05-25-2010, 05:36 PM
<!--quoteo(post=98281:date=May 25 2010, 04:34 PM:name=Butcher)-->QUOTE (Butcher @ May 25 2010, 04:34 PM) <{POST_SNAPBACK}><!--quotec--><!--quoteo(post=98279:date=May 25 2010, 04:29 PM:name=jstraw)--><div class='quotetop'>QUOTE (jstraw @ May 25 2010, 04:29 PM) <{POST_SNAPBACK}><!--quotec--><!--quoteo(post=98273:date=May 25 2010, 04:18 PM:name=Butcher)--><div class='quotetop'>QUOTE (Butcher @ May 25 2010, 04:18 PM) <{POST_SNAPBACK}><!--quotec--><!--quoteo(post=98270:date=May 25 2010, 04:02 PM:name=jstraw)--><div class='quotetop'>QUOTE (jstraw @ May 25 2010, 04:02 PM) <{POST_SNAPBACK}><!--quotec--><!--quoteo(post=98263:date=May 25 2010, 02:30 PM:name=leonardsipes)--><div class='quotetop'>QUOTE (leonardsipes @ May 25 2010, 02:30 PM) <{POST_SNAPBACK}><!--quotec--><!--quoteo(post=98250:date=May 25 2010, 12:58 PM:name=jstraw)--><div class='quotetop'>QUOTE (jstraw @ May 25 2010, 12:58 PM) <{POST_SNAPBACK}><!--quotec-->I have never before heard this idea that stat heads don't think hot and cold exist. Anecdotally, Posnanski mentions statisticians crunching numbers and not finding evidence of, well...essentially, streaks...hitting or missing a number of times in a row that represents a statistical out lier, in terms of distribution. I don't believe it. I think this casual mention is bullshit and is either made up or a complete misinterpretation or misrepresentation of some statistical analysis.<!--QuoteEnd--><!--QuoteEEnd-->
Read the quote. That is what they believe. I remember when the study fist came out in the 80's. At that time, I was in to stat as much as the next guy. It was what first made me realize that stat analysis is faulty.
The fatal flaw in all the analysis, is the results are based on luck not skill. Once they assume an AB is a random event, then you can assume for the purpose of analysis, all ABs are the same. Why is OPS better than RBI - because we can see that some hitters get more RBI chances than others, and we make the assumptions that quality of pitchers faced and type of ABs even out over a season
ABs are not all the same. In a clutch AB at the end of a game, a hitter will face a different pitcher than in a blow out and be pitched differently. Because baseball is made of 9 three out innings, there are a huge amount of different situations that come up, which lead to the pitcher pitching differently and/or a different pitcher in the game.
If instead of games, each team was given 4374 outs to see how many runs they scored, only 3 things would matter OBP, double plays, and runners left on base at the end of the year. Slg and speed would only matter to the extent that they helped avoid DPs. By taking the entire seasons worth of ABs and treating them the same, you are essentially smoothing out the baseball effect (impact of 9, 3 out inning games on the results). By considering runs scored in blow outs (which skews the data, because by definition, there are a lot of them) the same as game winning runs you minimize the effect of little things (what makes baseball, baseball) on wins.
When we watch a game, we see the actual situation. Statistics don't. That is why watching the game, we see things such as clutch, speed, productive outs.
<!--QuoteEnd--></div><!--QuoteEEnd-->
Absolute agreement.
I was thinking earlier that theoretically you could assign a value to an at bat...are there runners on, how many outs, who's hitting behind the batter, how late in the game is it, what's the score, etc. Then you could develop a rating based upon performance through this metric and quantify "clutchness."
<!--QuoteEnd--></div><!--QuoteEEnd-->
Isn't this already done to some degree? You can look at the splits for any batter and they have "close and late," "runner on third with less than two outs," and many other situational ABs.
And I think the conclusion is pretty much that good hitters have good stats in those situations, average hitters have average stats in those situations, and poor hitters have poor stats in those situations.
<!--QuoteEnd--></div><!--QuoteEEnd-->
I agree. I'm just musing about a unified stat that can rate the value of any at bat situation and weigh the outcome based on that value. In this way, one single or walk is not the same as just any other single or walk. An average could be calculated.
<!--QuoteEnd--></div><!--QuoteEEnd-->
Do it. You could revolutionize the game.
<!--QuoteEnd--></div><!--QuoteEEnd-->
Just musing. I'm not a statistician...and the formula for the weighting would be very controversial.
Read the quote. That is what they believe. I remember when the study fist came out in the 80's. At that time, I was in to stat as much as the next guy. It was what first made me realize that stat analysis is faulty.
The fatal flaw in all the analysis, is the results are based on luck not skill. Once they assume an AB is a random event, then you can assume for the purpose of analysis, all ABs are the same. Why is OPS better than RBI - because we can see that some hitters get more RBI chances than others, and we make the assumptions that quality of pitchers faced and type of ABs even out over a season
ABs are not all the same. In a clutch AB at the end of a game, a hitter will face a different pitcher than in a blow out and be pitched differently. Because baseball is made of 9 three out innings, there are a huge amount of different situations that come up, which lead to the pitcher pitching differently and/or a different pitcher in the game.
If instead of games, each team was given 4374 outs to see how many runs they scored, only 3 things would matter OBP, double plays, and runners left on base at the end of the year. Slg and speed would only matter to the extent that they helped avoid DPs. By taking the entire seasons worth of ABs and treating them the same, you are essentially smoothing out the baseball effect (impact of 9, 3 out inning games on the results). By considering runs scored in blow outs (which skews the data, because by definition, there are a lot of them) the same as game winning runs you minimize the effect of little things (what makes baseball, baseball) on wins.
When we watch a game, we see the actual situation. Statistics don't. That is why watching the game, we see things such as clutch, speed, productive outs.
<!--QuoteEnd--></div><!--QuoteEEnd-->
Absolute agreement.
I was thinking earlier that theoretically you could assign a value to an at bat...are there runners on, how many outs, who's hitting behind the batter, how late in the game is it, what's the score, etc. Then you could develop a rating based upon performance through this metric and quantify "clutchness."
<!--QuoteEnd--></div><!--QuoteEEnd-->
Isn't this already done to some degree? You can look at the splits for any batter and they have "close and late," "runner on third with less than two outs," and many other situational ABs.
And I think the conclusion is pretty much that good hitters have good stats in those situations, average hitters have average stats in those situations, and poor hitters have poor stats in those situations.
<!--QuoteEnd--></div><!--QuoteEEnd-->
I agree. I'm just musing about a unified stat that can rate the value of any at bat situation and weigh the outcome based on that value. In this way, one single or walk is not the same as just any other single or walk. An average could be calculated.
<!--QuoteEnd--></div><!--QuoteEEnd-->
Do it. You could revolutionize the game.
<!--QuoteEnd--></div><!--QuoteEEnd-->
Just musing. I'm not a statistician...and the formula for the weighting would be very controversial.