| Grand Slam Baseball View the Forum Registry |
| |
| 0 |
Subject: Pitching Formulas Posted by: cpa68 - [284131720] Fri, Sep 01, 16:20 This was originally posted on K2's message Board, by the owner of the K2 Mets. Allright-enough already about the meaningless of pitcher’s skill’s. As we all have observed there’s obviously something rotten- but no meaning whatsoever? I find that hard to believe-and since I missed the playoffs and had nothing better to do (actually plenty-but first things first..), I thought I’d dig into this stuff. Now for those who only barely passed math class this is how it goes: (If anyone is into this stuff-please bear with the simplifications-I just tried to make this somewhat accessible to the uninitiated) If we assume that one the outcome of one dependent parameter (in this case starting pitching performance) is depending on a number of more or less known set parameters (i.e. FB, BR, CH, CTRL) we have ways of determining statistically significant associations. This is known as multiple regression analysis. I’m not going to go into details but essentially we can determine, within certain limits of confidence, the degree to which these set parameters affect the pitching performance. So I basically made a spreadsheet of 63 K2 SP’s with skills and performances for last season. I then performed a number of analyses to determine the degree to which the skills predicted the outcome of various pitching stats. The analysis assumes a linear association between skills and performance so the results are depicting the statistically best straight line that fits the observed performances. Bear in mind that for the individual pitcher you cannot directly translate these formulae to their whip/era etc.: these are for the overall data and ONLY gives you an idea of the overall trend. Most important statistic is the r squared value. This gives you the proportion (based on a total of 1) of the parameter examined that are explained by the formulae. Thus r squared=1:every part of the performance is explained and r squared=0:no part of the performance is explained. The remaining proportion of performance cannot be statistically accounted for by the skills examined and are due to stochastic variation (i.e. computer randomized numbers) or gameplan factors unrecognized by skills and me. So here we go: WHIP= 2.69-0.00089*FB-0.0009*BR-0.0007CH-0.0008CTRL R squared: 0.348 ERA=10.42-0.0048*FB-0.0028*BR-0.0047*CH-0.0035*CTRL R squared: 0.275 BB/IP=0.57-0.00012*FB-0.00029*BR-0.00019*CH-0.00021*CTRL R squared: 0.197 K/IP=2.12-0.00078*FB-0.00061*BR-0.00052*CH-0.00061CTRL R squared: 0.277 CONCLUSION Well the operation succeeded but the patient died, and I’ll have to eat it:pitching skills sucks bigtime! All of these results are statistically significant-meaning that with a confidence of 95% we can tell that there is some association, but the magnitude of this is less than impressive.The best estimation predicts the whip where the skill factors determine about 35% (and the relative contributions of the individual skill factors are about equal)of the outcome which leaves some 65% to chance. Also rather sucky: the better skills tend to lower the K/IP. Whew-I’d better stop here since my wife thinks I’m busy writing up my Ph.D. Hopes this made sense to somebody.. This is some interesting stuff on how skills affect stats. | ||||||
| 1 | Renegade Salmon ID: 25431221 Fri, Sep 01, 16:48 | LOL, so they have "no meaning whatsoever" after all! | |||||
| 2 | Renegade Salmon ID: 25431221 Fri, Sep 01, 16:53 |
Pitching is the biggest consternation to me in the game, I have no idea what any of the numbers do. It pains me to see other team's pitchers doing better than mine with minor league fastballs or little league breaking balls. I always have good to terrific pitching, but don't ask me how because it seems to be intuition and sense of smell...;) | |||||
| 3 | Chaik ID: 17854120 Fri, Sep 01, 20:54 | This may refute it, but I always averaged the totals of fastball, change, curve and control... My pitching has always been above average too. | |||||
| 4 | Josie ID: 4441223 Fri, Sep 01, 21:13 | I think that things happen based more on hitting than pitching. We have figured the probability of a batter getting a hit, if that hit might be a double, or maybe a home run, regardless of pitching ratings. I am sure there is a formula but enjoy it being unknown. It adds fun to the game. I just wish relief pitching was more reliable. I guess this makes it more realistic tho. | |||||
| 5 | Toral ID: 544331814 Fri, Sep 01, 21:30 |
I have just enough stat knowledge to sort of understand the above. I did my own smaller multiple regression test on ERA only -- smaller sample -- and found nothing statistically significant, but I did it very crudely, as opposed to the above. I add some baseball knowledge, to say the following: 1) BB/IP and (especially) K/IP is essentially of academic interest only, as the only factor we should really be interested in is overall pitching performance. 2) I believe ERA is a better *baseball* measure than WHIP, and will become more so when HR ratings get improved. 3) So, if I understand the ERA equation correctly (which I may not), we learn that fastball and change are more important than breaking. This is counter to the message-board wisdom, which is that change is the least important #. 4) I find intriguing the idea that the random generator is *not* linear -- i.e., that you might find that a change is more (or less) important if you have a good fastball, or that control is more important if you have good breaking pitch (or maybe less), or whatever. 5) When the new pitching sim comes out, studies like this will be very important. BTW, I have had very good luck with my pitching. When a guy pitches good, I leave him where he is, and when he doesn't I change his role, or leave him out of the pitching scheme for the year (if he's maxed out and under 30). Shouldn't make a dime's worth of difference, but has worked for me. | |||||
| 6 | cpa68 ID: 284131720 Fri, Sep 01, 21:44 |
From my understanding these formulas suggest a few things. First, that WHIP is determined by 34.8% skill, and 65.2% luck. ERA is 27.5% skill, and 72.5% luck, thus explaning why Thompson takes a beating while Mr.AAA has a 2.00 ERA. Second, Fastball is the most important pitch, followed by Change-Up, then Control and Breaking (which has signifcantly less impact on ERA and WHIP then the other pitches) Third, Control is not the most important skill when it comes to walking batters, Breaking is. Hopefully the "New Sim" will fix all of these things. | |||||
| 7 | Red Rum ID: 1665108 Sat, Sep 02, 08:43 |
Let me start by saying that I have not been able to understand pitching, but over 4 seasons and a few teams, I have seem that generally better pitchers on average do perform better in the long run, so the ratings must be taken into account somehow. The low number of games they play makes it hard to do any sort of meaningful analysis, however my theory is the sim works as follows for pitching to determine a hit. 1) A random # between -50 to +50 is created for the pitcher for the game (this would explain the deviation between games). For each pitch, the sim picks a pitch to throw (a 1 in 4 chance). It takes the rating for that pitch, adds the random #, multiplies that # by 2.5. It then generates a random # between 1 and the number created and compares it to the batters avg rating less 150. There could also be a random # generated for each innings say -25 to + 25 that would explain the deviation in pitching performance between innings. Example Batter avg = 500. Pitcher's pitch rating selected by sim= 500. Pitchers random number for game = 100. Pitchers random number for innings = -25. Then max pitch rating for pitch = (500+100-25)*2.5 = 1387. Random # generated for pitch therefore has a max of 1387. Assume actual # generated = 250. Batters rating = 500-150=350. Therefore, as 350>250, the ball is a hit. This would explain why hitters are more consistent (less variables, and no random #), and why the pitchers performance varies from games to game, and from innings to innings. The random numbers generated for each game could be larger to explain why bad pitchers can still pitch well. This is just a theory without any sort of detailed analysis, but what do the rest of you think ? | |||||
| 8 | Chris ID: 44443318 Sun, Sep 03, 04:30 |
Toral - I agree ERA is a better baseball measure of a pitcher's success. However, the formula is not applied using projected ERA. It is applied to the sim in a per at-bat basis. So therefore hits/batter, walks/batter, and strikeouts/batter become the only stats that might coincide with ratings, especially with homers out of whack the way they are. BB/IP is surely of more than academic interest if it coincides directly with a particular rating. Walking batters obviously contributes to more runs, thus affecting his overall performance(ERA). There really isn't an "ERA equation". There can't be. It is impossible to program that, to my knowledge. You're best bet is looking at H/IP and BB/IP(K/IP may play a factor as well, although a much smaller one). If there IS an ERA equation, I'm crying foul. As this would suggest that a pitcher will pitch better if he gives up more runs than he should, etc. It would force past performance to affect future performance which should not be the case. Other than that, I'm not a statistician, and I may be acting like a dumbass. This whole discussion may be moot, as we're not sure how the new sim and the addition of realistic homers will affect pitchers. I have a feeling that there will be no given pitchers rating that affects how many homers a guy gives up(ie, the Jose Lima rating). Walks seem to be completely random as it stands(certainly it is for hitters), though I havn't done any extensive research. Thanks you for bringing this here cpa...a very thought-provoking piece... | |||||
| |||||||
Post a reply to this message: (But first, how about checking out this sponsor?) | |||||||