August 2, 200719 yr Preliminary findings on the best OBP / SLG ratio for OPS: In the interest of time and convenience, I relied solely on team runs scored, OBP, and SLG for this season, which makes small sample size alarms go off everywhere. I'll try to get a more in-depth analysis done soon where I look at those team stats for each season since the strike 1994-2007. I think it's probably the best period to use in terms of defining a particular era and having a big enough sample size. That said, I ran a multiple regression on where I use a computer to figure out the optimal values of A, B, and C for the equation: A*OBP + B*SLG + C = Runs Scored. The important thing is the values for A and B because B/A ought to give you the optimal ratio of OBP:SLG. Surprisingly, I got a rate of 1.342 OBP : 1 SLG as the best fit for the existing data. Additionally, my R-Squared value (which tells you how high the correlation is) was 0.89 (Perfect correlation = 1.0). Since I was kind of surprised that the coefficient for OBP wasn't higher, I decided to test the new formula against other variations of the OPS formula with different coefficients for OBP. The results are in the included graph. As you can see, the difference isn't all that significant although 1.342 remains the optimal coefficient. I'm going to see if I get the same thing once I include a wider sample size probably tomorrow. I've always thought the best ratio was 1.25 to 1 but I don't remember why. I must have read something a while back with that ratio and it must have made sense to me.
August 2, 200719 yr CC, I'm not going to sit here and type great post and make it seem like I understand what you just posted. I know what you're trying to figure out (bobbob explained that one to me :mischief ), but the graph lost me. I should have paid more attention in class.
August 2, 200719 yr I've always thought the best ratio would be 4 to 1, but thats only in a perfect world. Obviously, we are never going to have a player who outslugs his OBP by 4, so I guess a ratio closer to 1:1 would be more realistic. I was thinking a simpler way to do this would simply find a selection of players who have an extreme difference between their OBP and SLG (like, around .300 OBP and above .500, think Tony Batista in 2000) and then a group who are more OBP, less SLG (OBP above .370, SLG below .400, think Luis Castillo in 2003) and then a group of players who are average in both (.330/.430, Mark Kotsay in 2005) and simply find which type of player creates more runs per 27 outs. It wouldn't tell you the exact ratio between their values, but it would do well in telling you which type of player is more valuable. It's imperfect, and it could be better, but for those of us who aren't intelligent (Me), it's a rough way to tell you whether which type of player is the most valuable. Though, I'd suggest listening to anything Crimson Cane says over me.
August 2, 200719 yr What was your value for C? Where does it come into play for the equation? Wouldn't it need to be some sort of constant? (math isn't my forte) For the purpose of this analysis, you can disregard C because all we're worried about is correlations and relative values. With the formula A*OBP + B*SLG + C = Runs Scored, losing or keeping C doesn't alter order at all, which makes it acceptable to disregard. For example, suppose we're trying to evaluate something like the effect of height on a person's salary. You got 4 people (5' 7", 5' 10", 6'0, 6"4) whose salaries are $40,000, $50,000, $48,000, and $60,000. Your regression formula would be something like A*Height(In inches) + C = Salary. Now suppose you subtracted C from each salary. Since each value is being subtracted by the same constant, that operation doesn't change the order (the 6' 4" guy is still the richest, the 5'7" is still the poorest). If calculate the correlation of A*Height to Salary or A*Height to (Salary - C), you'll get the same values.
August 2, 200719 yr CC, I'm not going to sit here and type great post and make it seem like I understand what you just posted. I know what you're trying to figure out (bobbob explained that one to me :mischief ), but the graph lost me. I should have paid more attention in class. There are 5 expressions to the right of the graph: OBP+SLG (that's the light blue line) 1.34(OBP) + SLG (that's the red line) 2(OBP) + SLG (Yellow line) 3(OBP) + SLG (Green line) 4(OBP) + SLG (Purple line) The y-axis is OPS (in this case xOBP + SLG = OPS). It goes as high as it does because of the multiplier in front of the OBP. Look at the purple expression. Instead of having an OPS of .700 when OBP = .300 and SLG =.400, you'd have an OPS of 1.600. If you plug in the numbers into the equation you see why: 4(.300)+(.400) = 1.600. The points scattered around the lines represent different points in the data when they are plugged into each expression. The slope of the line represents the accuracy of the modifier in front of OBP. The closer the points are to the line, the more reliable the modifier is in predicting Runs scored. R-Squared represents how reliable the modifier. The most reliable result for R-squared would be 1.00, the least reliable would be 0.00. The r-squared of the 1.34 modifier = .89. That's pretty reliable, but not fantastic. He wants to make his inquiry over a larger sample to see if he can increase his r-squared value.
August 2, 200719 yr What was your value for C? Where does it come into play for the equation? Wouldn't it need to be some sort of constant? (math isn't my forte) For the purpose of this analysis, you can disregard C because all we're worried about is correlations and relative values. With the formula A*OBP + B*SLG * C = Runs Scored, losing or keeping C doesn't alter order at all, which makes it acceptable to disregard. For example, suppose we're trying to evaluate something like the effect of height on a person's salary. You got 4 people (5' 7", 5' 10", 6'0, 6"4) whose salaries are $40,000, $50,000, $48,000, and $60,000. Your regression formula would be something like A*Height(In inches) + C = Salary. Now suppose you subtracted C from each salary. Since each value is being subtracted by the same constant, that operation doesn't change the order (the 6' 4" guy is still the richest, the 5'7" is still the poorest). If calculate the correlation of A*Height to Salary or A*Height to (Salary - C), you'll get the same values. You wrote + C earlier. Multiplying by C makes much more sense. I see what you mean.
August 2, 200719 yr What was your value for C? Where does it come into play for the equation? Wouldn't it need to be some sort of constant? (math isn't my forte) For the purpose of this analysis, you can disregard C because all we're worried about is correlations and relative values. With the formula A*OBP + B*SLG * C = Runs Scored, losing or keeping C doesn't alter order at all, which makes it acceptable to disregard. For example, suppose we're trying to evaluate something like the effect of height on a person's salary. You got 4 people (5' 7", 5' 10", 6'0, 6"4) whose salaries are $40,000, $50,000, $48,000, and $60,000. Your regression formula would be something like A*Height(In inches) + C = Salary. Now suppose you subtracted C from each salary. Since each value is being subtracted by the same constant, that operation doesn't change the order (the 6' 4" guy is still the richest, the 5'7" is still the poorest). If calculate the correlation of A*Height to Salary or A*Height to (Salary - C), you'll get the same values. You wrote + C earlier. Multiplying by C makes much more sense. Sorry for the confusion. I actually meant + C. (Edits earlier post) Hope this doesn't make it just as confusing now. Also, thanks for helping explain the graph.
August 2, 200719 yr I find this very interesting and look forward to what CC comes up with and I hope he goes through with data from '94 - present. That said, not to go elementary in this discussion or to pimp myself but an easier way of looking at the basics of an argument of OBP v. SLG is to break it down in to extremes: Team A has an OPS of 1.000 Team B has an OPS of 1.000 Team A has a 1.000 OBP and a .000 SLG% Team B has a .333 OBP and a .666 SLG% Team A would score infinity runs because they will never record an out. Team B will score many runs because they will have at least two doubles per every three at bats. That is an easy way of spelling out why OBP > SLG, it's incredibly extreme but from reading what I have in this thread I feel that may be a good way to illustrate the point. I don't know how many of you are familiar with Aaron Gleeman or read his site or have seen his work on The Hardball Times but I think he has come up with the best easy formula for evaluating a player's value when it comes to OBP > SLG in an easy to read and understand formula and number. Gleeman and his fellow SABR members have figured that OBP is 80% more valuable than SLG. With that 'solved', Gleeman created Gross Production Average (GPA). GPA being calculated by (1.8)OBP + SLG/4 1.8 being the value of OBP over slugging and dividing the number by 4 to get something similar to batting average. So let's compare two players using GPA. Miguel Cabrera gets on base at a .406 clip and sluggs at a .610 mark. (1.8).406 + .610/4 = .335 Dan Uggla gets on base at a .328 clip and sluggs at a .510 mark. (1.8).328 + .510/4 = .275 It gives you an easy, batting average like number quickly and a number that is pretty telling. It's not as good as EQA but it's good enough and quickly figured out. Again, not to distract from the good stuff CC is doing but I felt this would be a good way to answer the original question posed by Jon.
August 2, 200719 yr Mike's putting that Belen education to good use.....that makes one of us! :| Jokes, jokes
August 2, 200719 yr If he was playing in a different market he would be getting MVP consideration. If Hanley and/or Cabs doesn't finish in the top 3...I'll...well I don't know what I will do, but I will be pissed.
August 2, 200719 yr Mike's putting that Belen education to good use.....that makes one of us! :| Jokes, jokes I try. I try.
August 2, 200719 yr Seriously, I can't think of a time when the left side of the infield for any team was this productive. Maybe A Rod and Jeter were at one time, but these guys are creating so many runs, its nuts. Oh yeah, I think someone already mentioned it, but The Numbers Game by Alan Schwarz is a good statistical analysis book, and the BP handbook is excellent, although horribly difficult to read. I still haven't finished it, as it gives me severe pain in the back of my head.
August 2, 200719 yr I'm pretty much finished with the analysis for 1997-2006, I'll be posting it once I get back from my lunch break. Since I think Hanley's awesomeosity is deserving of it's own thread and I can imagine some questions forthcoming, I'll be posting it in the Major League Baseball section as it's own thread. Back to Hanley...
August 2, 200719 yr I'm pretty much finished with the analysis for 1997-2006, I'll be posting it once I get back from my lunch break. *eats brain food in order to comprehend the forthcoming numbers*
August 2, 200719 yr I'm pretty much finished with the analysis for 1997-2006, I'll be posting it once I get back from my lunch break. *eats brain food in order to comprehend the forthcoming numbers* :Switches brain food for simply brains just to watch the madness that would ensue: Looking forward to this write up
August 2, 200719 yr Author Sadly it looks like Hanley missed out on player of the month to Ryan Braun who was also pretty well deserving. http://mlb.mlb.com/news/article.jsp?ymd=20...sp&c_id=mlb
Join the conversation
You can post now and register later. If you have an account, sign in now to post with your account.
Note: Your post will require moderator approval before it will be visible.