Offense vs. Defense article

[maligned]

Bbrool,
If you go back through our discussion, I was actually questioning NERP, because I didn't understand how it was calculating this earning of extra plate appearances or lack of out consumption.
If I understand your conclusion correctly, you're outlining the way NERP adjusts value based on lack of outs consumed (not just RC).

What I'm saying is this (Please help me where I'm misunderstanding NERP!):

I think NERP is not comprehensive in its ability to predict how a player will function in future situations. In my example on the previous page, Player cards A and B would have the following projected results against an average pitcher (.315obp/.415slg allowed).

Player A: .398obp/.481slg
Player B: .338obp/.536slg

See exact details of player card examples on previous page.

If these stats play out over a season, my contention is that the two will produce the same results in a given static set of plate appearances, but will not help their teams produce the same amount of runs because Player A will consume far fewer outs and will thus generate more plate appearances for his team (what I hear you trying to explain is that NERP considers this fact and adjusts the value; I was trying to say I don't think it does--please clarify the final points of your last post!).

Please look back at the discussion of Dean and me on the previous page and help me understand what I'm missing. Thanks so much! I'm still trying to understand how this all works.

[Munich_Man]

My brain hurts. You guys are awesome.

[childsmwc]

Ok let me see if I can lay this out in a format that is easy to understand:

Ok first lets put out the RC formula we will be using for this example:

BB- .32 runs
Hit- .16 runs
TB- .32 runs
Out (.10) runs

We will not complicate this example with SB's, CS, DP, or the finer rates that I actually use, these numbers are sufficient for this example.

Second lets establish our base team:

PA's- 6,552
AB's- 5,904
BB- 648
Hits- 1,557
TB's- 2,502
Outs- 4,347 (this will be the constant for this example a team has 4,347 outs over a 162 game schedule)
Avg. .264
OBP- .337
Slg- .424

Total Runs Scored using our formula above = 822.4 Runs a season

If all 9 players on that team our equal, then the average player has a line of:

PA's- 728
AB's- 656
BB- 72
Hits- 173
TB's- 278
Outs- 483
Avg. .264
OBP- .337
Slg- .424
Runs contributed- 91.4

While the other 8 players total:

PA's- 5,824
AB's- 5,248
BB- 576
Hits- 1,384
TB's- 2,224
Outs- 3,864
Avg. .264
OBP- .337
Slg- .424
Runs contributed- 731

Ok that is all of our team data.

[childsmwc]

Now lets compare the value of two different random players not on that team over an expected 728 Plate Appearances:

Player A has the following stat line:

PA's- 728
AB's- 618
BB- 110
Hits- 180
TB's- 291
Outs- 438
Avg. .291
OBP- .398
Slg- .471
Runs contributed- 113.3

Player B has the following stat line:

PA's- 728
AB's- 673
BB- 55
Hits- 190
TB's- 355
Outs- 483
Avg. .282
OBP- .337
Slg- .527
Runs contributed- 113.3

So at this point I think we can all agree that these two players are equal and contribute the same runs to a team if both players get 728 PA's.

We have just completed step #1 and have determined their RC value, but lets see what happens when we put these individuals on our base team in place of the average player.

[childsmwc]

Player A Team Totals:

PA's- 6,552
AB's- 5,866
BB- 686
Hits- 1,564
TB's- 2,515
Outs- 4,302
Avg. .267
OBP- .343
Slg- .429
Runs contributed- 844.3

Player B Team Totals:

PA's- 6,552
AB's- 5,921
BB- 631
Hits- 1,574
TB's- 2,579
Outs- 4,347
Avg. .266
OBP- .337
Slg- .436
Runs contributed- 844.3

So our linear RC model is still working great, I replaced a guys worth 91.4 runs in both examples with guys worth 113.3 runs and both team totals increased by 21.9 runs.

[childsmwc]

However, there is one problem with our team totals, Team A has only committed 4,302 outs, when we have established that a team will commit a total of 4,347 outs over 162 games. I have made the wrong assumption about the plate appearances for the other 8 hitters on the team.

Using my base hitter line of .264/.337/.424, I should have assumed these additional PA's:

PA's- 68
AB's- 61
BB- 7
Hits- 16
TB's- 26
Outs- 45

So my adjusted Team A totals now look like:

PA's- 6,620
AB's- 5,927
BB- 693
Hits- 1,580
TB's- 2,541
Outs- 4,347
Avg. .267
OBP- .343
Slg- .429
Runs contributed- 853.0

So when I discuss replacement value of Player A, his value will be computed as his initial RC value of 113.3 +8.7 additional runs produced by the rest of the lineup due to addtional plate appearances. So from a replacement value standpoint Player A is worth 122 runs.

His RC value on this team is still 113.3, but when you take Player A and put him on a team the plate appearances for the other members of the team must change. In this case they get another 68 Plate appearances and score another 8.7 runs.

Again the linear RC values haven't changed in any of the examples, but what has changed is that in Player A's case, I made an erroneous assumption when I assumed the plate appearances for the other 8 individuals would remain constant with Player A on the team. His OBP above the team average created more Plate appearances and therefore by definition more runs as more events were created before reaching the team out total of 4,347.

So if I were assigning prices to individuals Player A should be given a run value of 122 for pricing purposes, because he creates more events in the team context (ie. PA's) than Player B, even though their individual contribution over 728 PA's is the same value.

I hope this has helped in some way,

Bbrool

[maligned]

EXACTLY!!

I have a spreadsheet with the exact same type of example using NERP as the calculator. I misunderstood your original post, Bbrool. What you just illustrated has been my point all along.
In my example of Player A and Player B from the previous page of posts, the two players have equal NERP values on their player cards, but Player A helps his real-life team score 8.4 more runs over a season (I was estimating 7.5 in a previous post).

NERP, ESPN's RC, Bill James' RC, or whatever can only work correctly to predict value when you consider outs as the base--not plate appearances. This is the base of real baseball games.

[childsmwc]

Yes I think the take away here is that for an RC model to work correctly you must have the correct input. OBP or more accurately the consumption of outs (this would include CS & DP) directly impacts plate appearances and therefore impacts the input going into the RC model.

For pricing purposes you must consider an additional value beyond the base RC value, for each hitters impact on the team's Plate appearances.

maligned- I sent you some follow up information to your private message. I sent it a few days ago so can't quite remember what I said, but I know I sent you a follow up directly to your private box.

Bbrool

[maligned]

Ok, that was helpful info in my inbox. I had a couple follow-up thoughts I shared with you there.

Bottom line: real-life outcomes are realized in a string of interdependent results based on a standard number of outs--not plate appearances. The only way to evaluate true value is to figure out how a given player will contribute offense within a team concept until his team reaches the common standard number of outs. Raw data must be considered in its accrual per out, not per plate appearance.