FAQ and Forum on Advanced Stats

[shysocks]

I think the trick is that "you're paying $6-$7 million based on what the guy did last year".

 

You then wind up paying $9.5 million because "what they did last year" doesn't take into account the chances of a guy going all LaRoche/Melky/Rios/etc. And that is so common - guys underperform so incredibly much - that failures like them are the norm, not the exception. You sign a couple guys and you're bound to find someone who completely disappoints.

But teams don't pay based on what a guy did last year, they're paying based on their projections and they understand that some decline is expected given the age of most free agents. They don't expect a guy to go all LaRoche/Melky/Rios, but they do expect probably half a WAR less, give or take depending on the guy's baseline. That's why the $6M has been surpassed.

 

I did my own checks based on this link and came up with 38 names, not 31 like you had. Not sure why we differ but I did cut a few names because they were either international signings or they merely had options picked up. I extrapolated their final 2015 WAR just like you did and found there were 17 guys who do better than a 0.5 WAR decline. So yes, if 17 out of 38 will do better than you think, those are pretty bad odds.

 

There are 6 more guys due to decline by 0.5-1.0 WAR. 6 more between 1-2 (the moderate failures?), and 9 who decline by more than 2.0. Those are the catastrophes. The Sox have two of those (LaRoche, Melky), along with one moderate failure (Duke) and one big win (Robertson).

[Balta1701]

But teams don't pay based on what a guy did last year, they're paying based on their projections and they understand that some decline is expected given the age of most free agents. They don't expect a guy to go all LaRoche/Melky/Rios, but they do expect probably half a WAR less, give or take depending on the guy's baseline. That's why the $6M has been surpassed.

 

I did my own checks based on this link and came up with 38 names, not 31 like you had. Not sure why we differ but I did cut a few names because they were either international signings or they merely had options picked up. I extrapolated their final 2015 WAR just like you did and found there were 17 guys who do better than a 0.5 WAR decline. So yes, if 17 out of 38 will do better than you think, those are pretty bad odds.

 

There are 6 more guys due to decline by 0.5-1.0 WAR. 6 more between 1-2 (the moderate failures?), and 9 who decline by more than 2.0. Those are the catastrophes. The Sox have two of those (LaRoche, Melky), along with one moderate failure (Duke) and one big win (Robertson).

Yeah, a quick look shows that mine didn't include Kang or Tomas, so that's part of what's different. Out of your list, >50% underperform and 25% are absolute disasters. So we did get unlucky somewhat in disasters, but that's not unexpected with odds that bad.

[OmarComing25]

Yeah, a quick look shows that mine didn't include Kang or Tomas, so that's part of what's different. Out of your list, >50% underperform and 25% are absolute disasters. So we did get unlucky somewhat in disasters, but that's not unexpected with odds that bad.

What if you did a multi-year analysis on this? It'd be interesting to see data points beyond this year.

[shysocks]

OK, good, looks like I have found somebody who studies these things. Maybe you can explain: why does BABIP fluctuate so much year to year considering a random coin flip or die roll would never give such wild results. For example if the true odds of a weighted coin were 30% heads, over 400 flips (full season balls in play), the odds of getting 36 or 38% heads would be very very small. Yet for a given player, BABIP is supposed to be random.

Moving my response to the advanced stat thread.

 

I would say it's because baseball isn't a controlled probability experiment like flipping a coin. It's more complicated. There are dozens of factors that determine what happens to a ball in play. It involves a human being hitting a round ball hurled by another human being at varying speeds and trajectories with a cylindrical bat; the ball can then travel to any part of a 3-dimensional playing field, where any of nine human beings must react to turn it into an out.

 

To get less fluctuation you could do two things. One is simplify the experiment. I would bet that if you could have the same pitcher in front of the same defense throw straight fastballs down the middle of the zone at the same velocity to a hitter for 400 ABs, his BABIP would fluctuate a lot less.

 

Two, you could increase the sample. For a coin flip 400 is a lot of trials, but with all the factors in baseball, it's not that many. But in the last five years, with 180,000 plate appearances per season, the league BABIP has only ranged from .295 to .299.

[Hatchetman]

So when a guy with 1200 career balls in play has a .300 BABIP and the next year he has 400 balls in play with a .330 BABIP, what's his BABIP going to be next year?

[Eminor3rd]

Moving my response to the advanced stat thread.

 

I would say it's because baseball isn't a controlled probability experiment like flipping a coin. It's more complicated. There are dozens of factors that determine what happens to a ball in play. It involves a human being hitting a round ball hurled by another human being at varying speeds and trajectories with a cylindrical bat; the ball can then travel to any part of a 3-dimensional playing field, where any of nine human beings must react to turn it into an out.

 

To get less fluctuation you could do two things. One is simplify the experiment. I would bet that if you could have the same pitcher in front of the same defense throw straight fastballs down the middle of the zone at the same velocity to a hitter for 400 ABs, his BABIP would fluctuate a lot less.

 

Two, you could increase the sample. For a coin flip 400 is a lot of trials, but with all the factors in baseball, it's not that many. But in the last five years, with 180,000 plate appearances per season, the league BABIP has only ranged from .295 to .299.

 

Right. Put simply, there are two answers:

 

1) Expected value will converge to the mean as sample grows. One season may be far off, but five season won't be.

 

2) BABIP is NOT supposed to be truly random, it's just that random effects explain most of its value. Players that hit lots of line drives or ground balls, for example, can increase their BABIP.

[Eminor3rd]

So when a guy with 1200 career balls in play has a .300 BABIP and the next year he has 400 balls in play with a .330 BABIP, what's his BABIP going to be next year?

 

There's no way to know -- but the most likely outcome will be whatever the largest total sample shows. So say that after that 400PA year of .330 BABIP, his career total is now .312 or something -- the most likely outcome over the next 400PA is .312 (though you'd have to adjust more for aging curves and anomalies in batted ball profile, which is essentially what the projection systems do).

[Hatchetman]

So we don't know to what degree BABIP is random chance and what might attributable to changes in the player's TWIW, do we?

[Eminor3rd]

So we don't know to what degree BABIP is random chance and what might attributable to changes in the player's TWIW, do we?

 

I think we do know the degree -- I don't personally, but researchers have been able to attach factors to the influence of batted ball profile, ballpark, and quality of opposition. The end result is, roughly, that the majority of it is random chance, but speed, GB/FB rate, and LD rate are significant factors as well.

 

If you look up a list of all the MVPs, for example, you'll notice that all of them run high BABIPs in their MVP years. You'll also notice that they all run high line drive rates in their MVP years. High LD rates correlate strongly with high BABIPs. High LD rates are not shown to be stable from year to year at all, though; they basically occur when a player is hot for an extended period of time. And that makes sense -- guys that win the MVP are "having good years" and/or "playing to the peak of their abilities." So it isn't that it's random, but it ACTS as randomness because we can't predict its peaks and valleys and is better predicted by its collective average than it is by its recent values. We don't know when a guy is going to have the best year of his career, but we do know what it looks like when it's happening, and we do know that it isn't very likely to occur twice if it's driven by high BABIP.

[Hatchetman]

OK. I appreciate you getting me up to speed on this. Seems like it has limited predictive value then, which is what my initial hunch was.

[Eminor3rd]

OK. I appreciate you getting me up to speed on this. Seems like it has limited predictive value then, which is what my initial hunch was.

 

No problem!

[caulfield12]

http://www.kansascity.com/sports/mlb/kansa...le41870337.html

 

On the PECOTA system and the real player himself...

Edited October 30, 2015 by caulfield12

[shipps]

I had a random thought about Jose Valentin and how Hawk would always talk him up even when he had what seemed like statistically down years (advanced metrics weren't really out there). So I looked at what his WAR numbers were and it really did statistically back up the notion that he was a damn good ballplayer even when he had a .237 BA in 2003 where he posted a 4.0 WAR. In 2000 he was a 4.9 WAR player. For four years 2000-2003 he didnt post a WAR lower than 3.0.

 

 

 

 

[shipps]

I had a random thought about Jose Valentin and how Hawk would always talk him up even when he had what seemed like statistically down years (advanced metrics weren't really out there). So I looked at what his WAR numbers were and it really did statistically back up the notion that he was a damn good ballplayer even when he had a .237 BA in 2003 where he posted a 4.0 WAR. In 2000 he was a 4.9 WAR player. For four years 2000-2003 he didnt post a WAR lower than 3.0.

 

In essence I believe Frazier is going to at least produce comparable numbers to what Jose put up. That is pretty awesome to look forward to.

[CaliSoxFanViaSWside]

I still question having a thread pinned that no one has posted in for 4 months especially when game threads sometimes have good discussions going on in them and are booted to the archives almost immediately.

 

Not only that but since now I've posted , it will be the very 1st subject listed on page 1 . Shouldn't this be in the diamond club ?

Edited April 19, 2016 by CaliSoxFanViaSWside

[southsider2k5]

Was watching the game the other day, and had a thought/question. Specifically there was a high fly ball into the gap where Eaton called off Jackson to make a play, and it got me thinking. With Eaton's crazy D ratings, could those actually be hurting Jackson, as is Eaton getting to balls that Jackson would usually get to with a typical RF, thus cutting into Jackson's rating and accounting for why his numbers have been slightly negative this year?

 

I know by my eye test, Jackson has played a pretty decent CF, it hasn't been nearly as great as Eaton, and he does have a couple of misplays, but I have not seen it as an overall negative.

 

Thoughts, especially on the idea that Eaton's huge rating could actually be taking away from Jackson?