Revisiting Avoid K's and Contact

[Luis_Rivera]

Based on information gleaned from a couple minor studies on Avoid K's done by Sandman7103 and gmo, it's come to light that if possible... how these ratings effect the game should be reconsidered.

Rather than rehash everything... you can start with post #9 in this thread: Avoid K's

After looking that over, you can comment here.

Basically my feelings on the matter:

Originally Posted by Luis_Rivera

It would seem to me that if Markus would want to create the most realistic DIPS engine possible, then the contact rating would affect BABIP rather than batting average.

Although, the thing with this is that you'd have to have a huge number of players near the middle because there really aren't too many Derek Jeters out there that can consistently put up the .350 BABIP.

Doing this would also give much, much more meaning to the avoid K's rating... as well as a much more realistic meaning, because as it stands, if it were to be realistic then there would have to be some sort of algorithm that intertwined avoid K's and contact.

[Luis_Rivera]

Also, this is a fairly important point that can help drive the importance of the issue home:

Originally Posted by Luis_Rivera

What's funny is that as it stands, you're better off having a lower avoid K's rating since it doesn't affect batting average. This is not an exaggeration... it really would help as your team would hit into less double plays with the benefit of less balls in play.

[Luis_Rivera]

Nevermind... it seems that this one has been solved.

[Davey Eckstein]

Originally Posted by Luis_Rivera

Nevermind... it seems that this one has been solved.

How so?

[Luis_Rivera]

Originally Posted by Davey Eckstein

How so?

Well, not so much completely.

It doesn't appear to be nearly as bad as I thought... it's still a very workable system, but I suppose my first post in this thread would be still a valid request.

So, uh... let's go with that.

[Luis_Rivera]

Continuing with this from the other thread:

Originally Posted by mrbill

But, is it in any way realistic that a group of hitters with TERRIBLE avoiding Ks has a .473 BABIP? They don't make any better contact than guys with 95 for avoiding Ks, but those guys have a much lower BABIP.

That to me, is a really botched up implementation of DIPS

It's not really anywhere near realistic for any group of hitters to have a .473 BABIP.

This is why I'd like to make Avoid K and Contact independent of one another and have contact just effect BABIP.

[Fallschirmjager]

But doesn't 'avoiding' Ks have an effect on BABIP in real life?

Lets say you have too players.
One who strikes out 100 times and one who does 0 times

The leagues BABIP is .300

It is pretty rough but could not you expect the 2nd player to get 33 more hits over the course of the season by simply putting the ball in play?

[Elendil]

Originally Posted by Fallschirmjager

But doesn't 'avoiding' Ks have an effect on BABIP in real life?

Lets say you have too players.
One who strikes out 100 times and one who does 0 times

The leagues BABIP is .300

It is pretty rough but could not you expect the 2nd player to get 33 more hits over the course of the season by simply putting the ball in play?

Yes, more hits and a higher BA, but not necessarily a higher BABIP. Players with higher BABIP are better able to "place" the ball where the fielders aren't, when they do put the ball in play.

[Fallschirmjager]

Originally Posted by Elendil

Yes, more hits and a higher BA, but not necessarily a higher BABIP. Players with higher BABIP are better able to "place" the ball where the fielders aren't, when they do put the ball in play.

Hmm


Im starting to see where Luis is coming from.

Contact would be basicly 'bat control'
How good a player is at producing a hit when he makes contact.

Avoid Ks would simply be a function of how often a player makes contact
Thus a player who strikes out more make contact less often and thus puts fewer balls into player and produces fewer hits.

Am I getting it right?

[Elendil]

Originally Posted by Fallschirmjager

Hmm


Im starting to see where Luis is coming from.

Contact would be basicly 'bat control'
How good a player is at producing a hit when he makes contact.

Avoid Ks would simply be a function of how often a player makes contact
Thus a player who strikes out more make contact less often and thus puts fewer balls into player and produces fewer hits.

Am I getting it right?

I think so, at least that's my POV as well.

[Luis_Rivera]

Originally Posted by Fallschirmjager

Avoid Ks would simply be a function of how often a player makes contact
Thus a player who strikes out more make contact less often and thus puts fewer balls into player and produces fewer hits.

Am I getting it right?

Yessir!

[mrbill]

Yes, that's what I hope to see in the future, and the same thread referenced in the 2nd post in this thread has A LOT more discussion on the topic.

What makes sense to me:

Contact ~= quality of contact -> affecting BABIP for the hitter

Avoiding Ks = quantity of contact -> affecting average for the hitter (indirectly, by affecting # of balls in play)

[Luis_Rivera]

Originally Posted by mrbill

Yes, that's what I hope to see in the future, and the same thread referenced in the 2nd post in this thread has A LOT more discussion on the topic.

What makes sense to me:

Contact ~= quality of contact -> affecting BABIP for the hitter

Avoiding Ks = quantity of contact -> affecting average for the hitter (indirectly, by affecting # of balls in play)

This is a beautiful way of boiling down my convoluted request. Thanks, MrBill!

[gmo]

Originally Posted by mrbill

What makes sense to me:

Contact ~= quality of contact -> affecting BABIP for the hitter

Avoiding Ks = quantity of contact -> affecting average for the hitter (indirectly, by affecting # of balls in play)

I rambled around this a lot in a post I just put up in the OOTP6 forum thread, but I believe my thinking and understanding of the discussion on improvements is crystalizing a bit, so I wanted to come here.

I have said I think that complications like most notably perhaps OOTP6 using both the old and the new game engines but still being expected to produce not wildly different results and allowing for import of OOTP5 leagues could be part of why some of the ratings in the DIPS engine do not translate into the results expected.

So for the future, it seems people that have voiced an opinion want something like:
*AvoidK determines strikeout rate ~(K/PA)
*Eye determines walk rate ~(BB/PA)
*Power determines HR rate ~(HR/PA)
...with those all occurring initially in the DIPS engine. Those ratings help decide whether a batter gets a K, BB, or HR, but if not, then the ball is in play, and...
*Contact determines BABIP ~(H-HR)/(AB-HR-K)
*Gap determines, well, not sure, maybe how often a hit is a 2B or 3B - probably not important for this discussion

As I understand it, previously in the game (up until the DIPS engine) Contact was all about the ability to get a hit, basically determining batting average. In a new system like this Contact would be generally very different if its new job was affecting BABIP. The higher BABIP hitters are often the batters with more power and who strikeout more but hit the ball harder when they do make contact. That is certainly not always the case though, so I have some concern about there being enough parameters in that scheme to get all the player types you ought to get, though maybe that is mostly being hesitant to see the meaning of a rating changed so much.

[rcbuss]

Originally Posted by Fallschirmjager

Avoid Ks would simply be a function of how often a player makes contact
Thus a player who strikes out more make contact less often and thus puts fewer balls into player and produces fewer hits.

I think that we have to be really careful with how we define avoidK.

In looking at the list of tops in BABIP, I noticed that there was quite a bit of variation in K rate (I used AB, as BB and HBP weren't readily available. I don't think they would've changes the figures all that much.)

Originally Posted by gmo

Top 25 BABIP careers, minimum 100 AB
(from thru2004 Lahman database)

name                 AB  Hits   HR   SO  BABIP K/AB
Cobb, Ty          11434  4189  117  357  .3715 .0312
Barnes, Ross       2392   859    6   65  .3675 .0272
Hornsby, Rogers    8173  2930  301  679  .3655 .0831
Suzuki, Ichiro     2722   924   37  247  .3638 .0907
Overbay, Lyle       845   246   20  201  .3622 .2379
Carew, Rod         9315  3053   92 1028  .3613 .1104
Jeter, Derek       5513  1734  150  972  .3607 .1763
Jackson, Joe       4981  1772   54  158  .3602 .0317
Abreu, Bobby       4140  1264  166  909  .3582 .2196
Bay, Jason          498   141   30  158  .3581 .3173
Michaels, Jason     519   147   17  137  .3562 .2640
Darr, Mike          542   148    5  135  .3557 .2491
Rios, Alexis        426   122    1   84  .3548 .1972
Meyerle, Levi      1443   513   10   14  .3545 .0097
Mouton, Lyle        805   225   22  209  .3537 .2596
Jefferson, Reggie  2123   637   72  451  .3531 .2124
Hamilton, Billy    6268  2158   40  218  .3524 .0348
Roomes, Rolando     406   103    9  130  .3521 .3202
Heilmann, Harry    7787  2660  183  550  .3511 .0706
McVey, Cal         2513   869   11   51  .3501 .0203
Terry, Bill        6428  2193  154  449  .3500 .0699
LeFlore, Ron       4458  1283   59  888  .3486 .1992
Delahanty, Ed      7505  2596  101  244  .3485 .0325
Figgins, Chone      829   244    5  137  .3479 .1653
Boggs, Wade        9180  3010  118  745  .3477 .0812

It seems that a global AvoidK rating would vary widely among the game's best in BABIP. It would make a lot more sense if AvoidK equalled AvoidK-swinging plus AvoidK-looking.

AvoidK-swinging would likely have a good correlation with Contact (as evidenced by the +20/-20 range of Contact-AvoidK depicted in a related thread).

AvoidK-looking would be a combination of "Patience" (i.e., #pitches seen/PA) and "PitchRecognition" (i.e., player skill in discerning balls from strikes on pitches that he decides to take).

These changes, along with linking Contact to BABIP, would produce realistic ranges of BABIP, while at the same time, allowing for all variations of BB- and K-rate for hitters.

As Luis_Rivera said earlier in the related thread, these ratings would be related to each other in many ways (e.g., Patience to BB and K, Contact to BABIP and AvoidK-swinging).


Other thoughts:

• optional global league settings for pitches/PA (e.g., set to low for the early 1900s - low BB, low K, more CG)
• the results of the combinations of these ratings would be even more apparent if the pitch-by-pitch feature were being used

[mlyons]

Including deadball-era players in a discussion of BABIP probably isn't a good idea, given the difference in fielding then.

[The Man They Call Jayne]

Originally Posted by Luis_Rivera

This is a beautiful way of boiling down my convoluted request. Thanks, MrBill!

Here's the part that confused me Luis - by definition, don't these two ratings have to be linked together, so to speak ? Take the Jose Hernandez's of the world - we would agree that they make contact less regularly, but when they do, the contact tends to be more productive. I agree with the baseline you propose - I'm just worried that the engine will not follow the rules when generating players (like the fact that OOTP generates power 2b in the same ratio as OF sometimes).

Hmm - looking at THT's 2006 statistics (great book by the way - the annual reminds me of Prospectus as it used to be, with less of the snark and more of the substance) - it seems the highest BABIP's are varied amongst different types- Young, Jeter, and A-rod are the 3 highest. I wonder if there is some way to access an OPSBIP or SLGBIP- I may be able to test this theory better with regards to that.

http://www.hardballtimes.com/thtstat...&qualified=Yes

[Luis_Rivera]

Originally Posted by The Man They Call Jayne

Here's the part that confused me Luis - by definition, don't these two ratings have to be linked together, so to speak ? Take the Jose Hernandez's of the world - we would agree that they make contact less regularly, but when they do, the contact tends to be more productive.

Yes, that's exaclty my point. Jose Hernandez has a relatively high BABIP (career .326), but he strikes out 30% of the time. So, he'd have a higher contact rating (to account for the high BABIP) and a low avoid K's (to go with the high amount of K's).

I wonder if there is some way to access an OPSBIP or SLGBIP- I may be able to test this theory better with regards to that.

It would essentially be adding doubles and triples to BABIP, so I don't see much reason to do that.

[The Man They Call Jayne]

Originally Posted by Luis_Rivera

Yes, that's exaclty my point. Jose Hernandez has a relatively high BABIP (career .326), but he strikes out 30% of the time. So, he'd have a higher contact rating (to account for the high BABIP) and a low avoid K's (to go with the high amount of K's).

Fair enough - the OOTP engine's history in recognizing paired ratings is not particularly impressive, and that's why I'm somewhat tentative about it.

It would essentially be adding doubles and triples to BABIP, so I don't see much reason to do that.

I wasn't particularly clear- I guess I'm not looking for a rating based on BIP, but rather on contact - thus it would differ by including HR's (strictly speaking, I'd be curious if all the BIP metrics include inside the park HR's). Its not particularly relevant here.

[Luis_Rivera]

Originally Posted by The Man They Call Jayne

Fair enough - the OOTP engine's history in recognizing paired ratings is not particularly impressive, and that's why I'm somewhat tentative about it.

Yea, well I guess that would be another issue... correlating BB/HR with a lower avoid K rating. Avoid K and Contact wouldn't have to be paired though... a higher K% doesn't have to mean a lower BABIP, and vice versa.

I wasn't particularly clear- I guess I'm not looking for a rating based on BIP, but rather on contact - thus it would differ by including HR's (strictly speaking, I'd be curious if all the BIP metrics include inside the park HR's). Its not particularly relevant here.

Ah... ok, gotcha. I think you could do that pretty easily yourself by just removing K's from the equation. I'm not sure I see the relevance all that much though as strikeouts correlate with power so well.

As for the BIP metrics including inside the parkers... I would say with a large degree of certainty that they do not. However, I would imagine that they are so far and few between at this point that they don't have a material affect in the data.