MLB THE SHOW - Ratings Renaissance

[jgb3]

One of the most exciting parts of increasingly available Statcast™ data is the fact that it gives the O.S. community the ability to make players more lifelike than they ever have been. Player ratings are a hugely important part of any sports video game - they govern the simulation engine and dictate outcomes on virtually every single play or pitch.

My hopes for this thread is that it will serve as a breeding grounds for ideas on how to apply Statcast data to generate incredibly lifelike player ratings. It is inevitable that this data will one day be the framework for how player ratings are generated. There is no reason the O.S. community can't be the pioneers who come up with the foundations and formulas for these ratings.

For example, a future post in this thread might look something like this:

ARM STRENGTH rating proposal:

Amongst all outfielders that registered a competitive throw in 2017, Yankees OF Aaron Hicks registered in with the fastest average competitive throw. Based on the same data, we also know that Oakland OF Khris Davis has the worst arm strength. This February article by MLB Statcast writer Mike Petriello provides us with some of the first publicly available data on player arm strength, and gives us a top 5 list of current MLB outfielders:

http://m.mlb.com/news/article/215475982/twins-project-for-strongest-2017-outfield-arms/

Quote:

Top five outfield arms on "competitive throws" in mph in 2016 99.4 -- Hicks, NYY 97.0 -- Starling Marte, PIT 96.8 -- Santana, MIN 95.8 -- Carlos Gomez, HOU / TEX 95.7 -- Jake Marisnick, HOU

Arm Strength Rating: On a scale of 0 - 99, a player is graded based on the velocity of his average *Competitive* throw.

We can now map these values to ratings - once we get the full pool of MLB starters plotted out onto a graph, we can figure out a clever way to group these players into distinct rating buckets: Players over the 100 mph get a 99 arm strength rating. Players over the 97.5 mark get a 90+ arm strength rating, players over the 96 mark get am 85+, etc. etc.

Arm Strength data isn't fully available to the public yet, but in the next few years we will have access to data on virtually every competitive throw for every single player.

Please feel to post your ideas for other ratings categories below!

[jgb3]

Catcher Arm Strength:

Pop time is becoming an increasingly popular statistic. It is used to quantify a catcher's ability to throw a stealing runner out.

Pop time measures the time it takes for a catcher to receive a pitch and and throw the ball to second base.

This February, 2017 article by MLB Statcast writer Mike Petriello gives us real data on which catchers are the best at getting the ball to second base on steal attempts:

Quote:

[L]ooking ahead to 2017, which collection of catchers look like they'll be the quickest at getting rid of the ball? Would you believe it's … the San Diego Padres? Thanks to Austin Hedges and Christian Bethancourt, the 2017 Padres look to have elite pop times. Hedges was drafted by San Diego in 2011, more for his glove than his bat... For our sample, 2.01 seconds is the Major League average, and while Hedges barely played in 2016, he did manage a pair of 1.86-second throws. Bethancourt, meanwhile, has such a cannon that despite catching only 41 games in 2016, he had three of the five hardest throws by any catcher on a steal attempt, topping out at 89.3 mph.. He averaged 1.88 seconds in pop time in both 2015 and '16, putting him squarely above average.

Once this data becomes fully available, we can plot catcher pop times onto a graph and then come up with a formula for determining what each pop time should represent, in ratings form.

Any ideas on what this formula could be?

[Caulfield]

Infielder arm strength and speed IMO are the two most difficult for me to get a consistent, accurate read on. I too have high hopes for statcasting in the coming months and years to follow to get a closer feel in attribute ratings.

I see a lot of talk of using BABIP for Contact but I much prefer just using your regular old run of the mill batting average.

Power will always and forever be AB/HR for me

Vision and K/9 I use K%
Discipline and BB/9 I use BB%

I'm one of those guys who dont really believe in clutch and if it does exist its so minimal to be worth trying to replicate. For this reason I use OPS and WHIP to arrive at clutch numbers. I know thats unorthodox but I only want players that I consider really good to be clutch.

Durability, I use games. If a player isnt playing for whatever reason, injury or DNPMD, I dont want him playing too much in my games either.

Steals I use SB% and aggression I use SBA/SBO (attempts/opportunities.

Bunts I use SH/AB and Drag (which also is hard to replicate), I use a combination of contact, bunt and steals. Its not perfect but it is what is is.

Fielding is a hot button issue for me. I dont really care to use uzr or any of the other sabermetrics, I prefer to use assist/innings to get reaction/range and errors/innings to get fielding. I know some folks dont want to use errors because that seemingly is just the official scorers opinion. I can live with the O.S. numbers for better or worse, as unpopular as that may be. Arm accuracy I use throwing errors/innings for infield and for outfield I use assists/innings.

When I'm trying to create an older pitcher from say the 70's or 80's, to come up with break/movement I use FIP. Thats another one of my unconventional methods but it works well enough for me. Control for those older era pitchers, I use SO/BB for their first pitch and WP/innings for his 2nd pitch. Whatever the difference in control is between those first 2 pitches (say he has a 75 for pitch 1 and a 68 for pitch 2)then I take that difference (in this case 7) and his 3rd pitch control would be 61.

For H/9 I prefer to use batting average against as opposed to H/9 and for HR/9 I'd rather use HR%, but thats just my personal preference.

I am constantly tinkering with these formulas but probably 10 years from now as statcasting evolves, I will eventually settle on one set of formulas and who knows, I may come around to UZR for fielding as well lol

[jgb3]

Quote: Originally Posted by Caulfield

Infielder arm strength and speed IMO are the two most difficult for me to get a consistent, accurate read on. I too have high hopes for statcasting in the coming months and years to follow to get a closer feel in attribute ratings.

Good news - Statcast does have data for both. For infield arm strength, they'll be able to leverage the top 5% of throws, a.k.a competitive throws. This is the same metric for outfield arm strength.

For speed, it's as simple as the average of a player's top 5% home to first times. The problem with using top speed alone is that some players are able to reach a faster top speed than others (say Cespedes, who can reach nearly 22 mph I believe) but take a bit longer to reach that speed. Ultimately, an acceleration rating needs to be added to the game. Point being, no player is busting it down the line every time, but harvesting home to first times should provide a solid barometer for how fast a player is going when he is trying.

Quote: Originally Posted by Caulfield

I see a lot of talk of using BABIP for Contact but I much prefer just using your regular old run of the mill batting average.

Not a great idea, if you ask me. Babip is mostly fueled by exit velocity and a player's speed. Guys like Starling Marte (20% K-rate) post elite BABIP not because they are proportionately great at making contact, but because they are either extremely fast down the line, or can make really meaningful contact.

This is a really complex issue, as there are so many inputs for true contact. Batting average is obviously a healthy representation, but at the end of the day, it's deeper than that. Jose Iglesias batted just .255 last year despite the top z-contact and o-contact rate combination in the MLB. That means that when he swung, he always made contact. He also had the second fewest swinging strikes. So in the traditional definition of contact, he is a 99. But that cannot be, as he is not a career .320 hitter. So obviously, something needs to change. Contact cannot continue to be named contact, as contact means ability to put wood on ball. This is what "vision" should be called. Contact as we know it should probably be completely abolished, as BABIP is a product of a player's EV and speed and launch angle. All three of those latter ratings should have re-vamped ratings.

I'll address the rest of this post later - gotta run now. But I think the general theme of this thread will be that SCEA has the basics down pretty well... A proper overhaul will have to be from the ground up. Current ratings system is not really compatible with sabermetrics.

[Mitchrapp]

This is awesome. Great thread. Thanks for starting it..

[garyjr33]

Quote: Originally Posted by jgb3

One of the most exciting parts of increasingly available Statcast™ data is the fact that it gives the O.S. community the ability to make players more lifelike than they ever have been. Player ratings are a hugely important part of any sports video game - they govern the simulation engine and dictate outcomes on virtually every single play or pitch. My hopes for this thread is that it will serve as a breeding grounds for ideas on how to apply Statcast data to generate incredibly lifelike player ratings. It is inevitable that this data will one day be the framework for how player ratings are generated. There is no reason the O.S. community can't be the pioneers who come up with the foundations and formulas for these ratings. For example, a future post in this thread might look something like this: ARM STRENGTH rating proposal: Amongst all outfielders that registered a competitive throw in 2017, Yankees OF Aaron Hicks registered in with the fastest average competitive throw. Based on the same data, we also know that Oakland OF Khris Davis has the worst arm strength. This February article by MLB Statcast writer Mike Petriello provides us with some of the first publicly available data on player arm strength, and gives us a top 5 list of current MLB outfielders: http://m.mlb.com/news/article/215475982/twins-project-for-strongest-2017-outfield-arms/ Arm Strength Rating: On a scale of 0 - 99, a player is graded based on the velocity of his average *Competitive* throw. We can now map these values to ratings - once we get the full pool of MLB starters plotted out onto a graph, we can figure out a clever way to group these players into distinct rating buckets: Players over the 100 mph get a 99 arm strength rating. Players over the 97.5 mark get a 90+ arm strength rating, players over the 96 mark get am 85+, etc. etc. Arm Strength data isn't fully available to the public yet, but in the next few years we will have access to data on virtually every competitive throw for every single player. Please feel to post your ideas for other ratings categories below!

really cool idea...good job

[jgb3]

Today, BaseballSavant.com posted publicly available Catch Probability data on its website for the first time. Even thought it may not seem like it, this is a big milestone for statheads and for the MLB the Show community. This gives us the power to truly quantify how good a player is at tracking and fielding without solely going on error numbers and "the eye test."

https://baseballsavant.mlb.com/catch...n=50&year=2016

The Statcast team had hinted that this data would become available before the start of the 2017 season, and lo-and-behold, the data has been posted.

What does this data show?

Quote:

Catch Probability Likelihood an outfield catch is made based on “opportunity time” and “distance needed”. * Opportunity time: starts when the ball is released by the pitcher * Distance needed: shortest distance outfielder needs to make catch 5 Star Ranking Breakdown: 5 Star: 0 - 25% 4 Star: 26 - 50% 3 Star: 51 - 75% 2 Star: 76 - 90% 1 Star: 91 - 95%

REACTION AND ROUTE EFFICIENCY RATINGS:

Across the board, this data reveals a few things about a player's defensive capabilities - mostly his route efficiency, his reaction, and his speed.

4 and 5 star plays can positively impact a player's fielding, reaction and (newly proposed) route efficiency ratings. If a player demonstrates a consistent ability to make these low percentage plays, it will show in the data. We can then plot these outcomes on a graph and assign ratings based on a weighted score.

1 and 2 star playsshould negatively impact a player's reaction and route efficiency, and to a lesser extent, his fielding. A slow player that cannot get to a 1 or 2 star ball should not have his fielding rating penalized if he catches everything he gets his hands on. Speed and reaction and (newly proposed) route efficiency ratings should be the main drivers in a fielder not making these plays. We can plot these 1-2 star outcomes on a graph and get a good idea on which players are bad fielders.

The best solution would be to come up with a formula that assigns a certain amount of points per each successful play. For example, 1 point for each 1 star play, 2 points for each 2 star play, etc. It would also penalize players for not making plays. For example, a player would be penalized 5 points for not making a 1 star play, and so on.

We could then plot each player's "JGB3 Score" (lol) against a traditional fielding metric such as UZR (to make sure it holds water) and come up with a R value to make sure that the data appears to be correlated. This is to validate the data and insure it is giving us what we want.

All in all, there is a lot to be derived from this data. Players we thought were particularly good fielders (looking at you, Starling Marte) turn out to be not-so-great fielders that might be making themselves look better with highlight reel plays because they are taking bad routes.

Players like Mookie Betts, who make most all routine plays but not an elite percentage of low percentage plays, should probably have solid fielding ratings, but not elite reaction and route efficiency ratings.

Post your thoughts below.

[Mitchrapp]

Catch probability data is sweet. It's nice to see players data from 2016. I could browse at that stat page all day.. Some players had better 4 star plays than 3 star plays, very interesting.