Player Ratings--Metrics

[DarkEnigma510]

I am starting to fool around with adjusting player ratings for my own purposes. I am considering a somewhat non-scientific way of implementing this for my own players but just wanted some imput from people who like to do their own ratings. I am not asking anyone to reveal any secret formulas they used and I know this can be made very statistical if someone has the time to do so. However, I'm more interested in knowing certain metrics people might use to assess a player;s ratings.

POWER-- I am considering Isolated Power (which is Slugging Pct less Batting Average) which I think is preferable over Slugging percentage by itself since a good average hitter like ichiro could bump up his slugging, not due to his power but his ability to generate hits. I think Ichiro is over-rated in the game in terms of power and if you sim a season you'll often see Ichiro hitting 15-20 homeruns when it should be more like 7-15.

DISCLIPLINE-- There are two metrics I am considering....I don't know if these actually have an official name...but (OBP-Batting Avg) is a pretty good major indicator of a players rating as a hacker or disciplined hitter, followed by a player's SO/BB ratio, with the higher the ratio, the more of a free swinger they are. You do seem to get anomolies using this (Again, Ichiro doesn't walk much but this is more because he always makes contact, but he certainly isn't undisciplined) but it seems to work OK for most players and can pick out the most patient players (Bonds, Sheffield).

CONTACT-- The simplest metric here is batting average but its possible for a player to have a high average but also strike out a lot (i.e. Jim Edmonds) I was thinking using (AB-SO)/AB to better assess contact since this measure the percentage of times a player puts a ball into play.

ANNUAL, MULTI-YEAR, PROJECTED-- What do most people here use? Life Time Stats? Last 3 years? Projected Stats? One Year?

Just curious about any feeback or comments on what stats people might rely on.

[uni91]

I am using formulas that I found in another post on this forum....cant remember...but ill post later if I find it. Anyway, I use those formulas in excel, which makes it real easy....just input the required stats from the player splits, and voila! Depending on the situation, I use a weighted average by averaging last years splits with the players splits from the last 3 years. Last years stats end up having a 50% weight on the rating while the other 2 years each have a 25% weight each. But for younger players, you have to make your own decision (ex. In limited AB's, J. Maur killed RHP last year...like .420, it doesn't mean he deserves a 99 in contact versus RHP, he needs to do it longer...so you have to make your own decision on those) If you want the spreadsheet I use...ill e-mail it to ya.

[wikitundra]

Can you email your formulae as well?

wikitundra@yahoo.com

Thanks in advance...

[poonani]

poonani3869@hotmail.com

id appreciate it as well

[boulet430]

I am using formulas that I found in another post on this forum....cant remember...but ill post later if I find it. Anyway, I use those formulas in excel, which makes it real easy....just input the required stats from the player splits, and voila! Depending on the situation, I use a weighted average by averaging last years splits with the players splits from the last 3 years. Last years stats end up having a 50% weight on the rating while the other 2 years each have a 25% weight each. But for younger players, you have to make your own decision (ex. In limited AB's, J. Maur killed RHP last year...like .420, it doesn't mean he deserves a 99 in contact versus RHP, he needs to do it longer...so you have to make your own decision on those) If you want the spreadsheet I use...ill e-mail it to ya.

I'd like this spreadsheet as well if possible.

bigtexyank@hotmail.com

[ZMan]

Below is what I posted a few weeks back to another item...while tinkering with the stats, I have found that, in terms of SX and PX (explained below), there is pretty good correlation with EA's ratings for speed and power. But to get that, you need to scale the ratings such that the lowest is around 40 or so and the highest 99-100 (that is at least the case for the speed rating, I forget what EA's power ratings bottom out at).

This seems to me to be the key...take a look at the ratings in the roster set, see what the low and high values for players are...use a separate formula to rank the players, and then fit them inside that low-high curve.

It seems that (BB:K) ratio should determine plate discipline, while bunting, the way EA does it, seems to incorporate some combination of player speed and contact rating.

It doesn't seem like batting avg. is a very good measure of contact. After all, hitting the ball and hitting them "where they ain't" are two different things. I'd second Enigma's suggestion: ((AB-K) / AB)...seems like a more true contact rating.....again, you'll need some way to scale all the players such that they fit EA's chosen highs and lows on the 100 point scale.

***

Potentially useful formulas:

Ron Shandler has two interesting measures for power and speed, the appropriately named SX (Speed Index) and PX (power index). Both run on a 1-200 scale, with 100 being the league average, so they could be easily divided by 2 to produce a rating for MVP Baseball.

SX is based off an old Bill James formula. It is calculated as the mean value of the following four elements:

1) Stolen base efficiency = (((SB+3)/(SB+CS+7))-.4) * 20

2) Stolen base frequency = Square root of ((SB + CS)/(Singles + BB)) / .07

3) Triples rating = (3B / (AB-HR-K)) and the result assigned a value based on the following chart:

< 0.001 = 0

0.001 = 1

0.0023 = 2

0.0039 = 3

0.0058 = 4

0.008 = 5

0.0105 = 6

0.013 = 7

0.0158 = 8

0.0189 = 9

0.223+ = 10

4) Runs scored as a percentage of times on base = (((R-HR)/(H+BB-HR))-.1) / .04

The SX is derived from taking this total (Batter's Spd.), dividing it by the League Avg. Speed, and multiplying by 100. For our purposes this would be halved. (SX/2)

The PX, or Power Index, is based off of Linear Weighted Power:

((Doubles x . + (Triples x . + (HR x 1.4)) / (At bats - K) x 100

Again, this number is divided by the league average, multiplied by 100 and you have a clean power rating. PX/2 would make for a pretty decent power rating in MVP.

[ZMan]

The other issue is that I think you have to be careful about rating minor leaguers too highly. It seems to me that you need a good set of MLE's and worry more about being judicious with awarding the star ratings.

For some serious roster work, I'd suggest:

Ron Shandler's Baseball Forecaster (has good metrics for contact, speed, power, plate discipline, all there. If you have the book you can download spreadsheets with the same info. Also has a great set of MLEs.)

Baseball America's Prospect Handbook (just released, I think. This is the best source for more subjective ratings material. Also indispensable for rating pitchers, for whom creating ratings are far more difficult.)

Baseball Prospectus: Not as useful as the other two for roster making, but it has two interesting ratings: Fielding runs for judging how good a fielder is, and Stuff for rating pitchers.

John Sickels' Baseball Prospect Book 2005: More help on getting down those "subjective" ratings. He is the guy I trust for determining the star rating, as he rates all the top prospects on an A-F grading system. Seems sensible that:

A= 5 stars

B= 4 stars

C= 3 stars, and so on.

[poonani]

John Sickels' Baseball Prospect Book 2005: Seems sensible that:

A= 5 stars

B= 4 stars

C= 3 stars, and so on.

That is simply not the case. In this game, the # of stars corresponds to the ceiling of the player and that is not what sickels is going by

[Nomarhits400]

It doesn't seem like batting avg. is a very good measure of contact. After all, hitting the ball and hitting them "where they ain't" are two different things. I'd second Enigma's suggestion: ((AB-K) / AB)...seems like a more true contact rating.....again, you'll need some way to scale all the players such that they fit EA's chosen highs and lows on the 100 point scale.

While I certainly agree with your reasoning- I am positive that the game doesn't operate in that fashion; i.e., if you order contact rating by the above metric- it will work for guys like Bonds and Pujols- but not so well for guys like Ramirez and Ortiz. So, unless you want guys like Pierzynski, , Polanco, Eckstein, Alfonzo, Hatteberg, Hillenbrand, Vizquel and Deivi Cruz leading the league in hitting (all in the 20 or 30 in k/ab)- K/AB won't work. If you're not sure- try reordering contact ratings in this fashion and simming a season or two. The plain and simple truth is EA doesn't account for enough variables to accurately represent baseball statistics. Plate discipline is a great example- but by itself it can't accurately simulate guys like Bellhorn, Thome and Abreu who strike out AND walk over a hundred times...Makes things difficult

[JaFFiTh]

I am using formulas that I found in another post on this forum....cant remember...but ill post later if I find it. Anyway, I use those formulas in excel, which makes it real easy....just input the required stats from the player splits, and voila! Depending on the situation, I use a weighted average by averaging last years splits with the players splits from the last 3 years. Last years stats end up having a 50% weight on the rating while the other 2 years each have a 25% weight each. But for younger players, you have to make your own decision (ex. In limited AB's, J. Maur killed RHP last year...like .420, it doesn't mean he deserves a 99 in contact versus RHP, he needs to do it longer...so you have to make your own decision on those) If you want the spreadsheet I use...ill e-mail it to ya.

I'd appreciate if if you send me the spreadsheet as well

jaffith2002@yahoo.ca (not com...CA! )

[MikeB]

Go watch Ichiro at batting practice. He can hit the ball as hard and the best but he is a leadoff hitter whose main goal is to get on base. What sort of leadoff hitter hits home runs? Not on my team.

[ZMan]

EA needs to employ more stats nerds.

Re: Contact:

What formula might we derive that takes into account both the (AB-K/AB) formula and AVG.? Perhaps a hybrid of the two? Maybe that is the point of EA's rating...that it is not just the hitter's ability to get bat on ball but also to be able to "direct" the ball into base hits. Which is extremely weird, but I'll buy it.

Re: Stars rating:

I'm not entirely sure how the EA player progression models work, or how many players they give 5 or 4 stars, etc. but it seems to me that, like the Sickels ratings, there should be few prospects with 5-star ratings. What I like about the Sickels grades is that he is very stingy with A's and B's....and that's realistic, given how few blue-chip prospects make it through to become regular starters.

[DarkEnigma510]

I thought I would chime in with my thoughts after player around.

I went into the Lahman database and looked at the stats for the past 3 years. I removed all pitchers and any player who had less than 100 at bats in a year. I was trying to figure out what the upper and lower levels for certain stats were. Specifically:

ISOlated power (SLG-BA)--Power

SO/BB--Discipline

(OBP-BA)*1000--Discipline Alternative

((AB-SO)/AB)*100--Contact

The 90th, 50th and 10th percentile for each was:

Isolated Power--0.238, 0.150, and 0.080

Discipline 1-- 4.22, 2.16, 1.17

Discipline 2-- 102.1, 65.0, 30.4

Contact-- 89.1, 81.8, 72.5

If you really want to be keen you can use 95th and 5th percentile as the outliers. This would mean a guy like david Ortiz would have elite power (maybe a 95+),

Guys like Bonds, Juan Pierre, Brian Giles and Todd Helton have the lowest SO/BB ratio.

Guys like Juan Pierre, David Eckstein and Sean casey have high contacts.

I can send this file to people if they are interested.

The next thing to keep in mind is what parameters MVP sets up for each stat. If you play with MVP edit (and remove all the pitchers), and look at MLB rating only, you should find that 60 or less represents the bottom 10% on the majors for contact while 50 represnts the bottom 10% for power. This means that someone with minor power like Ichiro should have power in the 40s (which would give him about 8-14 homers in a year which is about right).

Some guys are hard to get right though and you may have to play around. Jim Edmonds strikes out alot but has a high batting average so is he a high or low contact hitter? IN real life he is not really true contact hitter (because he only puts it in play about 70-something % of the time). Yet if you gave him low contact stats on MVP I don't think youll get his stats right if you simmed.

There are other hard guys to get right including Thome. Maybe we just have to accept there are outliers for any stat and some improvisation is required.

[DarkEnigma510]

Go watch Ichiro at batting practice. He can hit the ball as hard and the best but he is a leadoff hitter whose main goal is to get on base. What sort of leadoff hitter hits home runs? Not on my team.

Ichiro has too much default power in the game. IMO his power stats should be lowered to the 40s. Also see Sean Burroughs with the padres..I think his power stats are too high also. They are both contact hitters..not homerun hitters.

[DarkEnigma510]

It doesn't seem like batting avg. is a very good measure of contact. After all, hitting the ball and hitting them "where they ain't" are two different things. I'd second Enigma's suggestion: ((AB-K) / AB)...seems like a more true contact rating.....again, you'll need some way to scale all the players such that they fit EA's chosen highs and lows on the 100 point scale.

While I certainly agree with your reasoning- I am positive that the game doesn't operate in that fashion; i.e., if you order contact rating by the above metric- it will work for guys like Bonds and Pujols- but not so well for guys like Ramirez and Ortiz. So, unless you want guys like Pierzynski, , Polanco, Eckstein, Alfonzo, Hatteberg, Hillenbrand, Vizquel and Deivi Cruz leading the league in hitting (all in the 20 or 30 in k/ab)- K/AB won't work. If you're not sure- try reordering contact ratings in this fashion and simming a season or two. The plain and simple truth is EA doesn't account for enough variables to accurately represent baseball statistics. Plate discipline is a great example- but by itself it can't accurately simulate guys like Bellhorn, Thome and Abreu who strike out AND walk over a hundred times...Makes things difficult

Just from playing around I don't think this is how the game works either. Look at Edmonds and Thome who both have lows stats by that metric but in the game are rated at contact much higher. I can only guess they have taken BA as a factor also.

[ZMan]

DarkEnigma, great, great stuff.

One thing we haven't discussed any of is pitching ratings. Though a lot of scouting report fodder goes into determining ratings, I wonder how stats might be converted into sensible movement or control ratings..

I hope we can continue this topic's discussion, perhaps merging it with the stuff that GForce is talking about with his new thread..

This would all be very helpful for developing formulas for a possible new MVPEdit interface by RGlass.

[DarkEnigma510]

Where is Geforce's Thread?

[GForce22]

What you need to remember is that EA doesn't look at things sabermetrically...if you develop ratings based on something too far from what EA thinks its ratings mean, all you'll do is frustrate yourself.

That said, I also don't think EA thinks through their ratings well at all. They don't look at how power and contact work together, for example (a big reason guys with little power but good contact always hit too many homers in this game)...

The other problem is, and I noted this last year...the game's ratings are not designed for both played game and sim functionality. For example, a played game you have your discipline rating and that's fine...but what REALLY matters in the in-game are the take/chase/swing ratings uncovered in MVPEdit. I uncovered this last year and basically had to redo all of these ratings to get hitters in played games to perform accurately from a discipline standpoint.

Best thing to do, IMO, if you want to really do the rosters right...

-- Globally edit all pitchers to the same ratings, say 50s across the board. This eliminates a variable.

-- Globally edit each team to 50s across the board as well

-- Modify individual teams, every hitter to a certain rating (i.e., all Mets to 80 contact, 50 power, all Yankees to 70 contact, 60 power, etc) and go down the line. I

-- Sim and see the results you get and it will tell you what the ratings really control, how they work together, etc.

Inverse everything and do the same for pitchers.

I did this on a smaller scale last year to get me in the ballpark for TotalMinors, but ideally, it would be as detailed as above. This lets you develop a ratings chart, what means what.

You can do whatever formulas you want to spit out stats, but if your conversions are off for whatever reason, all you've really done is waste your time.

Guys want rosters immediately...that's nice. If they want them done right they'll be willing to wait.

Good luck.

GH

[plez]

Gforce,

i have edited all batters to 60's for contact and power andall pitchers to 5 for movement and 71 for control for the purpose of seeing what th e ballpark effects are (dh for both leagues)

how many times do you think i should run the season?

also, i can't seem to find 'team totals' for stats which means i have to add stuff by hand, any ideas?

Plez

[JohnE66]

Good stuff and topics. One other thing I don't see noted though is the batter tendencies of taking or chasing pitches. Here is where a batter with high average and lots of K's can be made or one with high contact and few strikeouts. If a high contact hitter has a high take pitches and low rating at chasing his walks should increase. If he takes AND chases he should have increases in both K's and BB's.

I have tested this year (busy month 8O ),but last year this did have an effect.

I am not saying it will be perfect, but it helps.

[uni91]

Gforce,

i have edited all batters to 60's for contact and power andall pitchers to 5 for movement and 71 for control for the purpose of seeing what th e ballpark effects are (dh for both leagues)

how many times do you think i should run the season?

also, i can't seem to find 'team totals' for stats which means i have to add stuff by hand, any ideas?

Plez

In the standings screen you can have access to some team stats, but it is not a very exhaustive list.

[yukon2k]

I am using formulas that I found in another post on this forum....cant remember...but ill post later if I find it. Anyway, I use those formulas in excel, which makes it real easy....just input the required stats from the player splits, and voila! Depending on the situation, I use a weighted average by averaging last years splits with the players splits from the last 3 years. Last years stats end up having a 50% weight on the rating while the other 2 years each have a 25% weight each. But for younger players, you have to make your own decision (ex. In limited AB's, J. Maur killed RHP last year...like .420, it doesn't mean he deserves a 99 in contact versus RHP, he needs to do it longer...so you have to make your own decision on those) If you want the spreadsheet I use...ill e-mail it to ya.

send me the formula too. i couldn't find anything like that when i searched the forum. thanks in advance yukon2k@adelphia.net