Defense

Question regarding my remedial statistics abilities...

Harrison, at 6 NERPs to the dollar, should then be a MASSIVE value over Altuve, at 3 NERPs to the dollar, no? (Twice the value at a third of the price?) Gordon and Murphy get you equal NERPs per dollar, but for a small 56 cent premium, Gordon will covert about 160 out of 170 x-chances on defense. Does that pricing make sense to anybody?

FWIW, I’ve tried all the Wins Above Replacement and other esoteric linear weights formula experiments, and all I know is that they are often among my very worst performing teams. I don’t know a linear coefficient from a continuum transfunctioner, but I do understand what makes a Strat-O-Matic team go, and I don’t think this is it. But just because I can’t understand it doesn’t mean it doesn’t exist.

However, it feels a lot like reading the scientific principles of why a bumblebee can’t fly, while watching a bumblebee fly.

So, what am I missing?

Here is my 2 cents worth--and I acknowledge that J-Pav has a much better W/L ratio than I do so he's probably right in the long run.

I am taking these three players from the list and putting them in Yankee stadium. I know NERP means "New Estimated Run Production: and have read about it somewhere in the posts probably from the Sporting News days archives, but they are complicated and need spreadsheets with macros.

from lpezzeme's list:
NERP/Salary/Player
19.83 05.34 GORDON
19.66 04.78 MURPHY
19.46 03.67 GENNETT


I am calculating total OPS in card rolls only for right hand side here--all players are worse vs LH so should be somewhat the same ratios though I think Gennett would lose more vs LH.--I did not purchase a strategy guide this year but this should be close--the defensive numbers come from baseballthinkfactory.


Gordon -- Offensive OPS 85.15 no BPHR added, defense OPS allowed -9.1, I give -6 for weak power--total = 70.05
Murphy -- Offensive OPS 94.35 no BPHR added, defense OPS allowed -38.7, BP HR power in Yankee = 28.7 Total = 84.35
Gennett - Offensive OPS 82.95 no BPHR added, defense OPS allowed -29.8, BP HR power in Yankee = 38 Total = 91.15


I'll give Gordon +15 points for speed, stealing, and the extra DP he will turn

SO: in Yankee stadium at 2B

Gordon: 85.05 for 5.34M
Murphy: 84.35 for 4.78M
Gennett: 91.15 for 3.67M

Gordon holds that number in any stadium (no BP singles added--as all players are the same there)--the other two fall off in lower BPHR stadiums, Gordon surely best in SF and other low HR parks, but in Yankee and high BPHR parks--for the money spent, I want Gennett and the 1.7M extra for more somewhere else.

I did some back of the envelope calculations myself, and came away fairly convinced that SOM Online player pricing for offense is pretty consistent. However, I remain unconvinced about the pricing of defense.

NOBODY plays Murphy at 2B. Why? You need to know the number of runs his glove actually costs you. Let’s say he has 170 x-chance opportunities and only converts half. How many runs does 85 baserunners cost your team? How many baserunners are already on base with those x-chance fails, leading to additional runs against? How many more runs are created as a result of the ensuing extra at bats? What about the effects on pitcher tiredness?

I’m guessing the current pricing of defense is wildly underappreciated, if not wildly undervalued.

The crazy number of Champs teams I’ve seen this year sporting six or sevens 1s on defense is my first clue. Several of my own best teams using all 1s (more or less) is clue number two. The historical record of Champs teams using only 1s and 2s up the middle over 90% of the time kind of makes this a no brainer.

I wouldn't play Murphy at 2nd at 80M but might at 60M---I know 60M is a different beast than 80M.

Bour at 1B, Gennett at 2B, DeJong at SS, Devers at 3B --- a noDH team. This is probably worst fielding infield of any team I ever fielded--but it was interesting.

https://365.strat-o-matic.com/team/1486390

J-Pav,

I have to give credit to Marc Pelletier. Once I figured out in the 2015 season what he was doing with NERP, how to generate my own data [though not nearly as well or extensively as Marc did], and how to use it, I have fielded 11 teams [2015-2107 seasons]; nine of them made the playoffs and five won championships. Prior to that time, I had 27 teams [2009-2015 seasons]; 11 of them made the playoffs and two won championships. The NERP data gave me a framework for team construction – maximizing NERP within a given budget. Before that I didn’t have a clear strategy for building a team. I like data.

Gennett is an experiment. I use 1s only rarely, but I haven’t used a 3 at 2B since my early days in the game. I tallied up the runs in an inning resulting from Gennett’s missed X Out chances. Gennett has had 12 missed X Out chances out 39 total chances in 42 games, and eight of them have resulted in 11 runs in the inning of the missed chance, which is quiet alarming. One of the runs was in a win, and 10 of the runs were in seven losses [one run in five games, two runs in three games], which is very alarming. However, we were outscored by 38 runs in those seven losses, and in only one of the games could the runs resulting from an X chance be considered to have affected the outcome. On the other hand, he has 13 HRs, 29 RBIs, and three GWBIs. Clearly, I would prefer a better fielder with the same offensive abilities, but at $3.67 million that’s probably not going to happen. I could still drop him and add Harrison. I’ll see how it plays out!

Here are the defensive numbers used in pricing the 2B:

First here are some key assumptions:
2B will get 169 X-chances on his card for a full season (before factoring in injuries)
errors act and look like base hits, so a single and error = a double etc.
20% of DP chances will result in a double play
The average 2B produces 163.6 outs (DP count as 2 outs)

So a 1e13 (Gordon) should produce 11.7 singles, 1.6 doubles, 130.8 outs, 24.9 double plays. Using my RC formula that equates to -10.8 runs. Additionally, he produces 180.6 outs which is 17 outs more than the average 2B, which lowers the team runs allowed by another 3, so 13.8 runs is the final tally defensively.

Murphy at 4e11 produces 54.5 singles, 3.6 doubles, .2 triples,104.3 outs, 6.3 double plays which equates to negative (17.8 runs). Additionally Murphy allows more hitters in the lineup, and gives up another 8.2 runs, for a total (26.0)

Several other things happen in the final pricing:

1) defensive subs are assumed for poor defenders 3.5 run adjustment in Murphys case
2) since defensive rolls roll less than offensive rolls, they are "discouted", for 2B this is a 10% discount, so about another 2.7 favorable runs in Murphy's case, and about a .9 unfavorable adjustment to Gordon
3)DH pricing- There is a defensive pricing cap in place to limit the "value" one can find at DH. As it turns out Murphy hit the cap, so he is priced to DH, playing him in the field would provide less value than his price.

This is not a full step by step discussion, but some of the steps will tend to overprice poor defenders, unless you replicate the assumptions. Also the discount factor will tend to favor better defenders, however, while a run saved on defense is equal to a run scored on offense, the frequency of the occurence and therefore the standard deviation of outcomes is different when comparing hitting runs to defensive runs, which is why defensive runs are discounted. Corner outfielders are discounted 20%, as a further point of reference.

I have found over the years that if I remove or lessen those levers, the community complains 1 defenders are significantly over priced. So some of the pricing is not just a RC value issue, but one of community perception and expectations. Also the DH does make it hard to really "penalize" a poor defender because then you are creating the best valued DH in the game.

Mark,

Thanks for all the work you do and your great explanations to those of us who benefit from your efforts -

lpezzeme,

I sent you a pm -

J-Pav,

I think I sent you a pm. Let me know if you didn't get it.

Question regarding my remedial statistics abilities...

Harrison, at 6 NERPs to the dollar, should then be a MASSIVE value over Altuve, at 3 NERPs to the dollar, no? (Twice the value at a third of the price?)

So, what am I missing?

First off I don't see how I could possibly critique or respond to someone as great as J-Pav. I'm a rather nascent owner but I do have a good statistical head. There were a few things in this thread that I disagreed with but I wanted to start by responding to this query.

First off realize these values are not built for 12-team leagues - that impacts value.

From my perspective in dealing with development of values above replacement, price is determined by value above the minimal acceptable playable player. So for example, in a 12 man league only 12 2B will play the field full-time, and maybe 2-3 more will play as a DH, with another 12-15 or so receiving partial playing time.

Using NERPs per dollar is not a viable method of valuing a player because it fails to take into account the amount of salary you have to spend. Take for example a $1 player who is worth 4 NERPs. Would you fill your team with such players because their NERP/$ ratio is higher than say Altuve at 3 NERPS? Of course not. So the trick is to spend all of your money while maximizing NERPs - you wont do this if you go by NERP/$ solely. Another way of looking at it would be if you had $999M salary who would you take - Altuve or Harrison?

Lastly, its the total value in relation to standard deviation around the mean. So Harrison may have $6 NERP/$ while Altuve ha $3 NERP/$ but Altuve delivers a higher gross NERP and in relation to the mean of the relevant player pool (which depends on # of teams in the league) and thus delivers a higher standard deviation of performance - making him a more valuable player. For example, assume an average NERP of 20 among 2B and a standard deviation of 3. At 18.16 Harrison is worth -.65 STDEV. Altuve at 28 is worth 2.67 ST DEV. So Altuve is worth 3 ST DEV above Harrison. If Harrison is the baseline for a 12 man league and is valued at $3.11 Altuve could easily be valued at 3 x $3.11 = $9.4M.

That's some very crude math made up on the fly but illustrates my point.

This rambled on but I hope it 1) was helpful 2) was correct and 3) made some lick of sense.

Sam

So in a perfect world a single allowed on defense costs the same as a single a batter hits. However, when you factor in standard deviation of outcomes, a hitters stats are more predictable (more outcomes in a given season) compared to their fielding stats. Therefore defensive stats are discounted based on position (i.e. shortstop might be discounted only 5%, while an OF is discounted 20%).

Think of buying hitters in strat the same as buying a stock. The more consistent you can predict the stock the higher premium paid for the stock price. Volatility comes at a discount, with more seasons coming both above and below the average. This discount originated when the community routinely complained defense was "overpriced", so ultimately I decided to layer on the discount factor, instead of pricing every run equally. I would argue that if you played enough seasons, the better defenders will deliver value above their price point.

And JPav technically defenders are priced so you should be able to play a 3 or a 1 at SS, because its factor into the cost. However, I think what happens is owners tend to remember the outlying seasons, which are more likely with a 3 than a 1.

A few notes on this post:

First off - you seem like a great statistician. I'd be interested to hear more about some of your calculation methods.

Hitter standard deviations over long seasons would tend to normalize - however 1/2 of their rolls come from pitcher cards, 1/3 of which rely on defense, and up to 13 rolls from their card rely on ballparks + Pitcher BP Effects, which vary by card. So there's quite a variety of factors that change from season to season so you can't really aggregate the seasons, I would argue.

Secondly, in regards to your stock illustration I would argue defensive outcomes are the reliable standard. The only variance in defensive performance is the # of x-chance, which may vary by 10% or so from season to season. But the same player, 1e10 for example, will have virtually the same fielding % season over season. The lower the fielder rating the higher the standard deviation of fielding outcomes as they become less predictable and more randomized. This is particularly true for outfielders who have 1/3 as many x chances as a middle IF. None the less, all other variables constant, fielding would certainly be more reliable than hitting. Because of this I pay for fielding first and then worry about hitting. I further compound this by removing ballpark luck by taking low single, low home run ballparks for many of my teams.

Another item that was mentioned earlier is the defense's impact on pitching endurance and opposing hitter opportunities. Hard to quantify for me, but less (AB for them) is more (for me).

I've tried doing some hard math on this stuff - but I am not that great at it. It is really hard to beat a simple 1s up the middle and 3s on the corners strategy.

I have tracked fielding % and found that teams that field both 85% or better and 80% or worse do poorly. Seems like 80-85% is a sweet spot. I built a spreadsheet to forecast the defensive metrics and adjust to stay within these limits. Thoughts on that?

Sam