facility in squaring up to bunt (Juan Pierre, Luis Castillo, Ichiro, etc.), and having good contact ratings certainly benefits these types of guys as well.

Certainly managerial decisions impact this, but they impact everything, pretty much. It's like saying a steal aggressiveness rating can't really work off steal attempts because of managerial decisions. Or pitcher endurance based on pitches thrown or innings, because some managers will pull a guy more quickly. I understand what you're trying to get at, but the variable you're trying to introduce can be introduced as some sort of factor in all sorts of ratings, and trying to evaluate that variable accurately is a tricky and tedious game.

Also, understand my perspective...I come from the standpoint of what a player does, not what people say he can do. Factoring in contact rating is a pandora's box...lots of guys have good contact ratings but can't bunt a lick. Fast guys often are better bunters because they are taught how early on so they can utilize their speed, but not because of reaction time...if their reaction times were better than slower guys then your best hitters would always be fast guys because they would react more quickly to the pitch or any changes in its trajectory.

One factor that does need to be included in addition to sac hits is bunt hits. Unfortunately, neither ESPN or MLB.com has this number. Not sure if one of the other sites has it, but it should definitely be in there for guys like Ichiro and Castillo, who bunt well for hits but aren't often called upon to sacrifice (5 and 2, respectively).

I'll see if I can find bunt hits data, certainly a necessary component. But as for factoring speed and contact into the equation, I'll agree to disagree.

GH

[bbythepier]

GForce-

i mistakenly put this in another thread. appreciate any comment.

GForce-

i assume when you say "weighted" 3-year stats and projections you are implying something like (general idea, not the #'s):

15%- ZIP projection

40%- previous year

30%- 2 years back

15%- 3 years back

may i ask, since you have experience as a professional roster developer, are these things that were agreed upon by:

1) a "round table" of stat geeks

2) starting with an actual year in the past and working backwards to find the best regression statistic to fit the "weights"

3) to fit agreed upon ratings (i.e., make the data fit the theory, as my former professors would call it)

4) something totally different

when you say some stats were weighted according to their stadium (location), how were you able to do this in the minors? did you have to it by league? are there splits available per stadium?

when evaluating speed, the most effective (almost universally) method is home to first times. A professional article posted at: http://www.asep.org/jeponline/issue/Doc/Ju...4/ColemanV2.pdf

"Some teams routinely time players during spring drills at 30 and 60 yard intervals. These data are useful,but the times are artficial. Players, who run good times in a speed test, don't always run as fast in games situations. Game times are more valid than speed tests because they tell coaches how fast a player is when it counts, in game situations." Scouts routinely keep records of these in-game times for nearly every player. does the public ever see any of these numbers? how about a company that spends million$ for licensing rights?

when using age "weight" do you have a pyramid model where say age 28-31 is a player's prime years and the "weight" falls off as you go to both sides of this 'prime' period?

just curious? picking your brain. thanks for the time.

[qpontiac]

GForce-

when you say some stats were weighted according to their stadium (location), how were you able to do this in the minors? did you have to it by league? are there splits available per stadium?

My knowledge is nowhere near that of GForce on baseball stats, but BaseballProspectus has touched on both of these issues in great detail, and I am an addict of BP...

(Correct me if I am wrong GForce)

With regards to your question above, BP uses an Equivilent formula or Eq...

Their definition is this:

Equivalent Average. A measure of total offensive value per out, with corrections for league offensive level, home park, and team pitching. EQA considers batting as well as baserunning, but not the value of a position player's defense. The EqA adjusted for all-time also has a correction for league difficulty. The scale is deliberately set to approximate that of batting average. League average EqA is equal to .260. EqA is derived from REqA, which is (H + TB + 1.5*(BB + HBP) + SB + SH + SF) divided by (AB + BB + HBP + SH + SF + CS + SB/3). REqA is then normalized to account for league difficulty and scale to create EqA.

With regards to Speed, they have a formula for Speed Score:

Speed Score (SPD) is one of five primary production metrics used by PECOTA in identifying a hitter's comparables. It is based in principle on the Bill James speed score and includes five components: Stolen base percentage, stolen base attempts as a percentage of times on first base, triples, double plays grounded into, and runs scored as a percentage of times on base.

Here are the formulae:

SpS1

Speed score based on SB%

((SB+3)/(SB+CS+7)-0.4)*20

SpS2

Speed score based on SB attempts

SQRT((SB+CS)/((H-2B-3B-HR)+BB+HP))/0.07

SpS3

Speed score based on triples

3B/(AB-HR-K)/0.02*10

SpS4

Speed score based on runs per time on base

((R-HR)/(H+BB-HR-HP)-0.1)/0.04

SpS5

Speed score based on GDP's

(0.055-GDP/(AB-HR-K))/0.005

Speed Score

Net Speed Score

Average of top 4 Speed Scores above

[bbythepier]

thanks q-

i'm familiar with BP and your comment is great.

i guess what i was asking is (even though i understand EQA) does GForce use anything different from BP's EQA. Our Pacific Coast League out here is widely considered a "hitter's friendly" league with a high % of small "bandboxes" for stadiums. Our player's HR's are not representative of the Minors as a whole. Have they developed a "weight" system to account for this disparity between leagues and within leagues, or is this encapsulated in the EQA.

Also, i'm aware of the speed calc's put forth by BP and GForce on a previous post in this thread. And as GForce suggests, this merely the "best attempt" at representation through "numbers".

When playing Madden, however, we know that Deon Sanders can run a 40 in 4.21 or Maurice Clarrett in only 4.7, so we scale accordingly.

First base coaches routinely use a stopwatch to garner a pitcher's "speed to home plate". They also routinely use a similar method to time a player's speed to first base, namely on 100% effort plays like trying to beat out an infield hit or double play (while compensating for righties vs. lefties). We know that Ichiro can do this in about 3.5-3.6. seconds. One could assume that a chart of these #'s is out there somewhere. My question, are they 'top secret', or available in some capacity. This way speed is determined by the clock instead of the calculator. My question may be answered in the fact that a guru like James needs to develop such a 'speed calculation', but it may also be baseball followers overreliance (or infatuation) on stats.

[bbythepier]

qpopntiac-

by the way, i'm making the annual trek to spring training this weekend, into your neck of the woods. a little sunshine and the "crack of the bat" will be a pleasant escape from the fog.

[JaFFiTh]

hey GForce..I was testing speed for some players with your forumlas provided, but for what i looked thru..wasn't quite that accurate. Alexis Rios has a pretty good speed, but it gave him only a 57, and Russ Adams only a 29 (doesn't help when he didn't play much) but they're kinda off.

[llcmac]

hey GForce..I was testing speed for some players with your forumlas provided, but for what i looked thru..wasn't quite that accurate. Alexis Rios has a pretty good speed, but it gave him only a 57, and Russ Adams only a 29 (doesn't help when he didn't play much) but they're kinda off.

hmm, using his formula I got 69 for Rios

[JaFFiTh]

Errmm.....I did exactly what he did and i got the exactly same number for Ichiro that he posted, so....I don't think i did it wrong..i have it all set up in Excel, so i don't have to calcuate them over and over

Also he still should be better than 69 lol. Maybe 75ish is good i think.

[GForce22]

hey GForce..I was testing speed for some players with your forumlas provided, but for what i looked thru..wasn't quite that accurate. Alexis Rios has a pretty good speed, but it gave him only a 57, and Russ Adams only a 29 (doesn't help when he didn't play much) but they're kinda off.

I'm looking at this more because I didn't like some of the numbers coming out either. Probably a weighting issue on my part.

GH

[GForce22]

hey GForce..I was testing speed for some players with your forumlas provided, but for what i looked thru..wasn't quite that accurate. Alexis Rios has a pretty good speed, but it gave him only a 57, and Russ Adams only a 29 (doesn't help when he didn't play much) but they're kinda off.

I'm thinking that the 1 year rating may be part of the culprit. Running Adams' 3-year minor league numbers through, he scores a 70.

Rios' problem seems to be that he hits into a good amount of DPs. This ties in with what I said about recognizing outliers...if you know a guy is fast, every scouting report says he's fast, adjust accordingly.

GH

[GForce22]

Errmm.....I did exactly what he did and i got the exactly same number for Ichiro that he posted, so....I don't think i did it wrong..i have it all set up in Excel, so i don't have to calcuate them over and over

Also he still should be better than 69 lol. Maybe 75ish is good i think.

Not bad...the Speed Score from PECOTA puts him at 76.2 (actually 7.62 but we move the decimal, obviously).

I'll run Adams through next.

GH

[GForce22]

Using Adams' AAA numbers, since that's where he played more often, his Speed Score is a 4.69, or 47.

GH

[llcmac]

I have gotten way too into this. I made a formula for contact and power. I have been trying to find a modifier for players who havent had enough AB's......I have been using 700 ab's, might adjust it.

[GForce22]

I have gotten way too into this. I made a formula for contact and power. I have been trying to find a modifier for players who havent had enough AB's......I have been using 700 ab's, might adjust it.

I've worked off 550 or 600.

GH

[GForce22]

I have been quietly doing a complete overhaul of the TotalMinors formulas and projections and am now coming very close to the well-regarded ZIPS projections used by Diamond Mind Baseball and sabermetricians. Some top prospects I've run through recently.

Victor Diaz - METS

TotalMinors Projection: .293 AVG/.458 SLG

ZIPS Projection: .292 AVG/.472 SLG

Justin Morneau - TWINS

TMP: .289/.533

ZIPS: .275/.539

Nick Swisher - A's

TMP: .240/.429

ZIPS: .234/.422

Jeremy Reed - Mariners

TMP: .307/.440

ZIPS: .285/.412

Dallas McPherson - Angels

TMP: .280/.551

ZIPS: .275/.513

Chris Burke - Astros

TMP: .285/.411

ZIPS: .284/.401

I am also revising the sheets so virtually everything can be computed on them, or at least entered manually to have in an orderly fashion, including my new calculations for speed, stealing and baserunning, as well as bunt ratings and defensive data.

It has take/chase/miss ratings, HR%, basically everything except spray charts because I'd be fooling myself if I thought I could do those accurately.

Quite pleased with how this looks right now. If I can get the defensive stuff looking how I want, this will be a very valuable tool.

GH