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

Re: [APBR_analysis] Anecdotal and other unnumeric comments

Expand Messages
  • bchaikin@aol.com
    I think BobC s simulation results are fascinating, if untestable.  Such highly-varying results from placing a given player on a historic team.  why might
    Message 1 of 6 , Apr 1 8:13 AM
    • 0 Attachment
      I think BobC's simulation results are fascinating, if untestable.  Such highly-varying results from "placing" a given player on a historic team. 

      why might you insist, without proof or evidence, that the results are untestable? what incorrect assumption are you making to flatly denounce the results?...

      think about it - there is an easy way to test it. simply recreate a trade that occurred one season using the simulation, and compare it to the real life results of the next season, i.e. for example, recreate a trade that occurred in between the 94-95 and 95-96 seasons using the 94-95 stats, and compare the simulated results with the real life 95-96 results. you can do this for any trade or free agent signing...

      or take a simple example where a couple of starters were replaced in a team's lineup from one season to the very next: in 87-88 the atlanta hawks had dominique wilkins as their main offensive weapon. but between 87-88 and 88-89 they traded for moses malone and reggie theus and started them the next season (88-89) alongside dominique, replacing tree rollins and randy wittman. you can, using the 87-88 stats, "trade" malone and theus to the hawks and run multiple 82 game seasons, and compare those results to the real life results of the 88-89 hawks...

      I assume you are replacing the equivalent player on the team?

      you can but you certainly don't have to - the whole point is to see how much better or worse a team would be with a certain player, replacing whomever you like. like trading tim duncan to any number of teams but playing him at center instead of power forward. or taking norm nixon off the lakers of the early 1980s who was playing the SG position and playing him as a starting PG on another team. for the penny hardaway example used previously, he could easily be traded to a team and realistically be used at PG, SG, or even SF...

      i wouldn't advise using the software by, for example, taking a PG from one team and using him at C on another - common sense dictates that to be unrealistic. but pretty much any other scenario that would happen in real life can be simulated...

      bob chaikin
      bchaikin@...


    • Mike G
      ... untestable.  Such ... team.  ] ... are ... denounce the ... I m assuming you can t take us back to 1996, trade Penny Hardaway to 8 different teams, and
      Message 2 of 6 , Apr 1 9:35 AM
      • 0 Attachment
        --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
        >[I wrote: I think BobC's simulation results are fascinating, if
        untestable.  Such
        > highly-varying results from "placing" a given player on a historic
        team.  ]
        >
        > why might you insist, without proof or evidence, that the results
        are
        > untestable? what incorrect assumption are you making to flatly
        denounce the
        > results?...

        I'm assuming you can't take us back to 1996, trade Penny Hardaway to
        8 different teams, and play 82 games with the whole NBA, 8 times
        over, to see how close your simulations are.

        >
        > think about it - there is an easy way to test it. simply recreate
        a trade
        > that occurred
        > one season using the simulation, and compare it to the real life
        results of
        > the next season, ...

        I recall you 'recreated' the Rodman-to-SA trade of some years back,
        and, as usual, the results were fascinating. No doubt, when you do
        these 'trades', you also get some surprising results, and some that
        aren't even close to the real-life situation.

        As long as players change from year to year -- even when their
        environment is unchanged (relatively) -- you can't predict with any
        certainty how they will perform the next season.

        An ascendant player might be projected to continue his improvement
        at least another season. One who is 'over the hill' can be
        projected to continue to slip.

        Any system of prediction will hit a few on the nose, and miss others
        by a mile. I honestly can't imagine one using sounder principles
        than yours, Bob -- at least, from what I understand of it.

        Rather than focus on perfectly-made predictions or resounding
        failures, I firmly believe individual player inconsistency will
        continue to confound all prognostication.
      • bchaikin@aol.com
        I m assuming you can t take us back to 1996, trade Penny Hardaway to 8 different teams, and play 82 games with the whole NBA, 8 times over, to see how close
        Message 3 of 6 , Apr 1 10:23 AM
        • 0 Attachment
          I'm assuming you can't take us back to 1996, trade Penny Hardaway to 8 different teams, and play 82 games with the whole NBA, 8 times over, to see how close your simulations are.

          and why not? you can simulate him on any team you choose, any number of times, with pretty much any player combinations you want...

          why couldn't i, say, using the 02-03 stats, trade karl malone of the jazz and gary payton of the sonics to the 02-03 lakers, team them up with kobe and shaq, and see what happens? let's see now, in 02-03 shaq averaged 27.5 pts/g (38 min/g) and kobe 30.0 pts/g (41 min/g), malone 20.6 pts/g (36 min/g) and payton 20.8 pts/g (41 min/g)....

          now i put the 02-03 malone and 02-03 payton on the 02-03 lakers, and adjust all their min/g to match what they are getting in real life now (shaq 36 min/g, malone 32, bryant 36, and payton 36 - in actuality malone is really getting 33, bryant 37, and payton 34, but in the sim i have to put them in in multiples of 4 min at a time) so the result is as realistic as possible and see what happens? well, playing 8200 games shaq averages 22.6 pts/g, malone 14.3 pts/g, kobe 21.2 pts/g, and payton 15.2 pts/g, and the team averages a W-L record of 54-28 playing 8200 games. tell me, how close are those numbers to what the real life numbers are? if they all played the same amount of minutes they did in 02-03, simulation shows they would win close to 59-60 games per 82 game replay...

          oh, btw, right now in real life shaq's at 21.7 pts/g, kobe 23.6 pts/g, malone 13.5 pts/g, and payton 14.8 pts/g. so tell me now, is that close enough?...

          I recall you 'recreated' the Rodman-to-SA trade of some years back, and, as usual, the results were fascinating. No doubt, when you do these 'trades', you also get some surprising results, and some that aren't even close to the real-life situation.

          just what kind of "...surprising results..." might you be speaking of? do you have examples? evidence? tell me how some trades "...aren't even close to the real-life situation..."? please give me some examples. or do you not have any and are just speaking without proof or evidence?...

          As long as players change from year to year -- even when their environment is unchanged (relatively) -- you can't predict with any certainty how they will perform the next season.

          you meaning who? you? tell me what is your sound reasoning as to why this cannot be done?...

          people have mapped out using sound fundamental scientific principles the space missions to jupiter (gallelio) and saturn (cassini), then developed them and put them in action, that despite their glitches have been of great success, yet just because you or i didn't have any part of it (darn!) doesn't mean that it can't be done...

          so yes you can predict real life situations. can the software predict the performances this season of players like carmelo anthony and lebron james? absolutely not - because there is no prior statistical evidence for either player on the pro level....

          Any system of prediction will hit a few on the nose, and miss others by a mile.

          the software does have its limitations in that if there is not enough prior statistical evidence for a player or players, you can't model their pro level "behavior" with any sound accuracy. but if a player or players does have some meaningful data to work with (previous stats on a pro level), you can...

          you (meaning anyone in general) would be surpriised at what you can glean about players by just looking at their statistical history, as most in this discussion group already know...

          I honestly can't imagine one using sounder principles than yours, Bob -- at least, from what I understand of it.

          * sniff *.... i love ya' man....

          Rather than focus on perfectly-made predictions or resounding failures, I firmly believe individual player inconsistency will continue to confound all prognostication.

          once again, a model is a model. it does what its programmed to do...

          bob chaikin
          bchaikin@...

        • Mike G
          ... kobe and ... 14.3 pts/g, ... L record of ... what the real ... Very close. And the young Kobe is the one whose PPG are better than the sim suggests, while
          Message 4 of 6 , Apr 2 5:44 AM
          • 0 Attachment
            --- In APBR_analysis@yahoogroups.com, bchaikin@a... wrote:
            > ... using the 02-03 stats, trade karl malone of the jazz and
            > gary payton of the sonics to the 02-03 lakers, team them up with
            kobe and
            > shaq, and see what happens? ...shaq averages 22.6 pts/g, malone
            14.3 pts/g,
            > kobe 21.2 pts/g, and payton 15.2 pts/g, and the team averages a W-
            L record of
            > 54-28 playing 8200 games. tell me, how close are those numbers to
            what the real
            > life numbers are?

            Very close. And the young Kobe is the one whose PPG are better than
            the sim suggests, while the 3 old guys' PPG are less.

            But what I find highly surprising is the Lakers' winning only 54
            games, in this sim. Many people thought they'd win 70-80 games.
            Did you have to knock out key players for substantial parts of your
            test?

            > just what kind of "...surprising results..." might you be speaking
            of?

            I'd be surprised if a simulation would 'predict' Jermaine O'Neal's
            rise in production, upon moving to the Pacers. Or if the Mavs
            should be so weak, after their offseason roster changes. Stuff like
            that.

            And probably I'm not the only one wondering where was your
            prediction that the Lakers should win only 54 games, at the
            beginning of the season? There's no shortage of after-the-
            fact "predictions".


            > people have mapped out using sound fundamental scientific
            principles the
            > space missions to jupiter (gallelio) and saturn (cassini), then
            developed them and
            > put them in action, that despite their glitches have been of great
            success,

            I for one think that humans are less predictable than planets and
            robots. I do know how to calculate gravitational acceleration as
            Galileo approaches Mars. But that's easier than predicting the
            influence of Antoine upon the Mavs.


            >... * sniff *.... i love ya' man....

            That's disgusting, Bob.


            > a model is a model. it does what its programmed to do...

            Your model seems to make the assumption that a player will do this
            year what he did last year, but in a different environment. If you
            can find a team whose lineup is exactly the same as the previous
            year's, you will still find significant changes in individual and
            team performances.

            As I said, you can focus on the most- or least-successful
            predictions. Even missions to Mars often go awry.
          • bchaikin@aol.com
            But what I find highly surprising is the Lakers winning only 54 games, in this sim.  Many people thought they d win 70-80 games.  Did you have to knock out
            Message 5 of 6 , Apr 2 9:07 AM
            • 0 Attachment
              But what I find highly surprising is the Lakers' winning only 54 games, in this sim.  Many people thought they'd win 70-80 games.  Did you have to knock out key players for substantial parts of your test?

              really? then many people would be overly optimistic. afterall how many times has a team in NBA history won even 70 games, let alone 70-80?...

              think about it, as great as the lakers were a few years back, shaq and kobe are slightly older now (shaq i think is about 32, kobe still a kid) and did not win a title last year despite each having outstanding seasons individually, and now added to them are 2 players that are 36 and 40 years old. as the sim shows, if each played upwards of 40 min/g, they would win 59-60 games, but i certainly did not think each would be playing 40 min/g all year long, but that rather barring injury PT would be down for each slightly until the playoffs. but even the great bulls teams of the mid-1990s that did win a ton of games (they won 72, 69, and 62 games in 3 consecutive seasons) did it with jordan, pippen, and rodman being only 33-35, 30-32, and 34-36 years of age during that time.....

              Did you have to knock out key players for substantial parts of your test?

              didn't for this simulation, but you can...

              I'd be surprised if a simulation would 'predict' Jermaine O'Neal's rise in production, upon moving to the Pacers. 

              jermain o'neal's player attributes (what he does once he gets the ball) have changed little in the last five years, even his scoring FG% (combining 2s, 3s, and FTs) has changed little in that time (its his worst now, was his best last season during those 5 years, but its basically fluctuated this and the past 4 seasons), and his shot blocking is as good as it was in 97-98. his playing time in the last 3 seasons has been pretty much the same. what's been different is that his touches per minute has increased each and every season, gradually, and while that can be hard to predict, the bottom line is that it increased each and every season over the past 4, so that the difference from one season to the next has been at most 10%-12%, meaning he has handled the ball more on offense each season by about that much....

              so that while over the span of 4-5 seasons you can say something like "..well how can you model that...", a change in touches/min of 40% over 4-5 seasons,  the truth of the matter is is that what he actually did - a change of 10%-12% in touches/min from one single season to the next - is not that uncommon, i.e. you can run a simulation at any time and easily change players' parameters such as touches/min by 10% (or even more) to try out different scenarios. players can have their touches/min fluctuate by as much as 10%-12% each year, but i will grant you that an increase in 4 consecutive years of 10%-12% each year is uncommon. but since the model can be updated yearly, it doesn't need to be able to predict a 40%-50% increase in touches/min in one single year by a single player who has played substantial minutes because it just doesn't happen (or happens very rarely, i just can't think of an example)...

              same with someone like kevin garnett. you look at his touches/min (its now about the same as it was in 98-99) and player attributes this and the past 5 seasons and there isn't much difference. his scoring FG% is also pretty much the same over that time, as is his playing time. so why is he so much better now than then? because he has incrementally increased his def rebounding each and every year to the point that he is now twice the def rebounder that he was in 96-97 (even his shot blocking is the same now as it was in 96-97), and i'm guessing his overall defense is probably better to (that's just a guess tho). but at least we can identify these reasons, and be exact about them...

              Or if the Mavs should be so weak, after their offseason roster changes. Stuff like that.


              as for the Mavs simulation clearly shows that one of the key reasons, probably the key reason, for their worse play compared to last season, has been their insistence to play antoine walker a ton of minutes, when both stats analysis (www.82games.com) and simulation show him to negatively affect that team compared to what other Mavs teammates can do in his place. the mavs would be better off giving his minutes to antawn jamison or shawn bradley, or even another team's benchwarmer, someone like brian cardinal if they could get him from golden state....

              take a look at what 82games.com shows for walker - other than being an average rebounder for a PF, his offense is awful, his defense is poor, and stats analysis and simulation shows he takes touches away from more productive players like nowitzki, finley, and jamison....

              And probably I'm not the only one wondering where was your prediction that the Lakers should win only 54 games, at the beginning of the season?  There's no shortage of after-the-fact "predictions".

              my, my.... hindsight is always 20-20. tell me, did you, or anyone else post predictions for all 29 teams for this season in this discussion group at anytime during this year? its always easy to knock someone else's predictions when you don't make any yourself (those who can't do - teach, those who can't teach - criticise)...

              the point is that the simulation showed after about 25-30 games into this season that basically, within a few games, what their record would be barring major injuries (grant replaced malone without too much difference) by season's end. but in actuality in the first week of january simulation had them winning only 52 games for the year, not 54. kobe's resurgence the 2nd half of this season has since increased that, so as the simulation is updated weekly with current stats those predictions change slightly for each team each week...

              back then the sim also showed sacramento on pace to win 60-62 games for the year, but we all know what happened to them. they bring in and start a player who hasn't played all season and one of their players having his career season (brad miller) goes down with an injury. you gonna expect a simulation to show this at the start of a season too?...

              I for one think that humans are less predictable than planets and robots.  I do know how to calculate gravitational acceleration as Galileo approaches Mars.... 

              the point is that there are many who believe that just because they themselves couldn't, or didn't, do it, doesn't mean it can't be done. my point is that if we can fly spacecraft to jupiter and saturn with relatively few glitches, why do people think simulation of pro hoops is any different, or so much harder? if brain tumors can be excised without damaging the integrity of the skull, is simulating pro basketball really any more different or difficult?...

              But that's easier than predicting the influence of Antoine upon the Mavs....

              again - says who? you? maybe for you it is, but i find the latter easier, much easier...

              >... * sniff *.... i love ya' man....

              That's disgusting, Bob.

              mmmmmmmmmmm......

              Your model seems to make the assumption that a player will do this year what he did last year, but in a different environment.... 

              its not an assumption, its pretty much a fact. typically players reach their peak performance in their 3rd-4th season, if they've played substantial minutes in each season, and that peak can be maintained until their 8th-9th seasons, prior to a decrease in performance due to age, injuries, etc...

              If you can find a team whose lineup is exactly the same as the previous year's, you will still find significant changes in individual and team performances.

              depends on how you define changes... i can show you that jermaine o'neal, other than his touches/min, is basically the same player now as he was 4-5 years ago. you can argue all you want about minor specifics, but that is a fact. a consistant increase year by year in touches/min is substantial, but the numbers show he's not much different once he gets the ball...

              As I said, you can focus on the most- or least-successful predictions.  Even missions to Mars often go awry.

              at least those of us that do make predictions are not the second guessers who never make them...

              bob chaikin
              bchaikin@...
            Your message has been successfully submitted and would be delivered to recipients shortly.