Bill James on WAR

[gunnarthor]

ESPN linked this Posnanski piece:
Friday reading list: Bill James hates WAR - SweetSpot Blog - ESPN

Here is Bill James on Wins Above Replacement, perhaps the hottest advanced statistic in the game right now:

"Well, my math skills are limited and my data-processing skills are essentially nonexistent. The younger guys are way, way beyond me in those areas. I’m fine with that, and I don’t struggle against it, and I hope that I don’t deny them credit for what they can do that I can’t.

"But because that is true, I ASSUMED that these were complex, nuanced, sophisticated systems. I never really looked; I just assumed that the details were out of my depth. But sometime in the last year I was doing some research that relied on these WAR systems, so I took a look at them, and ... they’re not very impressive. They’re not well thought through; they haven’t made a convincing effort to address many of the inherent difficulties that the undertaking presents. They tend to get so far into the data, throw up their arms and make a wild guess. I don’t know if I’m going to get the time to do better of it, or if it will be left to others, but ... we’re not at anything like an end point here. I assumed that these systems were a lot better than they actually are."

A few years ago, Posnanski also wrote that the A's internal WAR system ranked Cabrera ahead of Trout - sorry I don't have the link.

 

[steveringo]

The biggest concern I have is with the people that try to rely on one stat only. They try to create one all-encompassing statistic that lumps everyone together...

WAR is a great concept, and for the most part it tells a great story, but to say one person is better because of WAR alone, is not very wise.

 

[HammerDown]

They try to create one all-encompassing statistic that lumps everyone together...

That's what we all want; one legit stat that's easily understood.

 

[element1286]

Hard to determine what he means without specifics. But if you actually understand WAR, you understand it's inherent flaws and questionable assumptions, and there are many. When you break it down though, assuming you understand it's flaws, there is no all encompassing baseball stat that comes close.

 

[williewilliejuan]

I hate WAR, and I specifically hate it when wannabe mathematicians use it as the sole basis of comparison between two players. This whole sabermetrics fad has great application if used correctly, but you still have to watch and understand the game.

 

[gunnarthor]

Hard to determine what he means without specifics. But if you actually understand WAR, you understand it's inherent flaws and questionable assumptions, and there are many. When you break it down though, assuming you understand it's flaws, there is no all encompassing baseball stat that comes close.

But, honestly, how many people understand it? I'm a smart guy by most definitions but I don't know how to calculate WAR.

Of the two bits, I'd like to know more about the A's internal system that had Miggy ranked ahead of Trout.

 

[element1286]

But, honestly, how many people understand it? I'm a smart guy by most definitions but I don't know how to calculate WAR.

Of the two bits, I'd like to know more about the A's internal system that had Miggy ranked ahead of Trout.

It definitely provides cover for people to argue from ignorance, because it boils a player down to one number. I mean the formula is on the website if you are talking about just straight plug and chug, its fairly simple. Understanding the possible problems with the formula/theory really shows if you 'understand' it in the context of baseball, IMO.

I have a hard time believing the A's system had Cabrera ahead of Trout, especially in 2012 when there was very little difference in their offensive seasons, with Trout having huge advantages in defense and baserunning.

 

[ducky]

The defensive components for the WAR stats are almost completely useless. And WAR heavily relies on those defensive metrics for their final number which makes the end number nearly useless. Anyone who had looked into WAR years ago knew of these flaws and tried to warn people. Hell even the people who put out these stats have disclaimers in there that state the stat is completely useless and nothing more than a slightly educated guess.

The defensive issues are going to slowly be solved in the coming years with better defensive measurements. It won't be long until every inch of a ballpark is measured so it will give us a better idea on a defensive player's actual range. Right now "range" stats are more a component of who is pitching rather than who is playing the field. Eventually that will change.

But even when it does it is going to impossible for a decent WAR stat because there is no good way to determine positional importance which is a HUGE component in the WAR stat. There simply is not a good way to say that an above average defensive left fielder is worth X more or less than a below average defensive 3rd baseman.

To anyone with half a brain the obvious answer is to stop using the WAR stat (let alone relying on it) and only compare the tangible (in other words only compare offensive numbers) and then to factor in defensive attributes separately.

 

[element1286]

The defensive components for the WAR stats are almost completely useless. And WAR heavily relies on those defensive metrics for their final number which makes the end number nearly useless. Anyone who had looked into WAR years ago knew of these flaws and tried to warn people. Hell even the people who put out these stats have disclaimers in there that state the stat is completely useless and nothing more than a slightly educated guess.

The defensive issues are going to slowly be solved in the coming years with better defensive measurements. It won't be long until every inch of a ballpark is measured so it will give us a better idea on a defensive player's actual range. Right now "range" stats are more a component of who is pitching rather than who is playing the field. Eventually that will change.

But even when it does it is going to impossible for a decent WAR stat because there is no good way to determine positional importance which is a HUGE component in the WAR stat. There simply is not a good way to say that an above average defensive left fielder is worth X more or less than a below average defensive 3rd baseman.

To anyone with half a brain the obvious answer is to stop using the WAR stat (let alone relying on it) and only compare the tangible (in other words only compare offensive numbers) and then to factor in defensive attributes separately.

Debatable how useless the defensive component is, I wouldn't use a single season worth of data, but regress multiple years worth of data it's plenty useful.

Well I don't disagree that positional adjustments can be troublesome. Not sure I agree with your conclusion to just throw everything out. We do know it's harder to play 3rd base than LF, so there is some instruction there. Whether 10 runs in the correct adjustment is certainly up for debate.

To the bolded, I agree. Of course, under the assumption that positional adjustments are real.

 

[steveringo]

To anyone with half a brain the obvious answer is to stop using the WAR stat (let alone relying on it) and only compare the tangible (in other words only compare offensive numbers) and then to factor in defensive attributes separately.

I'll disagree with this statement. You have made a educated decision to disregard WAR, but the majority of WAR 'haters' have no idea what it means.... Anyone with half a brain would figure it out and use it for what it is worth... It is a very valuable measurement to compare players as long as it is not misunderstood.

 

[ducky]

... It is a very valuable measurement to compare players as long as it is not misunderstood.

It's isn't a valuable measurement because it uses to many inaccurate stats.

It's flat out easier to look at the important stats rather than looking at WAR and trying to take out what you know to be inaccurate. It's like baking a cake. It's easier to crack an egg and put it into the batter rather than throwing in the entire egg and taking out the shell later.

WAR isn't some magical number that uses super secret things that we already don't have access to. It is just a clumsy mixture of stats that are already available on any guy's stat page.

 

[steveringo]

It's isn't a valuable measurement because it uses to many inaccurate stats.

It's flat out easier to look at the important stats rather than looking at WAR and trying to take out what you know to be inaccurate. It's like baking a cake. It's easier to crack an egg and put it into the batter rather than throwing in the entire egg and taking out the shell later.

WAR isn't some magical number that uses super secret things that we already don't have access to. It is just a clumsy mixture of stats that are already available on any guy's stat page.

What are important stats?

If you had to rank the top ten RF seasons of the last ten years, where would you start?

 

[Darkstone42]

WAR is fine, and the folks who use it know it isn't all there is to look at. At least, the folks who use it right. It's a decent attempt at assessing a player's full, comprehensive value, and in general, the WAR tiers and the relative positions on the WAR lists are in decent agreement with what folks might assume.

Everyone knows Trout is amazing. WAR says Trout is amazing. And that's a fine result. And it's fine to say, "look at Trout's WAR!" because Trout's WAR is probably pretty reflective of Trout's ability and performance.

But, of course, you can also go deeper. Fangraphs, for example, has a ton of information you can sift through. So maybe you want to look at WAR as an initial guess, but as you get more granular, you can separate out the offensive components (like wRC+), look at baserunning and defensive metrics over players' careers to get an idea of their true values in those elements, and maybe also go into things like RE24, WPA/Li, and various other stats to determine context-dependent performance.

Sure, maybe WAR is a guess, but ultimately it tends to be a pretty decent one.

 

[ducky]

What are important stats?

If you had to rank the top ten RF seasons of the last ten years, where would you start?

Well first of all I start with the OPPOSITE premise that WAR starts with.

WAR is a an attempt to objectively rate a player and condense it down to one number. That is flawed from the get go. The simple fact is that there is no one measurement and there never will be. As much as baseball people like to think that stats can tell the whole story, it is simply not true. It is all a fool's errand.

A guy who has one obscenely good week followed by 3 weeks of poor play is not as good as a player who consistently shows up over the course of the month. Stat guys will tell you that it all evens out but it doesn't. Anyone who has followed baseball closely and long enough will tell you that. There are guys who are professionals who show up every day and grind out a season and then there are guys who are great for two weeks and can carry a team offensively for that time but is followed by a month and half long slump. Both types of players are vitally important to a baseball team but the consistently hard outs are more difficult to find and impact a team's success more.

And to go even further, there are times when there isn't even a single correct answer of who is better over a single season than another guy more often than not. Who had the better season isn't objective in baseball....its subjective when the seasons are somewhat comparable. Sometimes a guy who is a monster masher in the middle of the lineup isn't as important as the lead off spark plug for a team and vice versa.

And all that is probably my biggest problem with WAR. It starts with a completely flawed premise that you can boil a player down to a single number and put a single number on how important he was to a team. It's simply not possible to do. I do understand some of the utility behind it when you are a trying to make quick judgements on players but it is flawed and will always be flawed approach. And yet you still get people who use the stat in context of an argument way too often when they KNOW its a flawed and inaccurate stat.

 

[Darkstone42]

Well first of all I start with the OPPOSITE premise that WAR starts with.

WAR is a an attempt to objectively rate a player and condense it down to one number. That is flawed from the get go. The simple fact is that there is no one measurement and there never will be. As much as baseball people like to think that stats can tell the whole story, it is simply not true. It is all a fool's errand.

A guy who has one obscenely good week followed by 3 weeks of poor play is not as good as a player who consistently shows up over the course of the month. Stat guys will tell you that it all evens out but it doesn't. Anyone who has followed baseball closely and long enough will tell you that. There are guys who are professionals who show up every day and grind out a season and then there are guys who are great for two weeks and can carry a team offensively for that time but is followed by a month and half long slump. Both types of players are vitally important to a baseball team but the consistently hard outs are more difficult to find and impact a team's success more.

And to go even further, there are times when there isn't even a single correct answer of who is better over a single season than another guy more often than not. Who had the better season isn't objective in baseball....its subjective when the seasons are somewhat comparable. Sometimes a guy who is a monster masher in the middle of the lineup isn't as important as the lead off spark plug for a team and vice versa.

And all that is probably my biggest problem with WAR. It starts with a completely flawed premise that you can boil a player down to a single number and put a single number on how important he was to a team. It's simply not possible to do. I do understand some of the utility behind it when you are a trying to make quick judgements on players but it is flawed and will always be flawed approach. And yet you still get people who use the stat in context of an argument way too often when they KNOW its a flawed and inaccurate stat.

Everyone knows WAR is only good to the ones place. And if you want other stats, the same guys who developed WAR have also developed things like WPA (win probability added), RE24 (real runs created), and others. No one's trying to boil players down to one number. We're trying to get a good starting point and first guess at relative value.

 

[steveringo]

Everyone knows WAR is only good to the ones place. And if you want other stats, the same guys who developed WAR have also developed things like WPA (win probability added), RE24 (real runs created), and others. No one's trying to boil players down to one number. We're trying to get a good starting point and first guess at relative value.

Yep... for a starting point, WAR is great. If I wanted to find the best single seasons for a position over a period of time, I'd look up the top 20 WAR seasons, then evaluate each of them. Chances are high that the top ten seasons are included in the top 20 WAR years....

 

[element1286]

Well first of all I start with the OPPOSITE premise that WAR starts with.

WAR is a an attempt to objectively rate a player and condense it down to one number. That is flawed from the get go. The simple fact is that there is no one measurement and there never will be. As much as baseball people like to think that stats can tell the whole story, it is simply not true. It is all a fool's errand.

A guy who has one obscenely good week followed by 3 weeks of poor play is not as good as a player who consistently shows up over the course of the month. Stat guys will tell you that it all evens out but it doesn't. Anyone who has followed baseball closely and long enough will tell you that. There are guys who are professionals who show up every day and grind out a season and then there are guys who are great for two weeks and can carry a team offensively for that time but is followed by a month and half long slump. Both types of players are vitally important to a baseball team but the consistently hard outs are more difficult to find and impact a team's success more.

And to go even further, there are times when there isn't even a single correct answer of who is better over a single season than another guy more often than not. Who had the better season isn't objective in baseball....its subjective when the seasons are somewhat comparable. Sometimes a guy who is a monster masher in the middle of the lineup isn't as important as the lead off spark plug for a team and vice versa.

And all that is probably my biggest problem with WAR. It starts with a completely flawed premise that you can boil a player down to a single number and put a single number on how important he was to a team. It's simply not possible to do. I do understand some of the utility behind it when you are a trying to make quick judgements on players but it is flawed and will always be flawed approach. And yet you still get people who use the stat in context of an argument way too often when they KNOW its a flawed and inaccurate stat.

WAR already has a playing time component, IP and PA's. There is really zero evidence that consistency is better than ups and downs.

Whether or not it's possible to the precision you seem to think it should be is immaterial. WAR, and more precisely its components, taken in their own contexts are better than anything we have right now.