Loading ...
Sorry, an error occurred while loading the content.

Re: [APBR_analysis] Vin, Rants, and Roland

Expand Messages
  • john1974@u.washington.edu
    ... I have to disagree here. Fielding Runs for first base were way off too. Outfield had its own problems (stemming, I think, more from Palmer and Thorn s
    Message 1 of 11 , Nov 5, 2003
    • 0 Attachment
      On Sun, 2 Nov 2003, Dean Oliver wrote:
      > John's Defensive Rant
      >
      >
      >
      > I only call it a rant because I also rant about defensive evaluation.
      > Baseball has been able to evaluate D pretty well for a long time, well
      > before Win Shares. Hidden Game of Baseball presented defensive evaluations
      > for players that stand the test of time. The only position with some
      > question was catcher.

      I have to disagree here. Fielding Runs for first base were way off too. Outfield had its own problems (stemming, I think, more from Palmer and Thorn's willingness to estimate defensive innings than a real problem in how they put things together). 2nd, 3rd, and short aren't really great either.

      Probably the #1 problem with FR is that it rewards people on a simple 1-to-1 scale for each put-out, assist, and double play. You've ranted about how linear weights don't paint a true picture, so I won't get into that aspect, but in addition in baseball you've got the problem that, no matter how awful or how great your defense is, the game ends after you get 27 outs (or 24 if you're losing after 8 and a half and on the road, but you get the idea). The problem is, on a team-wide level unadjusted fielding stats for really crappy teams are nearly impossible to discern from stats for good teams.

      This is, I think, where James has really been a boon to the industry. Some have castigated him for ignoring the advances in sabremetrics over the last 20 years, but the whole point of the James system is that we don't have as much information for prior years and he wanted to find a way to "unlock" what we already have available. This is basically what Palmer and Thorn tried to do, too, only they did it wrong (no offense to them; much of what they came up with in the early 80s was revolutionary in its own right).

      > Maybe James has improved that. I can't say I've
      > looked. I consider baseball statistical analysis mature enough that I don't
      > see a lot of actual new (and highly relevant) things coming about these
      > days, even from James (other than highly situational stuff, such as the cool
      > study I mentioned a couple months ago on the situations when teams should
      > walk Barry Bonds).

      There's still a lot of interesting stuff coming out - like the firestorm that erupted 3 years ago around the notion that pitchers don't affect their team's performance except in terms of striking players out and not giving up homeruns and walks. Overall, though, I would tend to agree that there isn't any "killer ap" waiting on the horizon (horribly mixed metaphor, I know).

      > I will add that baseball is much easier to evaluate because the interactions
      > among teammates, even on defense, is fairly limited. The greatest
      > interaction is on double plays or relays from throws in the outfield. Most
      > defense is the pitcher fooling hitters and a single fielder catching and/or
      > throwing the ball with a 95+% chance of success. Getting to balls (Range
      > Factor or equivalents) and errors (fielding percentage) do say a fair
      > amount.

      It depends on the position. RF is all but useless for 1st basemen and needs to be adjusted to a tremendous amount to make sense for the other infielders and, for that matter, corner OFers. Then you've got the 27 outs problem. Still, I think you have a point that it is *possible* to figure out defense given what's available.

      > Basketball is very different. I do spell out the method for defense quite
      > thoroughly in the book (due to ship on 11/7 or 11/10 - yes, Brassey's has
      > been struggling to keep up with all the books they are putting out this
      > year). But a crude form of the method has been online for years at the
      > Journal.

      Aw, man... now I'm going to have to go buy it. ;)

      > A big part of the problem with evaluating individual defense in basketball
      > is that individuals are to blame for defensive breakdowns only a fraction of
      > the time. A fair fraction of defensive problems are systemic - the defense
      > chooses to give up a 3 pt shot in favor of shutting down Shaq, for example.
      > That is probably THE BIGGEST result that came out of the work I did with
      > scoring defense. If a perimeter guy comes down to double team Shaq, Shaq
      > gets it back out to Fisher at 25 ft, who makes the 3, whose fault is that?
      > Is it the low box defender who needed help on Shaq? Is it the guy who came
      > to help but couldn't get back? Is it the other perimeter guy who could have
      > rotated over? This happens A LOT.

      Yeah, especially since the early 90s, I agree. Do you just chalk these missed assignments to team defense, or do you attempt to assign blame?

      > I really like what Roland has done.

      Sorry, I didn't mean to sound like I didn't appreciate his work. In fact, what I'm wanting is *more* of it. In particular, I'd like to see the raw position vs. opponents figures. There is some great stuff there that might work out really well in terms of figuring out D.

      > Sure, he has this individual plus/minus
      > stuff that kind of passes for an overall rating, but it ain't. Interpreting
      > any single rating - that number, PER, MikeG's stat, Dupree's evaluation, my
      > own individual win-loss numbers (which do sum quite well to team values) -
      > as a true valuation is what is wrong. It's not wrong to create these stats.
      > It's wrong to treat them as though any single one of them is the Word of
      > God. They are indicators of quality. When people make arguments about an
      > individual's quality purely on their own Overall Basketball Rating, it kind
      > of loses meaning. For a while, I think MikeG was doing that somewhat, but I
      > think he's done a very good job looking at the supporting evidence for what
      > he presents these days. When he presents his lists of top whatever, it's to
      > spur discussions of the reasons why so-and-so may be under or over-rated.
      > That's useful. Same with Roland's plus/minus. People are pointing out
      > problems with the number as an isolated rating. Roland acknowledges those
      > problems and suggests that there are a lot of adjustments that can be used.
      > That is great. As JohnH said, it was horrible that a coach used a single
      > plus/minus to sit down Kendall Gill.

      That was ME, you bastige.

      > I would also say it's horrible to use
      > PER or MikeG's numbers or my numbers or Winston/Sagarin numbers or any of
      > these single numbers to sit down every statistician's least favorite player:
      > Michael Curry. Basketball is a team game with significant complex
      > interactions not reflected very well in statistics. You have to do a ton of
      > work to get something meaningful. I think we here do a pretty good job
      > getting at some of the meaning, but we ain't gonna get it all and we gotta
      > keep that in mind in looking at all summary stats, whether it's Roland's
      > work or anyone else's here.

      I would go so far to add that in basketball it is simply not possible to play "Moneyball"; that these numbers can be useful for fan illumination, but ultimately I'm not sure that they'll have a big impact on how the game itself is played. You've been employed by the Sonics, I know, so maybe your experience is different. It just seems to me that basketball involves such a complex, tangled skein of player attributes and talents that you can't simply make a statement like "Allen Iverson should shoot less" or "the Celtics should trade Antoine Walker" without talking about things in an at least partially non-statistical manner.

      John Craven
    • Dean Oliver
      ... evaluations ... too. Outfield had its own problems (stemming, I think, more from Palmer and Thorn s willingness to estimate defensive innings than a real
      Message 2 of 11 , Nov 5, 2003
      • 0 Attachment
        --- In APBR_analysis@yahoogroups.com, john1974@u... wrote:
        > >
        > > I only call it a rant because I also rant about defensive evaluation.
        > > Baseball has been able to evaluate D pretty well for a long time, well
        > > before Win Shares. Hidden Game of Baseball presented defensive
        evaluations
        > > for players that stand the test of time. The only position with some
        > > question was catcher.
        >
        > I have to disagree here. Fielding Runs for first base were way off
        too. Outfield had its own problems (stemming, I think, more from
        Palmer and Thorn's willingness to estimate defensive innings than a
        real problem in how they put things together). 2nd, 3rd, and short
        aren't really great either.

        It sounds like you've studied it more than I, so I defer, for now. In
        my looks, I didn't see a lot of difference.

        > > A big part of the problem with evaluating individual defense in
        basketball
        > > is that individuals are to blame for defensive breakdowns only a
        fraction of
        > > the time. A fair fraction of defensive problems are systemic -
        the defense
        > > chooses to give up a 3 pt shot in favor of shutting down Shaq, for
        example.
        > > That is probably THE BIGGEST result that came out of the work I
        did with
        > > scoring defense. If a perimeter guy comes down to double team
        Shaq, Shaq
        > > gets it back out to Fisher at 25 ft, who makes the 3, whose fault
        is that?
        > > Is it the low box defender who needed help on Shaq? Is it the guy
        who came
        > > to help but couldn't get back? Is it the other perimeter guy who
        could have
        > > rotated over? This happens A LOT.
        >
        > Yeah, especially since the early 90s, I agree. Do you just chalk
        these missed assignments to team defense, or do you attempt to assign
        blame?

        With a variety of volunteers, it's not perfectly clear how they all
        did it. I told them not to be too concerned about assigning "Team" in
        cases like this or when there was doubt.

        I'd actually say that a lot of this began in the '80's, not
        league-wide, but definitely on several teams.


        > > That's useful. Same with Roland's plus/minus. People are
        pointing out
        > > problems with the number as an isolated rating. Roland
        acknowledges those
        > > problems and suggests that there are a lot of adjustments that can
        be used.
        > > That is great. As JohnH said, it was horrible that a coach used a
        single
        > > plus/minus to sit down Kendall Gill.
        >
        > That was ME, you bastige.

        I'll do that every so often. The anonymous halls of the internet make
        it even more difficult to remember all that...

        >
        > > I would also say it's horrible to use
        > > PER or MikeG's numbers or my numbers or Winston/Sagarin numbers or
        any of
        > > these single numbers to sit down every statistician's least
        favorite player:
        > > Michael Curry. Basketball is a team game with significant complex
        > > interactions not reflected very well in statistics. You have to
        do a ton of
        > > work to get something meaningful. I think we here do a pretty
        good job
        > > getting at some of the meaning, but we ain't gonna get it all and
        we gotta
        > > keep that in mind in looking at all summary stats, whether it's
        Roland's
        > > work or anyone else's here.
        >
        > I would go so far to add that in basketball it is simply not
        possible to play "Moneyball"; that these numbers can be useful for fan
        illumination, but ultimately I'm not sure that they'll have a big
        impact on how the game itself is played. You've been employed by the
        Sonics, I know, so maybe your experience is different. It just seems
        to me that basketball involves such a complex, tangled skein of player
        attributes and talents that you can't simply make a statement like
        "Allen Iverson should shoot less" or "the Celtics should trade Antoine
        Walker" without talking about things in an at least partially
        non-statistical manner.
        >

        Respectfully disagreeing here. I do think that playing "moneyball" is
        a different game in hoops. It's not just identifying the best
        players. I do a lot more than that. It's identifying how players fit
        together. Chp 19 shows the basic optimization you can do, using
        numbers to optimize how players fit together. So, yes, just
        identifying the guys who have the best RC/27 outs may work in
        baseball, but no such thing is possible in basketball. Basketball
        adds a significant degree of freedom in allowing certain players to
        use more possessions, which means a big tradeoff in quality and
        quantity. Further, with that degree of freedom, you can actually use
        stats to IMPROVE players, as we did with Rashard Lewis (hopefully he's
        improved, it is still early). I also looked at similar stuff for
        Collison and Ridnour, the two draft picks, to see how they could best
        improve. Keeping them out of the trainer's room -- not much I could
        do about that. We'll see what happens.

        Yes, there is always non-statistical stuff in the conversation. There
        should be in baseball, too. But stats can be a significant
        contributor. I guess baseball stat guys tend to look at the stats and
        say they _are_ the player. Whereas I'd say that basketball stat guys
        have to look at stats and say they define a range of players that he
        could be. You have to set the environment (his teammates, his
        coaching) right so that he and the team become better.

        DeanO
      • wimpds
        - ... I think you ll find that where James takes on Palmer directly, he makes some good points. ... Could you expand on this? Very intriguing, but I m afraid
        Message 3 of 11 , Nov 5, 2003
        • 0 Attachment
          -
          >
          > It sounds like you've studied it more than I, so I defer, for now. In
          > my looks, I didn't see a lot of difference.
          >

          I think you'll find that where James takes on Palmer directly, he
          makes some good points.

          > quantity. Further, with that degree of freedom, you can actually use
          > stats to IMPROVE players, as we did with Rashard Lewis (hopefully he's
          > improved, it is still early). I also looked at similar stuff for

          Could you expand on this? Very intriguing, but I'm afraid I have
          almost no idea what you mean.
        • Dean Oliver
          ... now. In ... Where is that exactly? I ve read some of the rants on general lin weights vs runs created but not found them very instructive. Having asked
          Message 4 of 11 , Nov 5, 2003
          • 0 Attachment
            --- In APBR_analysis@yahoogroups.com, "wimpds" <wimpds@y...> wrote:
            > -
            > >
            > > It sounds like you've studied it more than I, so I defer, for
            now. In
            > > my looks, I didn't see a lot of difference.
            > >
            >
            > I think you'll find that where James takes on Palmer directly, he
            > makes some good points.
            >

            Where is that exactly? I've read some of the rants on general lin
            weights vs runs created but not found them very instructive. Having
            asked Pete about the disagreements, I tend to find his responses a
            little more, how do I say it, more emotionally detached and
            scientific. But I haven't seen much on the defense.

            > > quantity. Further, with that degree of freedom, you can actually
            use
            > > stats to IMPROVE players, as we did with Rashard Lewis (hopefully
            he's
            > > improved, it is still early). I also looked at similar stuff for
            >
            > Could you expand on this? Very intriguing, but I'm afraid I have
            > almost no idea what you mean.

            Right now, that's a trick of the trade. And it ain't simple.
            Basically, if you see a number of players with certain similar
            characteristics, but certain things distinguish some as being much
            better than others, what causes those things? Why did MJ not commit
            turnovers? Why did Dirk Nowitzki explode after a brutal frosh year
            whereas others don't do that? The right numbers shed interesting
            light on some of this. Some of those are things like similarity
            scores, but there is more. Some of it's in the book, some of it is
            not.

            DeanO
          • bchaikin@aol.com
            I d also like to see other people come up with lists that combine postseason and regular-season stuff.     you can do this at
            Message 5 of 11 , Nov 5, 2003
            • 0 Attachment
              I'd also like to see other people come up with lists that combine postseason and regular-season stuff.    

              you can do this at www.bballsports.com, using the web-based basketball players database (and the "subtotals" feature)...

              bob chaikin
              bchaikin@...



            • bchaikin@aol.com
              I would go so far to add that in basketball it is simply not possible to play Moneyball ; that these numbers can be useful for fan illumination, but
              Message 6 of 11 , Nov 5, 2003
              • 0 Attachment

                I would go so far to add that in basketball it is simply not possible to play "Moneyball"; that these numbers can be useful for fan illumination, but ultimately I'm not sure that they'll have a big impact on how the game itself is played........It just seems to me that basketball involves such a complex, tangled skein of player attributes and talents that you can't simply make a statement like "Allen Iverson should shoot less" or "the Celtics should trade Antoine Walker" without talking about things in an at least partially non-statistical manner.

                aahh... the naysayers exist in every venue :) .... seems to me i've heard this before, say before the revelations of people like bill james (or galileo for that fact)...isn't that what this group is for, to discuss and possibly collectively come up with the better basketball mousetrap?.... guess we'll just have to try to prove you wrong....

                and Roland's efforts will go a long way towards helping that attempt...

                There's still a lot of interesting stuff coming out - like the firestorm that erupted 3 years ago around the notion that pitchers don't affect their team's performance except in terms of striking players out and not giving up homeruns and walks.

                and the person who - i'm not sure if he actually came up with the idea, but who certainly popularized it - did this, voros mccracken, now works for the boston red sox, no?...

                Overall, though, I would tend to agree that there isn't any "killer ap" waiting on the horizon..

                not for basketball? we'll just have to do something about that...

                fyi - "Moneyball" is turning heads not just in the baseball world but the world of almost all professionals sports in the old US of A, to a greater extent than you may be aware of...

                bob chaikin
                bchaikin@...

              • Stephen Greenwell
                and the person who - i m not sure if he actually came up with the idea, but who certainly popularized it - did this, voros mccracken, now works for the boston
                Message 7 of 11 , Nov 5, 2003
                • 0 Attachment
                  and the person who - i'm not sure if he actually came up with the idea, but who certainly popularized it - did this, voros mccracken, now works for the boston red sox, no?...

                  Voros is technically the assistant to Bill James.  Both work for the Red Sox right now, and are allegedly drawing up four scenarios - Red Sox with Nomar and Pedro, just Nomar, just Pedro, or neither.

                  fyi - "Moneyball" is turning heads not just in the baseball world but the world of almost all professionals sports in the old US of A, to a greater extent than you may be aware of...

                  With the ChiSox hiring of Ozzie Guillen and continual bad trades by the Cubs and White Sox, I'm not so sure that this is true ;)  Until the A's or Red Sox or Blue Jays win a world series, I don't think people will be trying to copy them.  The last two winners (Angels and Marlins) have stated policies of relying on "smallball" to win, and the winter hirings have reflected this.

                  Steve Greenwell
                • Mike G
                  ... postseason ... HREF= www.bballsports.com www.bballsports.com , using the web- based basketball ... I believe you without checking, Bob. I ve used your
                  Message 8 of 11 , Nov 8, 2003
                  • 0 Attachment
                    --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
                    > I'd also like to see other people come up with lists that combine
                    postseason
                    > and regular-season stuff.    
                    >
                    > you can do this at <A
                    HREF="www.bballsports.com">www.bballsports.com</A>, using the web-
                    based basketball
                    > players database (and the "subtotals" feature)...

                    I believe you without checking, Bob. I've used your site many times
                    (and never found an error!) However, I was thinking along the lines
                    of comparing seasons and careers.

                    Lots of people have "ranked" player-seasons based on numbers more
                    sophisticated than raw totals. There are a few examples I'm aware
                    of that rank postseasons, as well.

                    A few people have players ranked by career -- but they've only gone
                    by regular-seasons. In the playoffs, your opponents aren't the
                    whole NBA, but a select sample; so any game-pace or competition
                    factors should be customized for each team, each year.

                    It's a pretty big job to do it, but I have done it, by my own
                    imperfect methods. If others do it, we could compare. That's what
                    I've been looking forward to.
                  Your message has been successfully submitted and would be delivered to recipients shortly.