Baseball for the Thinking Fan

Login | Register | Feedback

btf_logo
You are here > Home > Primate Studies > Discussion
Primate Studies
— Where BTF's Members Investigate the Grand Old Game

Monday, September 16, 2002

SUPER-LWTS ? 2001

Mitchel presents the results from last season.

Presenting the results for the 2001 Season

Last year I introduced a revolutionary new metric for evaluating the complete performance of a player.  It is called Super Linear Weights, Superlwts, or SLW,  and is actually an extension of Pete Palmer?s original linear weights (lwts) formula (defense, baserunning, and a few more offensive categories are added to Pete?s formula).  It is based on play-by-play (PBP) data, including hit type and location for all batted balls.  I have since revised some of the methodologies, but the essence has remained the same.

Like Palmer?s lwts, each of the categories is expressed as a number of runs above (plus) or below (minus) the league average, which is naturally defined as zero.  For example, a player who had a defensive lwts of -5 was a below average fielder.  In fact, he allowed 5 more runs to score (theoretically, of course) than an ?average? player.  By convention, regardless of the category, minus is always bad and plus is always good.  Because each category uses the same ?currency? (runs above or below average), all of the categories can be summed to produce each player?s total linear weights, or Super Linear Weights.  A player?s SLW tells us exactly how many runs above or below average he is ?worth? to an average team under average conditions.

It is only better than Palmer?s lwts, BaseRuns, Runs Created, or other similar metrics, because it adds in defensive and baserunning value, and a few extra offensive odds and ends.  However, this makes it a much better metric for evaluating total player value.  There are many players whose good or great offensive performance are completely negated by their poor defensive and baserunning skills, and vice versa.  As well, great defensive value can turn a good offensive player into a great overall player, poor defense can turn an otherwise marginal player into a terrible player, etc.  The permutations are unlimited.

2001 Super Linear Weights Results

Here are the Super Linear Weights categories and their brief explanations - (Note: A more detailed explanation of the various methodologies used in SLW can be found in the original two articles.  A link to them is provided at the bottom of the page)

  1. Player Name

  2. 2001 Team (if more than one team, the team with the most PA?s)

  3. Primary position (10 is DH or unknown)

  4. Number of PA?s in 2001 (PA=AB+SF+BB+HBP)

  5. Defense

    • UZR - Ultimate Zone Rating or defensive lwts - runs saved or cost on defense, park adjusted, based on PBP hit type and location data.

    • GDP defense - for infielders only.  Number of DP?s turned per DP opportunity, compared to the league average at each position, adjusted for league-average number of opportunities, and converted into runs saved or runs cost.  When a DP is turned, the fielder and the pivot man are both credited.  If the fielder and the pivot man are the same, only one credit is given.  Also park adjusted.

    • OF arms - UZR does not include throws by the outfielders.  This category includes an outfielder?s outs (assists), holds, and extra bases allowed, also compared to league averages, adjusted for league-average opportunities, and converted into runs saved or cost.

    • Catching - for catchers only, the combined run value of their sb/cs totals, errors, and passed balls, as compared to the league-average catcher.

  6. Baserunning

    • Taking the extra base - this is the exact “reverse” of the OF arms category - it is extra bases, outs (OOB), and holds by a baserunner.

    • GIDP - this is a batter?s number of GDP?s per opportunity, compared to the league average, adjusted for league-average number of opportunities, and, as always, converted into runs saved or cost.

  7. Hitting

    • Moving runners over - a batter moving a runner from 2nd to 3rd on a fly or ground out (giving oneself up), per opportunity, compared to league averages, and converted into runs.

    • Batting - traditional offensive linear weights (a la Palmer), including sb/cs.  They are park adjusted, using regressed, 3-year (if possible) component park factors.

  8. Super Linear Weights

    • Superlwts - all previous categories combined.

    • Superlwts per 162 games - above category prorated to the PA equivalent of 162 games for an average player.

    • Position adjusted Superlwts - adds in a simple positional adjustment.  For example, if an average SS in 2001 had a Superlwts value of -16 per 500 PA, and player A had a Superlwts value of -10 in 250 PA, he gets a plus 8 run adjustment, to give him a position adjusted Superlwts value of -2.  (In other words, in 250 PA, he was 2 runs worse than an average SS in 2001.)

    • Position adjusted Superlwts per 162 games - the above prorated to the PA equivalent of 162 games for an average player.

The chart below contains the results of the 2001 season only.

View the sortable Super Linear Weights, 2001

Note: after clicking the above link, you will be prompted: click NO and wait a few seconds for the results

Mitchel Lichtman Posted: September 16, 2002 at 06:00 AM | 18 comment(s) Login to Bookmark
  Related News:

Reader Comments and Retorts

Go to end of page

Statements posted here are those of our readers and do not represent the BaseballThinkFactory. Names are provided by the poster and are not verified. We ask that posters follow our submission policy. Please report any inappropriate comments.

   1. Mike Emeigh Posted: September 16, 2002 at 12:46 AM (#606198)

He has conclusively argued that you cannot possibly rate players against a norm, that you must take—heavily—into context team performance. In fact, team performance—did they win and how much??—is the starting point for James’ revolutionary analysis.

James’s so-called *revolutionary* analysis is hardly that; as I have pointed out before, both Clay Davenport and Charlie Saeger derived their defensive evaluation models based on team performance long before James did. I would go so far as to say that James doesn’t do enough to account for team context when evaluating defense, because of his use of a fixed pitching/fielding ratio in allocating defensive Win Shares and because he doesn’t place individual defensive performance into the context of team defensive failures.

—MWE

   2. tangotiger Posted: September 16, 2002 at 12:46 AM (#606201)

Bill James knows not what he speaks of. 

He starts off with the notion that “0” means no value, when “0”, on a relative scale, means “equal to average”.

Therefore, an average player will show as “0”.  That does NOT mean he has no value.  It simply means he has no MORE value, than an average player playing as much as he did.  A “0” regular player will make 4 million$ per year.

If you want to argue that “replacement level” is a better concept for “rating”, fine.  There’s not many people who would argue against it.  But the concept of Linear Weights is sound.  I would ignore anything Bill James has to say on the subject.

   3. tangotiger Posted: September 16, 2002 at 12:46 AM (#606202)

And the flawed fielding runs has been highly improved upon (really supplanted by) what Mitchel does here.

   4. Mitchel Lichtman Posted: September 16, 2002 at 12:46 AM (#606209)

Paul, you have a nice writing style, but you should stick to writing novels or short stories.

Win Shares is fine.  It has its problems and limitations as do most metrics.  Some of them have been documented here and there on the web and I suppose in other venues as well.

Win Shares and Superlwts are by no means mutually exclusive.  As well, they are apples and oranges.  I imagine they both have their place.

Win shares is a “gross” representation of a player’s value to his team, albeit one that takes into consideration “context”.  However, this inclusion of context does not nearly make up for the deficiencies related to using commonly available stats rather than PBP data, particularly in the area of defense (see my comment on the adjusted Range Factor article).

Lwts and Superlwts is a “what you see (hear) is what you get” type of metric.  It does not purport to consider context, nor should it.  It is designed and defined to be context-neutral.  In fact, it is a near-perfect estimate of a player’s theoretical value to an “as yet to be determined” team, with that player to be inserted into an “as yet to be announced” spot in the lineup.  As well, it is a near-perfect metric for comparing the overall “talent” (as defined by the previous sentence) of one player to another, again without regard to what team they did or will play for.  To characterize such a metric as antiquated or useless or compare it to medieval terra-centric theories or suggest that it has little practical value is preposterous!

Now, if someone wants to take lwts or Superlwts and make the leap from theoretical context-neutral value to context-specific value, by taking into consideration the surrounding players, the home field, and the lineup spot, it is not that difficult.  All one needs to do is to use “custom” lwts values for all of the Superlwts categories.  In reality, this is usually not necessary for a variety of reasons: 1) custom lwts will not vary all that much from team to team; 2) many players often occupy a number of different lineup positions on a regular basis; 3) we may WANT to know how a player will perform, i.e., his theoretical value, in an undetermined context (i.e., on an average team in an average environment).

Win Shares (although I am not entirely familiar with it) is quite different from a lwts-type metric.  It seems to be more of an “MVP” type stat, whereas a lwts or Superlwts stat is more useful for estimating talent and projecting performance.  IMO, and it is just MO, Win Shares is a relatively boring and impractical type of stat.  I personally don’t care who is/was “responsible” for a team’s wins.  To me, this is more of an intellectual exercise than anything else.  I prefer to know 2 things: How “good” is one player relative to another and what can I expect if I add player A to my team and/or replace player B with player A.  Superlwts is a near-perfect metric for this.

All that being said, the primary “objection” I have with Win Shares or any metric or tool that uses limited information, regardless of how well it is constructed, is that it is naturally and inherently unreliable, because of a lack of complete data (again,see my comment on the “adjusted Range Factor” article).  That is really the strength, if you will, of Superlwts - the fact that it takes advantage of complete PBP data in order to more reliably estimate a player’s overall “talent”.  The methodologies involved in Superlwts are neither complicated nor sophisticated.  Perhaps Win Shares is more “cutting edge” and creative.  However, because of what it apparently measures (“win shares” I guess), which is patently uninteresting to me, and because of the limitations in its reliability (based upon the fact that it uses incomplete and limited data), not to mention its use of some questionable methodologies, which I have not even discussed, I MUCH prefer Superlwts!  Of course, I am a tad biased!

   5. dlf Posted: September 16, 2002 at 12:46 AM (#606212)

Over on fanhome, there was some discussion of using a player’s actual opportunities versus using a league average number of opportunities for some of the categories.  Which choice was made in the presented data?

   6. Rob Wood Posted: September 16, 2002 at 12:46 AM (#606213)

I am a fan of both Win Shares and Super-LWTS.  They both have their uses, strengths, and weaknesses.  They are complements, not substitutes.

Perhaps the biggest difference between them is that Win Shares purposefully attempts to allocate actual wins (and is therefore good at backward looking descriptions) whereas Super-LWTS estimates the player’s context neutral value (and is therefore better for forward looking predictions).  If I can inject a personal comment, the same difference applies to my own new Win Values stat and other traditional pitching stats.

Having said all that, I am troubled by one aspect of Super-LWTS.  I believe that the “positional adjustment” method that Pete Palmer introduced is still used.  The idea is that shortstop is more demanding than left field, and one way to estimate this value difference is to look at the difference in shortstops’ offense vs left fielders’ offense.  (Implicitly, each defensive position is assumed to have the same value, counting offense and defense.)

The problem with this is that the way you estimate this value then depends upon the quality of the other shortstops in the league.  ARod is penalized, to some extent, because Nomar, Jeter, and Tejada currently play the position.  I have many concerns about Win Shares, but that system does not rely upon a “equal value” positional adjustment. 

Am I correct that Super-LWTS relies upon this assumption?

   7. Mitchel Lichtman Posted: September 16, 2002 at 12:47 AM (#606223)

Yes, the values in each category are “pretty much” adjusted for league average opportunities.  For example, if player A’s arm lwts were +3, but he happened to have 3 times the opportunities to either hold a runner or throw him out, his arm lwts would be presented as +1, rather than +3.  Is that your question?

As far as my positional adjustments, I’m not crazy about them either.  However, they are what they are (“what you see is what you get”).  I didn’t need to include them, of course, in a separate column.  Anyone could have figured them out if they bothered to figure out the average Superlwts value for each position.  That’s all they are based on.  They are simply a player’s Superlwts relative to an average player at that same position, rather than an average player in the league (NL and AL combined).  They are nothing special.  I don’t think you can “criticize” them any more than you can “criticize” that “2+2=4”.  If someone wants to “use” them, fine.  If not, that’s also fine.  People talk about using runs above/below replacement value rather than runs above league average or league average at a partiuclar position to “evaluate” a player.  That’s fine too.  All the info is there (in the chart and the methodogogy).  The reader can do any calculation they want to determine “runs above replacement”, or whatever they want that is not readily available in the Superlwts columns.  I take no position on how positional adjustments should or shoudn’t be done or whether we should “pay more attention” to runs above replacement or runs above average, or whatever…

   8. Rob Wood Posted: September 16, 2002 at 12:47 AM (#606232)

David & Mitchel,  I understand and appreciate the need for the positional adjustment.  Clearly the adjusted results are far superior to the unadjusted results.

At the risk of repeating myself, my point was that the SLW system cannot present independent evidence of the value of the different defensive positions.  The defensive value of a shortstop is taken to be the difference between the league average hitter and the league average shortstop-hitter, or something like that.

Thus, as a super-stat of overall value it is not quite perfect.  I would prefer a system that independently assesses the value of each defensive position, and then allocates value appropriately.  Bill James’s old saw about never seeing a shortstop carry his glove into the batter’s box springs to mind.

However, I admit that this is probably a very minor issue in the big scheme of things.  The purpose of my first post was simply to determine if this “character flaw” from the original LWT system had been remedied.  By the way, according to win shares, the average value across defensive positions (offense + defense) has indeed been roughly equal over time.  So the assumption underlying the positional adjustment is largely justified.

   9. tangotiger Posted: September 16, 2002 at 12:47 AM (#606239)

Rob: regarding positional adjustments. 

Several months ago, Mitchel provided me with data of players who played multi-positions (e.g., Jose Oquendo), and I looked at how each player’s defensive value was relative to the league average, so that I could determine the “difficulty” of each position.

I also looked at 1989-2000 offensive linear weights to determine if the positional difficulty spectrum matched the offensive spectrum.

Loosely speaking, they did.  The SS was about 12 runs from league average, and the 1B was about 15 runs the other way.  There were some inconsistency (between 2b,3b,cf), but overall it was very enlightening.  It is fair to say that over a long enough period, each position is roughly equal in value.

From that standpoint, it would make more sense to use the long-term positional adjustments, rather than the given year.  However, in this case, the 2001 adjustments are pretty darn close.

E-mail me at .(JavaScript must be enabled to view this email address) , and I’ll pass you on more information if you like.

   10. scruff Posted: September 16, 2002 at 12:47 AM (#606244)

David and Mitchel—you both keep hammering that not using all of the data is wrong. On one level I can see that, but what about those of us that don’t have access to this data, or those of us that want to compare players historically?

That’s what makes what James has done so important, you can compare players across time, using the same methods. I think it’s much more useful from a historical perspective to come up with method that are almost as accurate using the traditional stats, and then validate them using PBP data for the years it’s available.

Again I’m not trying to get into the validity of LWTS or WS or anything else. But I think it’s very important to be able to compare players historically, so I’m alway partial to methods I can plug into a spreadsheet and figure out myself, methods I that I can pick up my Stats All-Time Handbook or Total Baseball and plug those numbers into.

   11. Mitchel Lichtman Posted: September 17, 2002 at 12:47 AM (#606247)

I think I’ll lay the “positional adjustments” and “WS versus Superlwts ” arguments to rest.

As far as pitcher Superlwts, yes they can be done in a similar way. 
However…

One, you would need to use “custom” lwts for a pitcher’s component stats, as a pitcher’s individual stats are like a “player within a team”.  Some would argue, of course, that a pitcher’s lwts, even “customized”,  based upon his component stats, do not fairly or accurately represent his talent - that you would want to use his ERA or W/L record in addition to or in lieu of.  That may or may not be true.

Two, pitcher defense is hard to quantify.  I do not compute ZR’s or UZR’s for pitchers or catchers for obvious reasons.

Three, except perhaps for a few pitchers, pitcher batting probably does not make that much difference (maybe it does - by a few runs or so per season - in which case, every little bit helps - or hurts).  It would be nice though to include sac bunting prowess in a pitcher’s repetoire of value.

Four, pitcher baserunning almost for sure makes very little difference.


Now that you bring it up, it’s not such a bad idea…

   12. Rob Wood Posted: September 17, 2002 at 12:47 AM (#606263)

Thanks Tango (and others) for the information concerning positional adjustments.  I agree that the most reasonable thing to do would be to use a positional adjustment calculated over several seasons so as to minimize the season-to-season variations.

By the way, I don’t quite understand Mitchel’s remark that he’ll lay the argument to rest.  Independently assessing positional values has long been one of the holy grails of sabermetrics (along with allocating run prevention credit between pitchers and fielders, which is taken up in another current thread).  So of course the issue continues to comes up.  If Mitchel meant to say that SLW sheds no additional light on the subject, so be it.

   13. Mike Emeigh Posted: September 17, 2002 at 12:47 AM (#606264)

It does not purport to consider context, nor should it. It is designed and defined to be context-neutral.

It had *better* consider, and account for, context, otherwise it can’t possibly be context-neutral. And UZR doesn’t accurately account for defensive context; no static zone-based system that fails to account for differences in positioning based on ball distribution and pitching staff and ballpark characteristics can possibly do so.

—MWE

   14. Mitchel Lichtman Posted: September 18, 2002 at 12:48 AM (#606293)

Tracy, allowing lots of runs and non-HR hits (the 2001 Pads) means that your pitching probably stinks.  If you allow lots of non-HR hits per balls in play then your defense probably stinks.  So #1, let’s not associate a team’s runs allowed with their defense.  There is probably not a whole lot of connection, as the primary factor, by far and away, in runs allowed, is pitching and not defense.  #2, before we implicate a team’s defense in a high non-HR hits allowed, we need to see whether that is because they allowed lots of balls in play or they did not catch their fair share of balls in play.  Since my defensive ratings ARE essentially how many balls in play are caught (turned into outs), I think we can safely assume that the number of outs by the outfielders on the Pads last year, divided by the number of balls in play to the outfield, was much lower than average, otherwise I am doing something seriously and fundamentally wrong.

Let’s see what went on with the Pads last year…

If I look at my team defensive lwts files I find that SD had an IF defensive lwts of -5 and an outfield defensive lwts of +73.  The +73 reckons with your observation that the main 2001 Pads outfielders all had very high defensive lwts.  Even though this number +73, and the individual OF’ers defensive lwts are based upon a UZR, it is almost a sure thing that the Padres outfield should have a very high ZR as well.  Remember that a ZR is simply balls caught divided by balls in play.  For the outfield this is a very simple metric - it is fly balls and line drives caught by the OF’ers divided by fly balls hit to the outfielders (which is of course OF PO’s plus OF hits plus OF errors on dropped fly balls or line drives).  The -5 IF defensive lwts, also based on UZR, suggests that the entire Padres IF was only slightly below average in catching ground balls (I only look at ground balls for the IF’ers).  So unless the Pads had tons of line drives hit thru their infield (which I don’t count at all), my results suggest that the Pads should have had a very low OF non-HR hits per BIP and a slightly high non-HR hits per BIP (ground balls only) in the IF.  You claim that the Pads had a high number of non-HR hits allowed in general, so either they had lots of balls in play, compared to the average team, or I am full of it.  Let’s see…

You say that SD was 16th overall (I assume you mean in the NL) in non-HR hits allowed in 2001.  According to STATS ML handbook, they gave up 1519 hits and 219 HR’s, for a total of 1300 non-HR hits.  MON had 1509/190 or 1319 non-HR hits, CIN had 1572/198, or 1374 non-HR hits, and PIT 1493/167, or 1326 non-HR hits.  Where did you get your figures?

In any case, yes SD was near the bottom in non-HR hits allowed.  But again, we are only interested in non-HR hits allowed.  Well, SD pitchers had the 6th fewest SO in the NL last year and the third lowest TBB, yet they had the 5th highest TBF.  What does this tell you?  It tells you that they had one of the highest BIP, if not the highest, in the NL last year.  Let’s get down and dirty…

SD actually had 4439 BIP (non-HR).  This is defined as TBF minus HB minus TBB minus K minus HR.  They had 3107 BIP outs.  This is defined as IP times 3 (total outs) minus K minus DP.  Their BIP outs divided by their BIP (DER or defensive efficiency rating) was therefore 3107/4439 or .700.

The entire NL had 69045 non-HR BIP and 48953 BIP outs for a DER of .709.  So although SD allowed the 4th highest non-HR hits, they were only slightly below average in DER, or BIP caught divided by BIP.

Well, I must admit that I expected their DER to be higher.  On the other hand, you should be equally surprised that their DER is so high. 

Given the fact that I compute their OF lwts to be very high and their IF lwts to be slightly low, I must conclude that either the SD pitchers gave up lots of line drive hits thru the infield and/or few pop-ups on the infield (which is entirely possible since their pitching was indeed bad and that is one of the things that bad pitchers do), relative to the league average, and/or they gave up lots of line drives to the outfield (which would affect UZR but not ZR or DER), which is again entirely possible due to their poor overall pitching, or I made some error(s) in calculating their OF or IF defensive lwts, based on their UZR’s.

I’ll look into it…

   15. tangotiger Posted: September 18, 2002 at 12:48 AM (#606306)

Thinking of this intuitively, the idea that a team would be below average in DER yet (apparently) above average in UZR is something I’m not quite ready to embrace.

I would bet that Mitchel will find that in this case the Padres pitchers allowed alot of hits in the “outlier” zones.  That is, the Padres make it real tough on their fielders, and don’t give them easy balls to hit.

Think of it this way: if you have a pitcher who really really knows how to control the balls in play, he will get all the balls hit directly to his fielders.  The DER will be very very high.  However, UZR is aware that the DER SHOULD be high for those defenders because the range of balls hit is in the “high out-conversion” zones.

On the other hand, if you have pitchers who can’t control a ball in play, say like if I was pitching, then lots of balls will be hit in the gaps and in no-man’s land zones, where not even Andruw Jones can get to them.  UZR knows about this as well.

DIPS does NOT say that the pitcher has no control, but relatively little control.  And some pitchers have more control than others.  And in this case, I’ll bet that the Padres pitchers have little control.

   16. Mitchel Lichtman Posted: September 19, 2002 at 12:48 AM (#606338)

As I said I would, I checked on what was happening with the SD team’s BIP’s last year.  Interesting results…

Here are all the 2001 NL team’s: 1) total BIP’s; 2) GB’s;  3) LD’s, and ; 4) FB’s allowed.  Foul balls that are caught for an out are not counted in any category.

ARI 3565 .493 .171 .335
ATL 3714 .477 .204 .318
CHN 3542 .484 .211 .305
CIN 4065 .491 .202 .307
COL 3908 .472 .225 .302
FLO 3685 .481 .201 .319
HOU 3705 .487 .186 .328
LAN 3618 .488 .208 .304
MIL 3845 .492 .203 .305
MON 3717 .497 .192 .311
NYN 3689 .467 .211 .322
PHI 3714 .468 .215 .317
PIT 4011 .512 .186 .302
SDN 3886 .487 .218 .295
SLN 3724 .493 .207 .299
SFN 3929 .467 .210 .323

NL 60317 .405 .203 .312

As you can see, SD pitchers allowed the highest percentage of line drives per BIP, and the lowest percentage of FB per BIP.  (As you noted, they had a very high GB/FB ratio AND because of poor pitching, bad luck, or both, they allowed an inordinate number of line drives.) 
Now we can easily see why their DER (defensive efficiency rating, which is the same as team ZR, which is simply BIP outs divided by BIP) was slightly below average, but their team UZR (as computed by adding up the UZR’s of all their individual players) was very high!  Fly balls are caught the greatest percentage of the time (.864), followed by ground balls (.749), followed by line drives (.258).

SD, in fact, caught .735 of their GB’s allowed, less than the league average of .749, suggesting a sub-average defensive IF, .279 of their LD’s, more than the league average of .258, suggesting good luck (for IF line drives) and goood outfield defense (for OF line drives), and .893 of their FB’s, higher than the league average of .864, also suggesting a good outfield defense.

So the raw numbers jive very nicely with the total UZR ratings of the SD IF and OF and with the hypothesis that SD had a poor pitching staff (too many balls in play and too many line drives allowed). 

Keep in mind one more thing when it comes to comparing a team’s DER or ZR with team UZR or totaling the UZR’s of a team’s players:

If a team has a high GB/FB ratio, their DER or ZR will look inordinately bad, since GB’s are caught with a lesser frequency than fly balls.  However, FB hits have a much higher run value than GB hits, so that a team with a high GB/FB ratio, like SD, will actually have a better UZR, since UZR takes into consideration, among other things that ZR or DER does not, the hit value of the various types of hits (OF FB’s, IF LD’s, OF LD’s, GB’s) as well as the various types of hits in the various zones.

Here is the fallacy of your initial observation that “Because SD allowed lots of hits, their team UZR, at least as computed by summing the UZR’s of their OF’ers, which was high, must be wrong:”

The same fallacy occurs if we use the same logic to say that “Because the SD DER was lower than average, their team UZR also could not possibly be so high:”

If hits allowed jived with ZR or DER or UZR in every case, we would not need a ZR or DER metric, would we?  The idea of having a more rigorous metric, like ZR or DER, rather than simply using “non-HR hits allowed”, is to identify those teams whose “non-HR hits allowed” do not jive with their ZR or DER!  The logic that a team cannot have a high “non-HR hits allowed” and a high ZR or UZR is the same (fallacious) logic that would go into the statement “How can player A have a .240 average and be worth 20 runs above average in lwts or RC?”  The reason is, of course, that we have lots more information to add to the .240 average (presumably high walk total and high secondary average).  Same with hits allowed versus ZR and ZR versus UZR.  While most teams will have all of those measures jive with one another, there will be a few (like SD), that don’t jive, because of the reasons outlined above (in this case, because the SD pitchers allowed many line drives and few fly balls).

I hope that the above data and analysis “clears up” the controversy surrounding the SD defense in 2001.  I think that the data in the chart speaks for itself, in terms of why SD allowed a great number of non-HR hits, had a slightly below-average DER team ZR), and a signficiantly above-average UZR in the OF and slightly below-average UZR in the IF…

BTW, if you look at the above chart, which is BTW a snap to generate if you have the PBP database of course, you might begin to hate team DER as a measure of a team’s defensive talent, as you can see how easy it is to mis-evaluate that talent if a team does not have a fairly average distribution of GB’s, LD’s, and FB’s, especially in the LD department.  At the very least, DER should be asjusted for a team’s GB/FB ratio…

   17. tangotiger Posted: September 19, 2002 at 12:49 AM (#606339)

Great stuff, Mickey!

This breakdown occurs in different facets of baseball as well.  Take the saves, which STATS has nicely broken down into “tough”, “regular”, and “easy”.  The conversion rate for the tough saves is much much lower than the easy save.  So, what happens if you have John Wetteland getting lots of tough save opps?  Well, his overall save % rate might even be lower than the league average, but he could be higher than the league average in EACH of the three categories.

This is also true in other things like RBIs.  You can have say Tim Raines perform much better than average with men on base AND with bases empty, but because he has so few opps with men on base, he will actually drive in less runners than the league average.

Throughout baseball, there are countless categories like this.  This is why it is CRITICAL to know two things (at least): the frequency at which each of the sub-categories occur, and the success rate for each of these sub-categories.

The SD example is a prime example of this situation.

However, I’d like Mickey to expand on my comment about the “third” thing, and that is, within each of the sub-categories (FB,GB,LD), is the distribution between the “high-out-conversion” zones and “medium-out-conversion” zones and “low-out-conversion” zones similar for SD as it is for the league?  (use whatever “league-conversion rate” you want to identify the various zones)

   18. tangotiger Posted: September 19, 2002 at 12:49 AM (#606344)

I know that we’ve (Mickey and I) discussed this in the past, regarding defense.  This ties in directly to what you’ve discovered in your previous email.

My position has always been to show:
Jeter, .854, .786 ... which shows what Jeter’s actual ZR is, and what the league average SS ZR would be, if he played under Jeter’s conditions (the ball distribution would be different because of his pitchers, the park, etc).

The danger in “adjusting” his stats is that you get in the scenario like the Padres where they have a below avg ZR (DER), but above aveage UZR.

However, and you made this quite clear, you showed the various distributions of GB/FB/LD, and the league success rates on that.

Now, look what happens when we follow the approach I am advocating:

NL2001 lgDER
ARI 0.703
HOU 0.696
PIT 0.692
MON 0.690
FLO 0.688
CIN 0.685
ATL 0.685
MIL 0.684
SFN 0.683
NYN 0.682
LAN 0.682
SLN 0.681
CHN 0.680
PHI 0.680
SDN 0.676
COL 0.673

Ah-ha!  Now, it becomes very clear what’s going on.  BASED on the GB/FB/LD aspect of the PITCHERS/PARK, we see that SD and Colorado fielders SHOULD have a tough time converting balls into outs.  Because this is what a team of league average defenders would do in these conditions.

We could (and should) have more conditions: LH/RH hitters for one.

This is why it is very important, in my view, that we leave the player data alone, and instead “adjust” the league average player INTO the player’s conditions.

(Actually, we should do BOTH, since each answers a different question.)

You must be Registered and Logged In to post comments.

 

 

<< Back to main

BBTF Partner

Support BBTF

donate

Thanks to
Shooty Survived the Shutdown of '14!
for his generous support.

Bookmarks

You must be logged in to view your Bookmarks.

Syndicate

Page rendered in 0.6582 seconds
66 querie(s) executed