Announcement

Collapse
No announcement yet.

Fascinating findings from a study.

Collapse
X
 
  • Filter
  • Time
  • Show
Clear All
new posts

  • Fascinating findings from a study.

    OK...I just spent some time finding weighted means, standard deviations, z-scores, and therefore adjusted offensive wins created figures to account for the spread of offense through the years. This is far from perfect, given the fact that talent distributions aren't normally distributed and the typical spread of PCA ratings is a little bit skewed in a given season/league. There's also the problem that GREAT seasons will sometimes have the ability to impact the standard deviation of the league win scoring rate, so great players with a lot of great seasons will have a tendency to pull themselves down...although minimally.

    Just a first crack here and it raises a MAJOR eyebrow toward the possibility that Babe Ruth may not be lapping the field offensively as much as we think he is...

    The top 30 in standard adjusted offensive wins created (this is calculated assuming all seasons fall on the same bell curve...with the alltime win-scoring rate of 0.00633 Wins/PA and the alltime weighted standard deviation of 0.00614):
    Code:
    PlayerID	CAR AOWC
    cobbty01	265.20
    ruthba01	264.52
    bondsba01	258.57
    willite01	238.52
    henderi01	237.42
    musiast01	224.91
    aaronha01	222.69
    mayswi01	213.52
    wagneho01	213.05
    mantlmi01	207.43
    speaktr01	200.48
    ottme01 	199.65
    collied01	189.73
    hornsro01	188.69
    gehrilo01	187.89
    robinfr02	184.72
    morgajo02	175.97
    thomafr04	165.60
    yastrca01	164.91
    foxxji01	160.04
    jacksre01	156.24
    lajoina01	154.22
    rosepe01	153.56
    brettge01	153.09
    molitpa01	151.87
    murraed02	148.92
    matheed01	147.42
    kalinal01	147.22
    schmimi01	146.65
    palmera01	143.52
    I haven't attempted to plug that info into the GI method framework although doing so has the obvious advantage that the average scoring rate is always known so I don't have to go around calculating it. This is just career total wins, so obviously there are long-career biases, and I believe Ruth will still win the fight for "greatest" offensive player of all time...but look who leaped into first place on the career list...that would be one Tyrus Cobb.

  • #2
    ---------------
    Last edited by Sultan_1895-1948; 04-10-2006, 01:04 AM.

    Comment


    • #3
      Keep up the good work Matt......
      Are you ever going to publish your PCA work?

      Comment


      • #4
        Wow, so Cobb is in first. It's not that strange of a conclusion, though. The same thing happened with Win Shares. A lot of people don't know, Cobb is actually about 30 OWS ahead of Ruth on the career list, and also ahead in defensive Win Shares by about 20. So, going from only position players Cobb is actually ahead of Ruth in WS by about 50 I believe.

        It is only Ruth's time as a pitcher that lifts him over Cobb.

        Comment


        • #5
          Cobb was no doubt an absolute stud, and is more deserving on my #2 spot than anyone by far (including Wagner, no knock on him mind you), but what gives here.

          Don't WS have a lot to do with how many AB a player got? If so, it makes sense that Cobb will have more.

          I've shared my thoughts on their defense before.

          When they were at their best, Babe had the much better arm, was just as heady, and good with the glove, but didn't possess the sheer speed that Cobb had, even though he played people perfectly, got good jumps, and was quick in his early days. Cobb's arm went down the crapper and Babe's never did. Babe's legs went down the crapper starting in the early 30's, and Cobb's never really did, especially in comparison, despite the knee injury. Overall, because Cobb lasted longer and his legs held up much better, I'd give him an edge in fielding, although the advantage is slight, considering Babe's arm was much stronger, and he was just as heady/aggressive/smart in the field. When talking about fielding, not sure if that defensive WS thing takes into account Babe's fielding as a pitcher, which was very good.

          I'm not trying to knock Babe or Cobb just calling it as I see it. If there's anyone imo who has a case over Babe, it's Tyrus. My respect for his game will never wane.
          Last edited by Sultan_1895-1948; 04-10-2006, 05:30 AM.

          Comment


          • #6
            Originally posted by SABR Matt
            OK...I just spent some time finding weighted means, standard deviations, z-scores, and therefore adjusted offensive wins created figures to account for the spread of offense through the years.

            This is just career total wins, so obviously there are long-career biases, and I believe Ruth will still win the fight for "greatest" offensive player of all time...but look who leaped into first place on the career list...that would be one Tyrus Cobb.
            This is indeed a fascinating revelation (esp. considering a few years ago when we began interacting, I don't think Cobb was top 10 by PCA).

            This is just looking at offense, though, correct? What if defense were incorporated into the mix (I know that might be months of numbers crunching away, but clearly Cobb was more valuable defensively than Ruth.....).

            Comment


            • #7
              Okay I fail to see how this is fascinating. Put together virtually any offensive list and you are going to get the same names in roughly the same order. Raw runs created list gets you basically the same list, and I'm willing to bet if one factored in the offensive context that these players played in (basically turns run into wins) you would get a list even more similar to the one you did while jumping through all those hoops.

              Raw RC vs yours has 5 different names on it
              Waner
              Gehringer
              Simmons
              Winfield
              Bagwell


              While you have
              Frank Thomas
              Reggie Jackson
              Nap Lajoie
              Eddie Mathews
              Mike Schmidt

              With Thomas, Jackson, and Schmidt finishing 31 through 33 in raw RC.

              Comment


              • #8
                Thanks for the support Ubi...:\

                I wasn't claiming the list in and of itself was some kind of groudnbreaking leap in sabermetrics. If you knew me very well, you'd know that I'm gathering information right now in an attempt to move past this version of PCA entirely.

                What I found fascinating was the difference it made when you considered the spread of the distribution of run creation rather than just looking at the data in isolation. The initial list was offense...the thing I was keying on as fascinating...at least to me...was the changes from the original offensive leaderboard by PCA.

                To correct a minor point, to csh...Cobb was actually exactly tenth by the original PCA raw career leaderboard (11th in offense alone) when you first started talking to me. I remember you throwing a rather vocal fit to Rickey Henderson being 7th...three spots ahead of Cobb.

                In uncalibrated PCA, Cobb is 7th overall and 10th on offense alone...now he's 1st in calibrated wins created, and that's a pretty significant difference.

                I know Ubi that you've made it your personal mission to hate on seasonal sabermetrics...but for those of us who find the larger questions infinitely more fascinating than whether it's a good idea to bunt down by a run in the 9th inning...please try to keep the attitude to a minimum.

                Thanks.

                In any event...I completed the entire makeover of the GI method using calibrated offense, defense and pitcher win statistics. Big changes in some rather high profile places.

                Lefty Grove for instance went from 16th among all pitchers in GI to 4th...a look at the top 20 pitchers by the new GI...
                Code:
                First	Last    	Ps	P-Rk	CAR	RATE	PEAK	VAR	TOT
                Walter	Johnson 	SP	1	178.51	122.79	32.64	23.03	356.97
                Cy	Young   	SP	2	197.56	108.96	22.6	24.24	353.36
                Roger	Clemens 	SP	3	145	129.33	37.27	25.07	336.67
                Lefty	Grove   	SP	4	130.62	131.58	34.05	20.71	316.96
                Pete	Alexander	SP	5	152.11	118.77	22.25	17.17	310.3
                Christy	Mathewson	SP	6	136.87	115.25	27.46	25	304.58
                Greg	Maddux  	SP	7	118.9	114.9	26.79	24.98	285.57
                Carl	Hubbell 	SP	8	104.49	111.15	24.08	20.85	260.57
                Tom	Seaver  	SP	9	121.85	103.61	15.6	18.07	259.13
                Randy	Johnson 	SP	10	102.67	111.49	23.8	17.71	255.67
                Pedro	Martinez	SP	11	87.13	111.13	28.8	15.78	242.84
                Tim	Keefe   	SP	12	115.87	94.08	7.66	16.24	233.85
                Kid	Nichols 	SP	13	115	93.12	7.01	15.51	230.64
                Fergie	Jenkins 	SP	14	106.28	94.95	9.05	15.86	226.14
                Nolan	Ryan    	SP	15	114.41	85.95	5.87	17.91	224.14
                Dennis	Eckersley	SP	16	88.59	98.69	17.34	18.18	222.8
                Bert	Blyleven	SP	17	108.52	87.07	7.8	19.16	222.55
                Ed	Walsh   	SP	18	89.63	104.71	16.38	9.7	220.42
                Dazzy	Vance   	SP	19	86.29	100.67	16.25	14.38	217.59
                Bob	Gibson  	SP	20	94.13	95.89	10.98	16.34	217.34
                Also note Maddux dropped from 1st or 2nd to 7th.

                Other noteable changes...Ty Cobb moved from 7th to *2nd* in overall scoring...the top 30 overall, including a 35% positional rating bonus for catchers to accoutn for the physical demands of the position:
                Code:
                First	Last    	Ps	O-Rk	OFF	PIT	DEF	TOT
                Babe	Ruth    	UT	1	673.25	74.6	65.1	812.94
                Ty	Cobb    	CF	2	613.08	-0.14	92.26	705.2
                Barry	Bonds   	LF	3	628.58	0	64.95	693.54
                Ted	Williams	LF	4	614.22	0	50.37	664.6
                Rickey	Henderson	LF	5	541.24	0	98.38	639.62
                Honus	Wagner  	SS	6	510.05	0.04	111.14	621.23
                Tris	Speaker  	CF	7	446.12	0	155.47	601.59
                Willie	Mays    	CF	8	470.5	0	117.68	588.18
                Stan	Musial  	UT	9	499.34	0	83.56	582.9
                Mickey	Mantle  	CF	10	523.09	0	53.74	576.82
                Hank	Aaron   	RF	11	461.79	0	82.79	544.58
                Rogers	Hornsby 	2B	12	478.45	0	64.03	542.47
                Eddie	Collins 	2B	13	414.53	0	116	530.53
                Lou	Gehrig  	1B	14	465.59	0	44.68	510.27
                Mel	Ott     	RF	15	455.95	0	36.92	492.86
                Joe	Morgan  	2B	16	403.87	0	78.74	482.61
                Nap	Lajoie  	2B	17	368.27	0	113.73	482
                Frank	Robinson	UT	18	402.96	0	71.98	474.94
                Jimmie	Foxx    	1B	19	393.36	3.26	58.31	454.93
                Frank	Thomas  	1B	20	424.72	0	19.75	444.47
                Walter	Johnson 	SP	21	68.75	356.97	6.52	432.24
                Billy	Hamilton	CF	22	350.08	0	78.29	428.37
                Mike	Schmidt 	3B	23	337.68	0	78.45	416.13
                Al	Kaline  	RF	24	312.77	0	101.68	414.45
                Reggie	Jackson 	RF	25	353.84	0	58.91	412.75
                Carl	Yastrzemski	LF	26	333.26	0	79.14	412.4
                George	Brett   	3B	27	344.53	0	66.73	411.26
                Ed	Delahanty	UT	28	350.46	0	57.65	408.12
                Cy	Young   	SP	29	30.82	353.36	19.08	403.26
                Joe	DiMaggio	CF	30	316.98	0	82.51	399.49
                The major problem (and I am aware of this) is that the linear weights used in the original PCA put too much emphasis on stolen bases...my more current research suggests that stolen bases are often worth far less than I have them worth in this model...expect Rickey Henderson, Max Carey, Billy Hamilton and Joe Morgan to drop lower on these lists if/when I incorporate more accurate linear weights based on my current reading.

                Overall...the look of the ratings after I made the adjustment to account for win scoring distributions looks more in line with traditional lines of thought...which may or may not be a good thing...I'll leave that up to readers to decide. :\

                Comment


                • #9
                  personal mission? Hardly if it was a lot in my life would be very different then it is today. I know you are sensitive to any criticisms to your baby so I'll keep it brief. You are not inventing the wheel, all you are doing is going into a shed somewhere for a few years and coming out with a wheel and going "tadaah". And the people are saying, "yep thats a wheel alright, we've had them for a long time now". PCA in terms of offense will not answer any large questions that have gone unanswered. By the way what large offensive questions have gone unanswered anyway? You've spent a long time working on something that basically tells us the same answer that a dozen other offensive metrics are already telling us, and you are still not finished. But what is the end product going to tell us that we don't know already? That Rogers Hornsby was worth 20.2 wins in 1921 instead of 20.4 like we thought? Does it make a difference?


                  By the way large questions and large problems are made up of small questions and small problems and a large answer is generally a over-simplification of the real answer and one that is usually not all that accurate.

                  Why did the US fight in WWII?
                  Large Answer: Because of Pearl Harbor or if you like the growing belligerence from developing nations towards established powers.

                  I look at metrics and ask what can this metric do that others cannot? I have yet to see what PCA can do that others cannot. So then what is the value of it? What makes it unique so that one would want to use it or employ those who do use it? If I'm a baseball owner why should I pay you for your system? Why should I not use something that is freely available like Win Shares, TPR, WARP, or any of the other publicly available metrics? And more to the point what use does a major league team have for PCA to the point where they should spend money on it?

                  Perhaps thats not the goal so then what is the reason us baseball fans should use it over anything else? What is it telling us that other metrics are not?

                  Again in either case I don't see an answer that points in favor of PCA. Thats not to say that your system is inaccurate or useless just that it is another metric in a room full of hundreds of similar metrics.

                  Comment


                  • #10
                    Originally posted by 538280
                    Wow, so Cobb is in first. It's not that strange of a conclusion, though. The same thing happened with Win Shares. A lot of people don't know, Cobb is actually about 30 OWS ahead of Ruth on the career list, and also ahead in defensive Win Shares by about 20. So, going from only position players Cobb is actually ahead of Ruth in WS by about 50 I believe.

                    It is only Ruth's time as a pitcher that lifts him over Cobb.
                    Actually, Cobb had more PA's and played in more games as a hitter than Ruth.

                    Ruth is still the alltime leader in batting WS per-game, by a pretty large margin.

                    Comment


                    • #11
                      Originally posted by Taco De Muerte
                      Actually, Cobb had more PA's and played in more games as a hitter than Ruth.
                      Right, but that's to Ty's credit that he kept himself in shape and still had an excellent season for a 40 year old who had been in the league 23 years (and then still managed a pretty good season for a 41 year old in 1928). Ruth was completely done as a player by the time he was 40 and he came up later than Cobb.

                      Comment


                      • #12
                        But if Babe Ruth didn't start off as a pitcher his PA would be very very close to Ty Cobb's total.

                        Comment


                        • #13
                          Originally posted by Ubiquitous
                          personal mission? Hardly if it was a lot in my life would be very different then it is today. I know you are sensitive to any criticisms to your baby so I'll keep it brief. You are not inventing the wheel, all you are doing is going into a shed somewhere for a few years and coming out with a wheel and going "tadaah". And the people are saying, "yep thats a wheel alright, we've had them for a long time now". PCA in terms of offense will not answer any large questions that have gone unanswered. By the way what large offensive questions have gone unanswered anyway? You've spent a long time working on something that basically tells us the same answer that a dozen other offensive metrics are already telling us, and you are still not finished. But what is the end product going to tell us that we don't know already? That Rogers Hornsby was worth 20.2 wins in 1921 instead of 20.4 like we thought? Does it make a difference?


                          By the way large questions and large problems are made up of small questions and small problems and a large answer is generally a over-simplification of the real answer and one that is usually not all that accurate.

                          Why did the US fight in WWII?
                          Large Answer: Because of Pearl Harbor or if you like the growing belligerence from developing nations towards established powers.

                          I look at metrics and ask what can this metric do that others cannot? I have yet to see what PCA can do that others cannot. So then what is the value of it? What makes it unique so that one would want to use it or employ those who do use it? If I'm a baseball owner why should I pay you for your system? Why should I not use something that is freely available like Win Shares, TPR, WARP, or any of the other publicly available metrics? And more to the point what use does a major league team have for PCA to the point where they should spend money on it?

                          Perhaps thats not the goal so then what is the reason us baseball fans should use it over anything else? What is it telling us that other metrics are not?

                          Again in either case I don't see an answer that points in favor of PCA. Thats not to say that your system is inaccurate or useless just that it is another metric in a room full of hundreds of similar metrics.
                          I certainly grant that offensive metrics have come so far that actual advancement is limited in the field of pure empirical analysis, but I haven't spent "years" developing an OFFENSIVE metric. The offensive metric only develops/changes in that I go through drafts of how best to approach the ENTIRE method. PCA isn't an offensive metric exclusively. And I happen to believe work on trhe FSIA (which is something I've devoted a big chunk of my time to in the last six months) and PCA's DEFENSIVE analysis can and will teach something that has not already been taught.

                          Just what the hell have YOU done to advance the study of the game Ubi? Aside from sitting back and picking at the rest of the field. BTW, if you think I'm "sensative" to criticism about my work, then you don't know me as well as you claim, because all I've done in the last two years especially is listen to criticism and attempt to learn something from. I tend to put up a fight, but the changes in my thinking over the years should tell you that I am not sitting here closed off to criticism.

                          Comment


                          • #14
                            Just what the hell have YOU done to advance the study of the game Ubi? Aside from sitting back and picking at the rest of the field.
                            I've done as much as you have which is basically nothing in the large scheme of things. I've taught a few people on the person to person level a few things about the game. I've relayed information that was unknown or unavailable to some. It may not be up there with Bill James resume buts it better then some and not the worst possible.
                            As for sitting back and picking I tend only to pick that which I believe is not correct. If someone says 2+2=4 I'm not going to nitpick it. If someone says that Ross Young was better then Rogers Hornsby in a certain year I am going to pick at their work.




                            PCA's DEFENSIVE analysis can and will teach something that has not already been taught.
                            And that would be. . .

                            Comment


                            • #15
                              The Ross Youngs thing was quickly corrected because we agreed with you that it was odd and it led to the discovery of a problem in the James-style analysis of offense as team-dependent...I certainly didn't have a problem with you bringing that up as a problem, because it was one. But what value have you added to any discussion by spending time to say "I don't care about this, it's uninteresting"...yeah...thanks for your input...I'd say more but it wouldn't be G rated/

                              Comment

                              Ad Widget

                              Collapse
                              Working...
                              X