SoxProspects News
|
|
|
|
Legal
Forum Ground Rules
The views expressed by the members of this Forum do not necessarily reflect the views of SoxProspects, LLC.
© 2003-2024 SoxProspects, LLC
|
|
|
|
|
Forum Home | Search | My Profile | Messages | Members | Help |
Welcome Guest. Please Login or Register.
How many top prospects = lower good prospects?
|
Post by jchang on Jan 30, 2014 17:06:07 GMT -5
thanks, I will grind through the bWAR/fWAR details later as I should maintain the appearance of working (any complex Excel spreadsheet qualifies, but not looking at the BR and FG websites). Anyways, my maintain point is that the excellent article by Scott McKinney in RoyalsReview/SB (great information, as I have certainly inquired whether such information was available) over-estimates the bust rate of pitching prospects. Using Baseball Reference pitching WAR, I separated the starters from the RP, then restored the sort by WAR, then made sub-groups of 30 (30 MLB teams, even though talent is not evenly distributed) For Starting Pitchers, below is the WAR range for the top 30, 2nd 30, 3rd 30 etc
SP bWAR range 1-30 3.6 to 7.8 31-60 2.3 to 3.4 61-90 1.4 to 2.2 91-120 0.6 to 1.4 121-150 -0.3 to 0.5
For Relief Pitchers (0-3 starts), below is the WAR range for the top 30, etc
RP bWAR range 1-30 1.7 to 3.6 31-60 1.2 to 1.7 61-90 0.8 to 1.2 91-120 0.5 to 0.7 121-150 0.3 to 0.5
I am inclined to guess that ERA translates to win probability with other factors, and IP * WP translates to WAR? My point is that I think the rating/grade system is valid: 8 elite, 7 all-star, 6 above avg., 5 avg. Grade 6 can be either a 2/3 starter or 1st div closer Grade 5 can be either a 4/5 starter or late inning RP Grade 4 a good RP Yet the bWAR system puts 4/5 starters in the below avg. range (grade 3 to 4). The RP WAR values seem to be consistent with the grading principle.
|
|
|
Post by Chris Hatfield on Jan 30, 2014 19:07:47 GMT -5
Riffing a bit here, so please correct me if I'm wrong here, jchang or others.
I like that analysis because it would seem to bear out what some of us here repeat often: that the best way to think of a "no. 1 starter" is not to think of a top-30 starter, a no. 2 as one between 31 and 60, etc., as using this system above, you wind up with a no. 5 starter as being replacement level.
This can be confusing. One line of thinking might say, sure, a no. 5 is replacement level, right? But a no. 5 starter as the term is usually used doesn't mean replacement level. Usually, it means a guy you can count on to be in the back of a rotation, not a guy at replacement level (defined as a player you can acquire for no cost, i.e., league minimum). We all know that a number 5 starter isn't replacement level in that sense.
jchang, is there any chance you could make a graphical representation of the data? I'm thinking something like a bar graph showing the number of pitchers with each WAR on the year?
EDIT: As for the original concept of the thread (which I guess has come completely derailed?), I'm not on board with the logic, even if the math may be interesting and correct. Give me one Bogaerts over the entire group of prospects 71-100. Only the tiniest bit of hyperbole in that statement.
|
|
|
Post by jchang on Jan 30, 2014 20:25:49 GMT -5
I don't think I have derailed this thread - top prospect to lower prospect value using the success rate using the data in the McKinney article. I am only trying to clarify the assessment of the failure rate for pitching prospects based on WAR definitions, which seems to value SP and RP out of line relative to the definitions of the 20-80 grade scale. - yeah that my excuse. I am fully on board with grade 8 - 1 pitcher being elite, and that there are only a handful of elite pitchers, far less than 30, and this is reflected in the WAR list. My grouping in batches of 30 is purely to see whether we could fill the MLB roster spots according to definition. Also that grade 7 - all-star is good as there enough WAR 5 players to fill the all-star roster for both position and pitchers (there were 18 pitchers at WAR 4.4 and above in 2013, so the remaining pitching spots are for elite closers?) The tricky part is what grade corresponds to a 2-pitcher? I would like to say that a 3-pitcher should be grade 6 - above average, same as a 1st div position player, and 4/5 pitchers being grade 5. The BR WAR definition is 2+ Starter, 0-2 Sub, <0 Repl. So I would like to say that WAR 1.5-ish should be a decent 5 starter. Finally, the last 1-2 RP should represent the replacement level line. the Excel sheets with graph are below www.qdpma.com/tools/WAR_pitching_2013.xlsxwww.qdpma.com/tools/WAR_position_2013.xlsxoops I did the graphs differently.
|
|
|
Post by jchang on Jan 30, 2014 20:47:39 GMT -5
here is distribution by WAR for position, SP and RP
WAR Pos SP RP 0-0.9 xx 35 205 1-1.9 69 26 52 2-2.9 58 24 19 3-3.9 29 23 5 4-4.9 25 9 5-5.9 15 6 6-6.9 9 5 7+ 7 3
there were a few pitchers who started 4-9 games with positive WAR but were excluded from my SP/RP separation method
|
|
|
Post by jchang on Jan 30, 2014 21:29:19 GMT -5
below is the bWAR of starting pitchers in descending order, in groups of 30. SP1 does not represent a 1-pitcher, just the top 30.
The purpose of the grouping of 30 is to test whether the 30 team MLB roster can be filled in accordance to the bWAR definition: starter being WAR 2+, of course not all teams have 5 starters.
below are relief pitchers by bWAR in descending order, groups of 30
below are position players by bWAR in descending order, groups of 30
finally, the number of players: position, SP and RP by WAR bracket the number of low WAR (0-0.9) position players is difficult to assess as there are lots of players with few AB and low WAR. The main point is that the bWAR system is unfair to SP, as there are fewer than 150 (5 x 30) starting pitchers (10+ starts) with WAR > 0. A perfecting viable 5-starter might have negative WAR.
|
|
|
Post by dewey1972 on Jan 30, 2014 23:24:18 GMT -5
below is the bWAR of starting pitchers in descending order, in groups of 30. SP1 does not represent a 1-pitcher, just the top 30.
The purpose of the grouping of 30 is to test whether the 30 team MLB roster can be filled in accordance to the bWAR definition: starter being WAR 2+, of course not all teams have 5 starters.
below are relief pitchers by bWAR in descending order, groups of 30
below are position players by bWAR in descending order, groups of 30
finally, the number of players: position, SP and RP by WAR bracket the number of low WAR (0-0.9) position players is difficult to assess as there are lots of players with few AB and low WAR. The main point is that the bWAR system is unfair to SP, as there are fewer than 150 (5 x 30) starting pitchers (10+ starts) with WAR > 0. A perfecting viable 5-starter might have negative WAR. But this is the point: lots of teams throw a fifth (and often fourth) starter out there who IS a replacement level pitcher, someone who performs no better than the team's AAA starters. Furthermore, we need to keep in mind how often pitchers get hurt. Those injuries create lots of starts for pitchers who ARE replacement-level, since they need to be in the minors or in the bullpen until the injuries occur. We shouldn't expect 150 quality starters. By the way, this doesn't change much if we look at fWAR. The 121st starter in the past year was Wily Peralta with 1 win and the 150th was Kevin Slowey at 0.4 WAR.
|
|
|
Post by dewey1972 on Jan 30, 2014 23:30:40 GMT -5
WAR doesn't correlate with ERA because it's not based on ERA. That simple, really. Also, just a suggestion for the future, it might be a good idea to say whether you're using bWAR or fWAR, as they're calculated differently. Well, bWAR is based on runs allowed, which correlates pretty well to ERA. And Fangraphs also has RA-9 WAR, which is based on runs allowed.
|
|
|
Post by dewey1972 on Jan 30, 2014 23:52:31 GMT -5
I find the WAR value of pitchers to be totally baffling player IP ERA bWAR Buchholz 108 1.74 4.3 Uehara 74.1 1.1 3.6 Lester 213 3.67 3.0 Lackey 189 3.52 2.8 Breslow 59.2 1.84 1.6 Peavy 144 4.17 1.4 Tazawa 68.1 2.94 1.0 Doubront 162 4.08 0.9 Dempster 171 4.59 -.2 I would think an elite starter (top 5 in all MLB) might have a 2.5 ERA,
while the top 30 have ERA below 3.25, the top 60 starters have ERA below 3.75 and top 90 starter ERA below 4.25 for RP, elite - ERA 1.25, top 30 ERA 2.0, and top 90 ERA 3.0
Buchholz was on pace for an elite season until the injury, so his WAR 4.3 make sense. Lester and Lackey at WAR 3.0 and 2.8 makes sense as being above average. Peavy was above average but short on IP so his WAR 1.6 makes sense. Dubrount was also above average so I think he should be WAR 2.0+ Dempster is slightly below average for a starter, but should correspond to grade 5 - which would be WAR 2 for a position player. Breslow rates as "late inning RP", also grade 5, i.e. WAR 5 if he were a position player. Tazawa seems right as "reliable RP" - grade 4, WAR 1.
Doubront only pitched 162 innings last year, which is one reason his WAR isn't higher. Secondly, he wasn't above average. He had a 4.32 ERA and league average was 3.87. Also, bWAR adjusts for the quality of the opponent and the parks that the pitcher pitched in. I have no idea whether this impacts him, but it seems likely, given that his Fangraphs RA-9 WAR is higher than his bWAR (1.8 compared to 0.9). As for Dempster, his ERA was way below average, at 4.57. Though it is worth noting that his Fangraphs RA-9 WAR is again about a win higher than his bWAR (0.9 to -0.2).
|
|
ericmvan
Veteran
Supposed to be working on something more important
Posts: 8,925
|
Post by ericmvan on Jan 31, 2014 0:35:14 GMT -5
I've always regarded the top 15 starters as #1s. The idea being that below-average clubs lack a true #1, almost by definition.
Then take the next 30 guys as #2, and so on. You end up with the below average clubs typically having a 2, 3, 4, 5, and replacement level starter, while the good clubs have a 1 through 5.
Oh, and fWAR for pitchers is utter garbage and should always be ignored.
|
|
|
Post by patrmac04 on Jan 31, 2014 1:26:26 GMT -5
I don't think I have - top prospect to lower prospect value using the success rate using the data in the McKinney article. I am only trying to clarify the assessment of the failure rate for pitching prospects based on WAR definitions, which seems to value SP and RP out of line relative to the definitions of the 20-80 grade scale. - yeah that my excuse. I am fully on board with grade 8 - 1 pitcher being elite, and that there are only a handful of elite pitchers, far less than 30, and this is reflected in the WAR list. My grouping in batches of 30 is purely to see whether we could fill the MLB roster spots according to definition. Also that grade 7 - all-star is good as there enough WAR 5 players to fill the all-star roster for both position and pitchers (there were 18 pitchers at WAR 4.4 and above in 2013, so the remaining pitching spots are for elite closers?) The tricky part is what grade corresponds to a 2-pitcher? I would like to say that a 3-pitcher should be grade 6 - above average, same as a 1st div position player, and 4/5 pitchers being grade 5. The BR WAR definition is 2+ Starter, 0-2 Sub, <0 Repl. So I would like to say that WAR 1.5-ish should be a decent 5 starter. Finally, the last 1-2 RP should represent the replacement level line. the Excel sheets with graph are below www.qdpma.com/tools/WAR_pitching_2013.xlsxwww.qdpma.com/tools/WAR_position_2013.xlsxoops I did the graphs differently. You aren't derailing this thread at all. I think you are speaking towards the inconsistencies in the two original articles I posted. That is where the conversation was being driven to to see if we can better define success so we can have a better picture. I really liked the graphs as they visually helped quite a bit. I also really appreciate the saber posts that spoke towards alternative metrics as it was what I was curious about. All good stuff to me! Sent from my SGH-T999 using proboards
|
|
|
Post by jchang on Jan 31, 2014 2:48:10 GMT -5
The two articles patrmac04 started this thread with on prospect success rate were great. I had been interested in this subject for a while but did not have data to put it together + laziness. I had expected that success rate may only be 50% or so, consistent with the expression that one proven player (in hand) is worth 2 prospects (in the bush). What bothered me was the much lower success rate (based on the standard WAR 2 or 1.5 being a success) for pitchers compared to position players. Is something wrong with the way we evaluate pitching prospects?
The 51-100 prospects this year are all grade 55, and I am assuming that this is common. This should correspond to either a 5 starter or closer by the grade definition. There are typically 12 pitchers on an MLB roster. WAR definition is 2 - average or SP. I would expect that the average pitcher at WAR 2 should mean the 5 starter and best RP (6 of 12 pitchers at WAR 2 or above, 6 pitchers below, hence WAR 2 being average). It is the last one or two RP that should constitute the replacement level player valuation. And yet the WAR rating puts both the 5-starter and the 5th best RP at about the same value, Below is a distribution chart for SP, bWAR on the vertical axis, and RA9 on the horizontal
Below, the same for RP There were 186 pitchers who started 10 or more games in the top graph. There were 258 pitchers with >= 20IP but <= 3GS (ok, this may have a couple of spot starters). There a number pitchers with between 4 and 9 starts excluded from the above, 7 of whom between 0.9 to 2 bWAR.
Note the WAR 0 point is about 5.0 RA9 for SP and 4.6 RA9 for RP (eyeball, stats help would be appreciated). I would argue that the 5-starter with 165 IP over 30 starts (5.5 IP/game) at 5 RA9 should indeed be considered MLB average value for WAR 2 (ok, perhaps 1.5), while the RP good for 60 IP at 4.6 RA9 might be just above replacement level. Another curiosity: MLB 30 team total bWAR
BR bWAR WAR/team Avg WAR/roster spot pitching 410.0 13.67 1.14 batting 598.6 19.95 1.53 This would mean that the argument of WAR 2 = avg is untenable, also contributing to the low success rate of prospects using WAR 1.5 as the standard of success? Edit: actually this could work? See bottom for pitchers. Of the WAR 20 per average team for batters, assume the 9 starters average WAR 2, and the 4 bench average WAR 0.5?
Below is the bWAR + fWAR (previously I had just bWAR) Starting pitchers by WAR descending in groups of 30
SP bWAR range fWAR range 1-30 3.6 to 7.8 3.4 to 6.5 31-60 2.3 to 3.4 2.4 to 3.3 61-90 1.4 to 2.2 1.6 to 2.3 91-120 0.6 to 1.4 1.1 to 1.5 121-150 -0.3 to 0.5 0.4 to 1.1 Same for RP RP bWAR range fWAR range 1-30 1.7 to 3.6 1.2 to 3.3 31-60 1.2 to 1.7 0.8 to 1.2 61-90 0.8 to 1.2 0.6 to 0.8 91-120 0.5 to 0.7 0.3 to 0.6 121-150 0.3 to 0.5 0.2 to 0.6 151-180 0.1 to 0.3 0.0 to 0.2 Below is the number of players by bWAR bracket bWAR Pos SP RP 0-0.9 xx 35 124* 1-1.9 69 26 52 2-2.9 58 24 19 3-3.9 29 23 5 4-4.9 25 9 5-5.9 15 6 6-6.9 9 5 7+ 7 3
*(min 20 IP) same by fWAR bracket fWAR Pos SP RP 0-0.9 182 43 141 1-1.9 73 53 39 2-2.9 53 30 6 3-3.9 38 22 2 4-4.9 19 11 5-5.9 13 4 6-6.9 7 6 7+ 6 Updated Excel sheets to include fWAR, and to adjust the RP excluded pitchers with less than 20 IP) www.qdpma.com/zBaseball/WAR_pitching_2013b.xlsx www.qdpma.com/zBaseball/WAR_position_2013b.xlsx
Edit 31 Jan 6:49PM
Below, SP: IP by pitchers that started 1 or more games (308), then broken out by IP in GS, and in relief, RA are apportioned RP: IP by pitchers 0 starts (418 pitchers), IP in relief is then totaled IP IP_st IP_re RAst RAre RA9st RA9re RA9 IP/GS bWAR SP 30648.7 28676.3 1972.3 13823 951 4.338 5.90 272.5 RP 13004.7 4?? 13000.7 0 5479 3.793 137.8 All 43653.3 14973 6430 3.865 4.18 410.3
Pitchers with 1 or more starts have RA9 = 4.338 (based on apportioned RA by start/relief mix), and the composite RA9 in relief inning is 3.865, and the overall RA9 (all IP) is 4.18. The net result is that starters pitched 70.2% of the IP, while getting only 66.4% of the WAR. Given the SP total bWAR 272.5 over 30 teams and 5/team, the average SP is bWAR is 1.82, and the average RP bWAR is 0.66.
|
|
|
Post by jchang on Jan 31, 2014 10:36:25 GMT -5
One addendum. It would appear that (b&f))WAR is a complex calculation based on IP x win percentage above replacement (which I will call W%), with W% strongly influenced by RA9?. This would mean that a SP with 165 IP has about the same value as a RP with 60 IP if both have W% near zero. This seems to be the case for the 5th group of 30 SP and for the 5th group of 30 RP. My argument is that the 5th starter is more valuable than the RP when their WAA% are both near zero. The 5th starters job is to grind through enough innings so that even when he is getting hammered, the marginal RP's can finish the game and allowing the better RP's to be available to finish games when there is a lead to protect. This would mean WAR calculation must employ situational RA9?
|
|
|
Post by dewey1972 on Jan 31, 2014 21:54:19 GMT -5
Note the WAR 0 point is about 5.0 RA9 for SP and 4.6 RA9 for RP (eyeball, stats help would be appreciated).
I would argue that the 5-starter with 165 IP over 30 starts (5.5 IP/game) at 5 RA9 should indeed be considered MLB average value for WAR 2 (ok, perhaps 1.5), while the RP good for 60 IP at 4.6 RA9 might be just above replacement level. Another curiosity: MLB 30 team total bWAR
BR bWAR WAR/team Avg WAR/roster spot pitching 410.0 13.67 1.14 batting 598.6 19.95 1.53 This would mean that the argument of WAR 2 = avg is untenable, also contributing to the low success rate of prospects using WAR 1.5 as the standard of success.
I still don't understand why a fifth starter would be average. Shouldn't a third starter be average, by definition? We don't expect every starting third basemen to be an average player. We expect roughly half of them to be above average and half of them to be below average.
|
|
|
Post by dewey1972 on Jan 31, 2014 21:56:57 GMT -5
I've always regarded the top 15 starters as #1s. The idea being that below-average clubs lack a true #1, almost by definition. Then take the next 30 guys as #2, and so on. You end up with the below average clubs typically having a 2, 3, 4, 5, and replacement level starter, while the good clubs have a 1 through 5. Oh, and fWAR for pitchers is utter garbage and should always be ignored.Could you clarify why you think this? There seem to be a fair amount of smart analysts who disagree.
|
|
|
Post by jchang on Jan 31, 2014 22:48:13 GMT -5
if there are 5 starting pitchers, then the 3-pitcher should be the middle (or average) starter in terms of value. However, there are 12 pitchers on 25-man MLB roster. The 5 starters and the closer should be the best 6 out of 12 in talent/value, and other 6 RP should be least talent/value, hence the 5 SP + closer are above average pitchers and the 6 RP are below average. Depending on the difference between the 5 starter and closer, one should be close to being an average MLB pitcher, hence map to a grade of 5 (elite closers are 6). The 3 starter is an average starter (3 of 5) but an above average pitcher as he is 3 of 12, hence the grade of 6.
I would say that we should assess position players in the same manner. Assume an average team. There are 13 position players including the DH, 9 starters and 4 bench. Presumably 6 are above average, one is average, and 6 are below average. Hence only 7 of the starters are average or better, 2 in principle are below average along with the 4 bench players, barring a coach decision to do quirky alignment (NBA 6th man strategy?)
Apparently bWAR pitching is based on RA9, while fWAR is based on FIP? which attempts to remove luck?
Given that SP have higher RA9 than RP, I say that we need to rethink how pitching WAR should be calculated focusing first on correctly assessing starter pitching to properly reflect that between 2 starters with the same RA9, the SP that can sustain higher IP/GS is more valuable, not just in the higher season IP.
Lets assume that baseline Win is the following sequence. The starter pitches his average IP/GS and avg RA per outing. From here we calculate the probability that he has the lead against the opponents average SP. If the starter does 7 IP, then he is followed by the setup man (RA9 2.6?) and then closer (RA9 1.8?). Because his team has the lead, these last two innings are against below average RP (RA9 4?) If the starter completes 6 IP, he is followed by the 3rd RP, who might be RA9 3+. But given that the average starter has RA9 4.3, there needs to be some mechanism to assign a higher value to the starter who can put up more IP/GS, even though his RA9 is lower than the better RP?
|
|
|
Post by ethanbein on Feb 1, 2014 15:33:39 GMT -5
Hey guys, my first post here but I think I have a few things to add. On the difference between SP and RP WAR, there are a few things to keep in mind. One is pretty obvious, which is that starters throw a lot more innings than relievers. Koji threw quite a few last year, but still ended up under 74 IP. The best starters throw 3 times that many, so on a per-inning basis 3.3 WAR is huge. Another is that it's just a lot easier to be a reliever than a starter, so the replacement levels are different. A generic replacement level reliever from the minors can be expected to give nearly a league average ERA (~4.00), so any reliever that can't beat that by much isn't going to be very valuable. On the other hand, a starter with a 4.00 ERA can be hugely valuable (see Doubront, Felix). Another thing to keep in mind is that RP WAR is adjusted for leverage. So pitchers that throw in more important spots (closers) end up with higher WAR than a pitcher who pitches just as well in less important spots. In general though, I think looking at low elite RP WAR values and concluding that WAR undervalues them is a bit simplistic. RPs, even elite ones, just really aren't that valuable. I think that even with a replacement level closer instead of Koji, the Red Sox still win the division last year. Closers do become more important in the playoffs, though. As for fWAR vs bWAR, bWAR essentially starts with runs allowed and then adjusts for team defense. The problem is that it adjusts for total team defense, essentially assuming that every starter gets the same defense behind him, which is crazy. We know that different pitchers get drastically different run support just based on luck, so why wouldn't the same be true for defense? Fangraphs provides one WAR based on FIP and one based on runs allowed. Usually the FIP one is more accurate, but the RA one is there in cases where you think it's more accurate, especially if you have a larger sample size, say a whole career. Rivera for example beat his FIP year after year. This is a good article on the topic of fWAR vs bWAR for pitchers: www.fangraphs.com/blogs/why-our-pitcher-war-uses-fip/
|
|
|
Post by fenwaythehardway on Feb 1, 2014 18:15:56 GMT -5
Riffing a bit here, so please correct me if I'm wrong here, jchang or others. I like that analysis because it would seem to bear out what some of us here repeat often: that the best way to think of a "no. 1 starter" is not to think of a top-30 starter, a no. 2 as one between 31 and 60, etc., as using this system above, you wind up with a no. 5 starter as being replacement level.
This can be confusing. One line of thinking might say, sure, a no. 5 is replacement level, right? But a no. 5 starter as the term is usually used doesn't mean replacement level. Usually, it means a guy you can count on to be in the back of a rotation, not a guy at replacement level (defined as a player you can acquire for no cost, i.e., league minimum). We all know that a number 5 starter isn't replacement level in that sense.jchang, is there any chance you could make a graphical representation of the data? I'm thinking something like a bar graph showing the number of pitchers with each WAR on the year? EDIT: As for the original concept of the thread (which I guess has come completely derailed?), I'm not on board with the logic, even if the math may be interesting and correct. Give me one Bogaerts over the entire group of prospects 71-100. Only the tiniest bit of hyperbole in that statement. There is no more useless exercise in baseball analysis than trying to define a rigid statistical definition of a #1/2/etc starter. These classifications are not meant to be rigid, precisely defined or unambiguous. They are roughly defined shorthand, that's all. If you want precision, talk in terms of WAR, FIP variants, etc. Use the right tool for the right job. Trying to use the #1/2/etc as a precise measure of pitcher value is like using your screwdriver as a chisel. It can work in a pinch, but why on earth would you do it when you have access to any number of perfectly good chisels? What does it gain the analyst community to create a five-bin system to define pitcher quality? It would be like designing a version of WAR where 1, 2, 3, 4, and 5 are the only possible scores, and everything just gets rounded off into one of those grades. That would be insane, right? Well, that's essentially what people are doing when they get into these discussions about what exactly a #3 starter is.
|
|
|
Post by dewey1972 on Feb 1, 2014 22:55:32 GMT -5
if there are 5 starting pitchers, then the 3-pitcher should be the middle (or average) starter in terms of value. However, there are 12 pitchers on 25-man MLB roster. The 5 starters and the closer should be the best 6 out of 12 in talent/value, and other 6 RP should be least talent/value, hence the 5 SP + closer are above average pitchers and the 6 RP are below average. Depending on the difference between the 5 starter and closer, one should be close to being an average MLB pitcher, hence map to a grade of 5 (elite closers are 6). The 3 starter is an average starter (3 of 5) but an above average pitcher as he is 3 of 12, hence the grade of 6. I would say that we should assess position players in the same manner. Assume an average team. There are 13 position players including the DH, 9 starters and 4 bench. Presumably 6 are above average, one is average, and 6 are below average. Hence only 7 of the starters are average or better, 2 in principle are below average along with the 4 bench players, barring a coach decision to do quirky alignment (NBA 6th man strategy?) Apparently bWAR pitching is based on RA9, while fWAR is based on FIP? which attempts to remove luck?Given that SP have higher RA9 than RP, I say that we need to rethink how pitching WAR should be calculated focusing first on correctly assessing starter pitching to properly reflect that between 2 starters with the same RA9, the SP that can sustain higher IP/GS is more valuable, not just in the higher season IP. Lets assume that baseline Win is the following sequence. The starter pitches his average IP/GS and avg RA per outing. From here we calculate the probability that he has the lead against the opponents average SP. If the starter does 7 IP, then he is followed by the setup man (RA9 2.6?) and then closer (RA9 1.8?). Because his team has the lead, these last two innings are against below average RP (RA9 4?) If the starter completes 6 IP, he is followed by the 3rd RP, who might be RA9 3+. But given that the average starter has RA9 4.3, there needs to be some mechanism to assign a higher value to the starter who can put up more IP/GS, even though his RA9 is lower than the better RP? Ethan did a great job with his thoughts about WAR and the differences. I want to comment on two things here. The first is that I think our arguments about "average" are beside the point. WAR is based on replacement level. It happens to be that a 2 WAR player is an average player, but that doesn't have to be so. WAR shouldn't be adjusted so that an average player is at a set point, be that 2 wins or something else, it should be set so that a replacement level player who is freely available is worth 0 WAR. This is the crux of the matter. That being said, when we say "an average player" we don't mean "an average major leaguer" we mean an "average full-time player." The FanGraphs WAR link under the glossary states this. For context, last year there were 25 first basemen listed under "qualified" on the FG leaderboard, which gives us a pretty good estimate of full-time players. #14 had 2.2 wins. #15 had 1.9 wins. Second base has a bigger drop-off, with #14 having 2.5 and #15 having 1.3. At short, the drop-off is quite dramatic, and earlier. #11 had 2.1 wins, #12 and 13 had 1.6, and #14 had 1.1. So clearly we're talking about full-time players when we say an average player is a 2 win player. Furthermore, I really think that starters and relievers should be considered different positions. If a team has a starter get hurt, most of the bullpen arms are not available to replace him in the rotation, so when we talk about a starter's replacement level, we're talking about one or two bullpen guys on most teams and then AAA starters. But again, all of this is pretty much irrelevant as WAR is about replacement level, not about making an average player worth 2 wins. The second note is about fWAR for pitchers and FIP more generally. It's a common misconception that FIP is about removing luck. That's not really its purpose. Its purpose is to focus on what a pitcher is solely responsible for. With a little help or hindrance from their catchers (and perhaps some great home-run robbing catches by outfielders), pitchers alone are responsible for strikeouts, walks, and home runs, the three components of FIP. Everything else is a complex interaction of how hard the ball is hit, where it is hit, how good the fielders are, and where they are positioned. FIP is not meant to state that whether a ball in play is a hit or an out is purely luck, but that we simply don't know how much credit or responsibility to give to the pitchers and how much to give to the fielders. The confusion comes in that, most often, a pitcher's ERA will regress towards his FIP, because when more (or fewer) balls in play than usual are becoming hits the likelihood is that this won't continue. Often this is because balls were finding holes or screaming liners were being hit at fielders, but sometimes it's because a pitcher was missing his spots in a way that he won't continue to do and still be in the majors as a starter (or hitting his spots in a way he won't continue to do), or he's been facing a number of particularly good (or bad) lineups, or something else that won't continue to happen. When Voros McCracken realized that BABIP almost always regressed to a certain point (right now it's around .290) and other sabermetricians read what he had discovered, people often referred to this as luck. More recently, people have recognized that it's not always the ground balls finding holes, etc. and have begun referring to particularly high or low BABIPs as unsustainable, for whatever reason. This is all true, of course, for a pitcher who has shown the skill level to reach the major leagues. If I took the mound, my BABIP would be about .800, since my two-seamer sits at 65 and I don't throw any other pitches.
|
|
|
Post by jchang on Feb 2, 2014 12:39:33 GMT -5
Its complicated. Baseball Reference provides this as a guide (same on position and pitcher): 8+ MVP, 5+ A-S, 2+ Starter, 0-2 Sub, < 0 Repl. which clearly says a Starter is 2 and that 0-2 (lets say it really means 0-1.9) is a sub (I will consider a good closer to be equivalent to a starter). And yes, I do acknowledge that this is different from FG. This is (kind of) the definition used in one of the articles cited at the top of this thread, except that they used 1.5+ as the standard for prospect success instead of 2. Also, most of the WAR data I am using is BR, not FG.
But you are right in that one of the 2 anchor points of WAR is that 0 is a replacement player (and a replacement team wins 48 games). The second anchor point of WAR is that the average team is WAR 33 (48 + 33 = 81), otherwise WAR would not be a win above replacement.
OK, now that we have stated the description, what do the actual WAR values assigned say? The average team is WAR 20 in batting and WAR 13.7 in pitching for an excess of 0.7 (meaning the replacement team should have 47.3 wins).
On the position side with WAR 20, this would support WAR 2 for each of 9 positions (NL no DH really messes up my model) and WAR 2 for the 4 bench roster spots. Note that this is by roster spot. Given that playing time is distributed with platoon and AAA players, expect to see viable players between WAR 1-2.
On the pitching side, there is sufficient information in the Baseball-Reference data set to further assess that the WAR 13.7 pitching is split 9.1 on IP starting and 4.6 on IP in relief. This works out to WAR 1.82 for 1/5th of the starts for each team (32.4), but starting is really tough so we could normalize SP WAR to 32 games and 192 IP (starters average 5.9IP per start) to assess an individual pitcher against the WAR 1.82 per roster spot.
So the actual WAR values assigned match your definition of average, even though the stated definition is otherwise, i.e., its complicated. The WAR values assigned to position players do seem to make sense. It is with starting pitchers that the values are goofy. Because SP and RP seem to be graded on the same scale, a perfectly reasonable 5 starter at 32GS, 180IP, RA9 5.25 get assigned WAR -1.06 (Wily Peralta). By comparison, a relief pitcher with 36 IP RA9 6.94 has WAR -0.92. Also, a call-up with 1 ok start gets WAR ~0, below that of a 5-starter.
The whole point of my line of argument is that using WAR 1.5 as the standard of success for assessing prospect contributed heavily to the apparent low success rate of pitching prospects. If we consider that most of the top 50 prospects are grade 60 and 65, this maps to a 3-starter, and 51-100 are grade 55 mapping to a 4/5 starter. In the actual WAR results, a 3-pitcher might be in the 1.25 to 2.25 range, and the 4/5 starters in the -1 to 1.25 range. So the fact that 20% of the 51-100 achieved WAR 1.5+ actually represents the over-achievement group. There are about 20% in the WAR 0.5 to 1.0 range that were classified as busts should actually be considered to be successful projections.
Edit: League W L bWAR-pos bWAR-pit bWAR-tot AL 81.3 80.8 21.05 14.36 35.4 NL 80.7 81.3 18.9 13.00 31.9 What interesting is that AL has 3.5 WAR higher than NL, but only 0.5 extra W to show for it?
Edit 2: By the prospect grading scheme, a grade of 5 can be assigned to a prospect projecting to be a 4/5 starter or late-inning RP/2nd div closer. In other words, a 5 starter producing 160-180 IP at RA9 4.8 or so is much more than a generally available replacement level player. But in actual WAR values, the 4/5 starter is ranging from -1 to 1 (avg 0) while the group 2 RP get WAR ranging from 1.25 to 2.0, averaging perhaps 1.5. As it
|
|
|