Weighing Batting Ratings

[Argonaut]

Originally Posted by Mizzery

I ignore AvK in my evaluation, since it appears to only impact how an out is made, not whether or not an out is made. If someone has data to the contrary, please share.

Since most high POW hitters have high K rates, it would seem counterintuitive to try and maximize both POW and AvK, especially because POW is a critical component of SLG and consequently OPS.

In a normal league with a regular standard deviation of talent, like the MLB 2019 start date, you'd be mostly right. Allenciox has data that suggests differently in Perfect Team.

We know that BABIP (hidden), POW, and AvK make up the CON rating -- you can tell this by going into commissioner mode in OOTP and editing a player. But in high levels of Perfect Team, the pitchers will have outsized STU and MOV. So much so that the effect of POW on a player's CON is overstated, and the effect of AvK on a player's CON is understated.

My one caution about the AvK rating is to be wary of players with high AvK and low Speed. They'll be grounding into double plays a lot. I'm looking at you, Record Breaker Tony Gwynn.

[waittilnextyear]

Originally Posted by Argonaut

In a normal league with a regular standard deviation of talent, like the MLB 2019 start date, you'd be mostly right. Allenciox has data that suggests differently in Perfect Team.

We know that BABIP (hidden), POW, and AvK make up the CON rating -- you can tell this by going into commissioner mode in OOTP and editing a player. But in high levels of Perfect Team, the pitchers will have outsized STU and MOV. So much so that the effect of POW on a player's CON is overstated, and the effect of AvK on a player's CON is understated.

My one caution about the AvK rating is to be wary of players with high AvK and low Speed. They'll be grounding into double plays a lot. I'm looking at you, Record Breaker Tony Gwynn.

Since the BABIP rating is hidden, I've resorted to loading up the main game to import in different historical batters in their respective years I want to know more about. Taking a little peak at the editor. Not sure now accurate this is for the PT mode, but I'm not sure they'd have a completely different set of BABIP ratings for it either.

[Josquin]

Originally Posted by waittilnextyear

Since the BABIP rating is hidden, I've resorted to loading up the main game to import in different historical batters in their respective years I want to know more about. Taking a little peak at the editor. Not sure now accurate this is for the PT mode, but I'm not sure they'd have a completely different set of BABIP ratings for it either.

The BABIP ratings for the initial set of cards were released in the spreadsheet that the devs released at the beginning of the season. I've played around with the editor as well to try to figure out the relationship between CON, POW, AVOIDK and BABIP, but there doesn't appear to be a simple linear relationship. It is more like several linear functions glued together for different ranges of values.

In any case, based on my analysis of perfect league data, the number of hits (1B/2B/3B only, excluding HR) per 550 AB on average is:

H/550 = 0.83 * BABIP + 0.63 * AVOIDK - 0.10 * POW + 38

So AVOIDK does have a large impact on the production of a hitter.

[Mizzery]

Originally Posted by Josquin

The BABIP ratings for the initial set of cards were released in the spreadsheet that the devs released at the beginning of the season. I've played around with the editor as well to try to figure out the relationship between CON, POW, AVOIDK and BABIP, but there doesn't appear to be a simple linear relationship. It is more like several linear functions glued together for different ranges of values.

In any case, based on my analysis of perfect league data, the number of hits (1B/2B/3B only, excluding HR) per 550 AB on average is:

H/550 = 0.83 * BABIP + 0.63 * AVOIDK - 0.10 * POW + 38

So AVOIDK does have a large impact on the production of a hitter.

In your calculation, how do you account for the league totals ultimately being a finite or fixed number? To keep runs scored at a reasonable rate, league totals at all levels are normalized, which means that even if you have an entire league of hitters who rate well with your formula, they can’t all perform at an optimized level.

I would think you would also have to factor in park effects and how they impact the distribution of the hits you are observing, the defensive factors of the players the hitters are playing against, the specific pitching factors such as High STF or low STF (how many Ks are likely to be generated), etc.

I have not run the math, so I am very prepared to be wrong, but I am not clear how BABIP or AvK can possibly be that predictive given all of the other factors that go into the determination of what is an out and what is a hit over thousands of at bats.

[Josquin]

Originally Posted by Mizzery

In your calculation, how do you account for the league totals ultimately being a finite or fixed number? To keep runs scored at a reasonable rate, league totals at all levels are normalized, which means that even if you have an entire league of hitters who rate well with your formula, they can’t all perform at an optimized level.

It doesn't really matter. Even if the absolute numbers are rescaled, the relative performance between players will remain the same.

I would think you would also have to factor in park effects and how they impact the distribution of the hits you are observing, the defensive factors of the players the hitters are playing against, the specific pitching factors such as High STF or low STF (how many Ks are likely to be generated), etc.

I did take park factors into account, but I also used a large enough sample (10 seasons at Perfect level) for the defensive variances and other factors to average out.

I have not run the math, so I am very prepared to be wrong, but I am not clear how BABIP or AvK can possibly be that predictive given all of the other factors that go into the determination of what is an out and what is a hit over thousands of at bats.

I stopped trying to make sense of the exact mechanics of how the engine calculates the outcomes, since it didn't conform to my expectations of how things should work either. But the data has the final say, so I don't question it anymore. Seems to be working so far for me.

[OMGPuppies]

Originally Posted by Josquin

The BABIP ratings for the initial set of cards were released in the spreadsheet that the devs released at the beginning of the season. I've played around with the editor as well to try to figure out the relationship between CON, POW, AVOIDK and BABIP, but there doesn't appear to be a simple linear relationship. It is more like several linear functions glued together for different ranges of values.

In any case, based on my analysis of perfect league data, the number of hits (1B/2B/3B only, excluding HR) per 550 AB on average is:

H/550 = 0.83 * BABIP + 0.63 * AVOIDK - 0.10 * POW + 38

So AVOIDK does have a large impact on the production of a hitter.

where do the parentheses go?

[Josquin]

Originally Posted by OMGPuppies

where do the parentheses go?

Standard precedence rules apply, so multiplication before addition.