How do you feel about context-dependent advanced stats?

[Omar 382]

Not RBI's or wins, but RE24, WPA/LI, and Clutch? Sabermetricians are largely maligned for citing stats that have nothing to do with context, because they believe that a hitter cannot control how many men are on base or how many runs a pitcher's team scored. Those statistics, and a few others, give weight to how well a batter performs in a higher-leverage situation.

I don't hate them, but I'm not a fan. I agree, of course, a grand slam with two outs in the ninth and a team down by three runs is more important than a home run with a team up 10 runs. But there really isn't any evidence of those statistics being repeatable, which tells me that it is not (for the most part) a skill that can be learned. Let's just look at Barry Bonds in his full seasons in the context of WPA. Bonds played significant parts of every season every year from 1986 to 2007 except for 2005. He never had a WAR below 3.2, and for 16 of the 20 seasons, it was above 5. But looking at WPA, he had 8 seasons, not counting '05, below 5 WPA. If it was such a repeatable statistic, I think there would be more consistency.

Of course, this is just one player, and I cannot find a correlation chart between WPA and WAR. If anyone knows of one, please post it. But looking at lots of other great players, I find a similar trend. Tony Gwynn never had an OBP beneath .350 in a full season. Roger Clemens never had a FIP above 4.20. I would like to see some repeatability before I gave context-dependent stats any weight.

 

[Fountain City Blues]

I generally pay attention when it comes to relievers in this regard as they are one of the few players that aren't context independent. A manager can choose to use a closer- or use a low leverage loogy. If you happen to have a string of tier 1 relievers, things get weird. My guess is that is how some of the past few teams pushing 2+ standard deviations from their projected win total are doing some of their work (besides other shenanigans like run sequencing) even if the pythags and run differentials don't quite add up.


Wade Davis for instance had a very, very healthy bWAR.

 

[rmilia1]

I don't think they're dependable on a career long basis but I think they should be more widely used when debating things like MVP or Cy Young on a season by season basis. If 2 guys have similar stats but one performs worse in high leverage situations or even in games vs quality opponents then I think that's important. I agree that they are largely not repeatable though. On fact most guys seen as "clutch" really aren't better than usual in those situations, they simply aren't worse. Clutch generally just means you don't choke.

 

[poprocksncoke]

 

[Villain]

I don't typically bother with them, if only because I haven't taken the time to really and truly understand what goes into them. I like to stick with what I know. That's regarding stats like RE24 and WPA.

Things like Clutch, ehhhhh. I feel like these situational-dependent stats end up having sample sizes that are too small to be relied upon, then they get used in a narrative-driven argument. Clutch seems like something that exists to be used for confirmation bias, while dismissed when its results aren't with the narrative.

 

[DirtDirtDirt]

Omar's intelligence intimidates me

 

[Omar 382]

Omar's intelligence intimidates me

 

[Omar 382]

I don't think they're dependable on a career long basis but I think they should be more widely used when debating things like MVP or Cy Young on a season by season basis. If 2 guys have similar stats but one performs worse in high leverage situations or even in games vs quality opponents then I think that's important. I agree that they are largely not repeatable though. On fact most guys seen as "clutch" really aren't better than usual in those situations, they simply aren't worse. Clutch generally just means you don't choke.

ummmmm... what?

 

[rmilia1]

ummmmm... what?

Pretty simple. Most guys who are deemed "clutch" are just guys who's postseason or late game stats mirror their regular stats. Most guys perform worse in high leverage situations,than they do regularly. "Clutch" guys largely just perform the same.

 

[The Q]

Not RBI's or wins, but RE24, WPA/LI, and Clutch? Sabermetricians are largely maligned for citing stats that have nothing to do with context, because they believe that a hitter cannot control how many men are on base or how many runs a pitcher's team scored. Those statistics, and a few others, give weight to how well a batter performs in a higher-leverage situation.

I don't hate them, but I'm not a fan. I agree, of course, a grand slam with two outs in the ninth and a team down by three runs is more important than a home run with a team up 10 runs. But there really isn't any evidence of those statistics being repeatable, which tells me that it is not (for the most part) a skill that can be learned. Let's just look at Barry Bonds in his full seasons in the context of WPA. Bonds played significant parts of every season every year from 1986 to 2007 except for 2005. He never had a WAR below 3.2, and for 16 of the 20 seasons, it was above 5. But looking at WPA, he had 8 seasons, not counting '05, below 5 WPA. If it was such a repeatable statistic, I think there would be more consistency.

Of course, this is just one player, and I cannot find a correlation chart between WPA and WAR. If anyone knows of one, please post it. But looking at lots of other great players, I find a similar trend. Tony Gwynn never had an OBP beneath .350 in a full season. Roger Clemens never had a FIP above 4.20. I would like to see some repeatability before I gave context-dependent stats any weight.

It doesn't mean it's not valuable. It shows the value for a single given season of each individual play.