California 2038.11 - An arrogant Jerk's Guide to Sustained Winning, Part 4

[Ted]

Time for sabermetric nonsense.

An Arrogant Jerk's Guide To Sustained Winning Part 4: A Detour Into WAR and Team Building

In Part 1 we talked about make a big ol' pile of baseball value. Part 2 went into what that value is made of. Part 3 superficially discussed how to get it/grow it.


Part four of this series was supposed to be about value leaks and how to avoid them and plug them. However, I decided to take a detour into roster building, because I think it's necessary. This entire effort is designed to be top down. Too often people ask for advice about tactics when they don't have a strategy. The merits of individual tactics are difficult to evaluate without an understanding of strategy. If that sounds opaque, we can go back to the question from part one of "Should I sign this expensive FA that would upgrade my biggest weakness from last season?" If you remember, I answered, "I don't know." Because I don't know your strategy. Not only do I not know it, I don't know if it's any good. Some strategies are better than others. That's what this series is about. It's about the strategy behind building a perpetually winning 88-92 win team. Its about how to make the playoffs 80 percent of the time, and never rebuild.

The following is from Wiki:

"The terms tactic and strategy are often confused: tactics are the actual means used to gain an objective, while strategy is the overall campaign plan, which may involve complex operational patterns, activity, and decision-making that govern tactical execution."

Making trades, signing players, etc are tactics. You cannot properly evaluate them if you do not know the strategy. What could be a good move in one system would be terrible in another. It's largely why Ron and I bicker after these. I say you shouldn't do something. He says that something can be a fine tactic, and he's right. But it's not a good tactic if it doesn't fit into the overall strategy.

Why am I bringing this up? Well, parts 1-3 have involved largely strategy. The planned part 4 about how to not lose value starts to blend in tactics somewhat. But we need a bridge. So this part and at least the next will create that bridge.

Players Have Different Values in Different Situations
I am NOT talking about how a star is more valuable to a winning team than a losing one. Or how a team with a hole in the bullpen might value a certain reliever more than a team that does not need one. Get that out of your head. This is not about that. In fact, that line of thinking is very dangerous. It lets you talk yourself into what are likely overpays. Be very careful when you start thinking about immediate need. You always need to keep in my a player's overall league value, even when looking at how he's valued for you.

So what am I talking about? Let's look at the WAR model. (For those of you who aren't as sabermetrically inclined, don't worry. WAR is the example here. But you can use any way that you evaluate a player). A "replacement level team" should win around 48 games. So add in 40 WAR worth of players, and you have an 88 win team. A wild card contender, right?

Well, not really. Not in the Brewster. I guarantee you I can construct rosters that underachieve whatever metric you use to evaluate them consistently.

To understand this, let's use WAR again. You have to understand what WAR is an how it works.

WAR is an approximation of the number of wins a certain set of outcomes (hits, wlks, HR, etc) will create in a league neutral environment.

LEAGUE NEUTRAL ENVIRONMENT!!!!!

That's the key. How many of us play in a league neutral environment? Not very many. The Brewster, with it's wacky parks and drastically different types of hitters and pitchers, makes simple use of WAR very difficult. I still talk about it a lot, but it's as a shorthand, because it's easier to say "3 win player" than it is to say "right handed, moderate strikeouts, low walk, moderate home run starter who will likely throw 180 innings that should be worth around 3 wins in X park."

Note that you only have to change "X park" to a park that allows lefty hitters to hit home runs easily to completely change the assessment of the players. His WAR should calculate similarly in both cases, but he's much more useful in the first park than the second one.

Hitters are the same. WAR is calculated using what are called linear weights. I'm not going to get into it here, because the fangraphs library can explain it better than I can. If you care, read it there. It will take some time. It's basically a small book. But the concept is simple. A single on average generates X runs, a double Y, etc etc.

But here's the flaw traditionalist point out. All singles do not happen in the same setting. All doubles do not happen in the same setting. And so on. We'll they're both right and wrong. Over time, the settings differences kind of cancel each other out, and WAR becomes reliable. That's what sample size is about. And a necessary sample size for WAR is more than one season.

But this also points out how you can take advantage of certain types of players. Put together a team where your players strengths complement each other. You're not making a WAR soup. You're making an OBP and power soup. Or a pitching and defense soup. Some soups are not as good as others. A high power, low OBP, high speed soup is not very good. Those guys in that last group could be just as talented overall as the high OBP, high power group, but their skills don't complement each other.

This isn't just about park effects. It's about the impact of the OTHER PLAYERS on a team. The perfect example is Jon Mick. Mick, by WAR alone, is not a hall of famer. Not even close. But by traditional counting stats alone, he'd be a shoe in. Two things explain that. One is that traditional counting stats simply don't account for walks, which is a huge oversight. For a big time slugger, Mick walks poorly. That's old school versus new school stuff.

The more important thing is Mick has played his whole career surrounded by tremendous OBP guys. That means he gets more at bats with guys on base, and more at bats than other players in weaker lineups. Madison's lineup is PERFECT for Jon Mick. Does that make him a hall of famer? That's another debate and not relevant to this topic. The point is, Madison's lineup leverages Mick's power potential better than some others would have. He's worth more than his WAR to them. If you think Jon mick was worth only 1.6 wins through his 130 games in 2038 (His WAR at the time of this posting) you are dead wrong and don't get how this works.

Here's a comparison to another player to drive the point home. Rupert Grant is a bit older, but before the last few years before he declined, he had nearly identical avg/obp/slg numbers to Mick. Both players were 8/10/6 right handed bats at their peak. Both batted 3 or 4 their entire careers. Grant has had a very good career. I'm not trying to say anything bad about him. He was part of some very good Calgary teams. But LOOK at the RBI difference. That's not sequencing. That's not park effect. (Calgary is actually RH hitter homer friendly and Madison isn't.) That's TEAM DESIGN. Jon Mick/Rupert Grant, is simply worth more to those Madison teams than either would have been in Calgary.

Again, I'm not saying Calgary did something wrong here. They were actually better over that span that Madison I believe. The point I'm making is that you must understand team makeup to evaluate the utility of a player to the team. These guys are not plug and play.

So that's it. WAR is an approximation of the true talent of a player in a neutral environment. It's very very good at that. We need to be good at realizing what the strength that contribute to that WAR are and how they fit into the STRATEGY behind our team building.

Or don't use WAR. It doesn't matter. Use any stat. Or use the player's raw ratings. You still have to understand what the strengths of a player are and how their crooked edges fit into your puzzle. So with that rather boring material out of the way, we can get to the strategy of roster building next time. After that, then discussion of how to maintain your value will make much more sense.

[RonCo]

Another outstanding piece. But, of course, ideas behind Team Building and Roster Construction is always interesting to think about.

What could be a good move in one system would be terrible in another. It's largely why Ron and I bicker after these. I say you shouldn't do something. He says that something can be a fine tactic, and he's right. But it's not a good tactic if it doesn't fit into the overall strategy.

I've spent literally hundreds of hours discussing the difference between a strategy and a tactic with equally hundreds of corporate managers. Yikes!

We always agree teams should have strategies, though.

In the case of these guides I think we often bicker here because I think you're trying to simplify the playbook (tactics) one should employ to create a team that wins all the time (strategy) into basic rules of thumb, while I'm suggesting GMs have a bigger view of that playbook. Things like: "to enact a strategy of hording talent to win forever (*), never use the tactic of overpaying a relief pitcher." This isn't a bad rule of thumb for limiting your tactics as long as your only real goal is to avoid making a mistake. But there are also ways to leverage over-priced relief pitchers while enacting a strategy that increases your horde of talent and enhances your ability to win forever.

(*) Or is hording talent a tactic and win forever the strategy? See, it goes deep.

WAR is an approximation of the number of wins a certain set of outcomes (hits, wlks, HR, etc) will create in a league neutral environment.

LEAGUE NEUTRAL ENVIRONMENT!!!!!

That's the key. How many of us play in a league neutral environment? Not very many.

I don't think I would characterize WAR that way. Or maybe I'm not sure what you're trying to say.

WAR attempts to remove the effects of parks and teams so that you can compare apples to apples. So, yes, in context of what WAR is _supposed_ to be, all players in the BBA do play in a neutral environment. We can question how well it does this, but that's a different problem. If, however, WAR did do it's job perfectly, and it was a perfect predictor, which no one says it is, then yes, you should actually be able to merely pour WAR into the bucket and make WAR soup.

Of course, no one, not even the people who developed it, say you hsould view it that way.

Another issue in using WAR here is that it's quite unlikely that OOTP calculates WAR in a way that perfectly matches Fangraphs. So even if real WAR were perfect, OOTP WAR would be a little flawed.

WAR is calculated using what are called linear weights. I'm not going to get into it here, because the fangraphs library can explain it better than I can. If you care, read it there. It will take some time. It's basically a small book. But the concept is simple. A single on average generates X runs, a double Y, etc etc.

But here's the flaw traditionalist point out. All singles do not happen in the same setting. All doubles do not happen in the same setting. And so on. We'll they're both right and wrong. Over time, the settings differences kind of cancel each other out, and WAR becomes reliable. That's what sample size is about. And a necessary sample size for WAR is more than one season.

I like this on the whole...but...hmmm.

Your comment about the sample size required to WAR to become reliable is interesting, but it's more useful, I think, to see war as the components it's based on. WAR is composed of offense (essentially wOBA, which is a linear weight kind of thing), base running (which I'm sure OOTP mangles), fielding (ditto), and then both ballpark and league adjustments (ditto/maybe ditto). Of these, defense is probably the most variable.

WAR is a nice stat to toss around, but I think it's wisest to look at these composites wOBA, base running, and fielding (as augmented by playing time).

I note, too, that it's important to understand that WAR is greatly influenced by how you use players--epecially those with splits. A guy with big splits might post .2 WAR if you play him full time, but 1.8 if you platoon him. This is because his performance against the platoon advantage can be so far under replacement that it wipes out the value he gave when he had that platoon advantage.

These things are important tools to keep in mind when using WAR (or whatever, really) to construct your roster.

This isn't just about park effects. It's about the impact of the OTHER PLAYERS on a team.

Your example of Mick and Grant are interesting, though. Grant's career wOBA is 20 points less than Mick's (which is a main reason why Mick's career WAR makes him a little more valuable than Grant...the offensive component of WAR is driven by wOBA). So if you were to plug Grant into Madison, theoretically he'd have a a bit fewer RBI. But in using WAR as a way to construct your club, you'd certainly pick Mick over Grant...simply because his WAR is better. (Of course, Mick could regress and his career WAR could possibly wind up less than Grant's).

I'm thinking that you're getting at the idea that Mick will help Madison score more runs if Mons Raider is also on the team (rather than, say my own Mauro Saucedo). This is true, but that's because of Raider's value, not Mick's. It's fair to say that Madison would score more runs if they had Emilio Morales in place of Mick, but that's because Morales' wOBA (which is park adjusted and which I'll use since he'd have to play 1B to make the comparison valid) is much better than Mick's this year.

Anyway...

We need to be good at realizing what the strength(s) that contribute to that WAR are and how they fit into the STRATEGY behind our team building.

Yes. There's the money quote.

The key to all value judgements is to understand what their roots are.

[Ted]

The point of this was to not really debate the utility of WAR, Ron, but to show how it, or any metric, or the ratings, are great in a vacuum, but have to evaluated in the context of how the player is used.

Regarding the Mick versus Grant wOBA difference, it essentially doesn't exist if you take away Grant's last three seasons, which are all his serious decline phase. Through age 32, both guys are essentially the same player. They'd have the same WAR within less than a point or so. So he wouldn't theoretically have fewer RBI if plugged into Madison's lineup. He'd have been the same as Mick. Age 23-32, they are virtually identical players. And yes, a better hitter being plugged in would be better, but that's not the point. The point is that the other players surrounding Mick caused him to create vastly different results, despite being nearly identical to Grant.

[RonCo]

THat's kind of where I was going by saying Mick could regress and match WAR. The magic 8-ball says "check back later."

[Ted]

The reason I feel that understanding how environment impacts the utility of a player is so critical, is because that's a key way to pick up value. A certain player may simply be worth more to you than someone else. You can trade something that doesn't work as well in your system for something that doesn't work well in another person's, and both GMs come out ahead.

There are teams out there that could benefit from trading a "great" player for a couple "barely above average" ones, simply because what they have currently fits together so poorly. This is an extreme case, but I can think of at least one situation off the top of my head where this is certainly true.

[Ted]

THat's kind of where I was going by saying Mick could regress and match WAR. The magic 8-ball says "check back later."

Whoops, missed that part. Sorry.

[RonCo]

The point of this was to not really debate the utility of WAR, Ron, but to show how it, or any metric, or the ratings, are great in a vacuum, but have to evaluated in the context of how the player is used.

Sure, but WAR is often mis-understood, too...especially the idea that, especially for guys with big splits, WAR is often deeply affected by how a player is used even beyond playing time. The idea that WAR is effectively a counting stat, but can also go backward, is easy to forget.

[Ted]

The point of this was to not really debate the utility of WAR, Ron, but to show how it, or any metric, or the ratings, are great in a vacuum, but have to evaluated in the context of how the player is used.

Sure, but WAR is often mis-understood, too...especially the idea that, especially for guys with big splits, WAR is often deeply affected by how a player is used even beyond playing time. The idea that WAR is effectively a counting stat, but can also go backward, is easy to forget.

Yeah, I probably should have stayed away from it and just gone with ratings for this discussion. It's such a weird stat. It's a counting stat, but it's also predictor AND fair value evaluator. There are so many ways to use it, and thus, so may ways to use it incorrectly and misinterpret it.

[RonCo]

The reason I feel that understanding how environment impacts the utility of a player is so critical, is because that's a key way to pick up value. A certain player may simply be worth more to you than someone else. You can trade something that doesn't work as well in your system for something that doesn't work well in another person's, and both GMs come out ahead.

Yes.

There are teams out there that could benefit from trading a "great" player for a couple "barely above average" ones, simply because what they have currently fits together so poorly. This is an extreme case, but I can think of at least one situation off the top of my head where this is certainly true.

This is true, too...but it's also true that when someone does that kind of thing around here (and around other online leagues!), they often get yelled at because the yellers have a specific view of market value that doesn't match the guy making the trade.

[Ted]

The reason I feel that understanding how environment impacts the utility of a player is so critical, is because that's a key way to pick up value. A certain player may simply be worth more to you than someone else. You can trade something that doesn't work as well in your system for something that doesn't work well in another person's, and both GMs come out ahead.

Yes.

There are teams out there that could benefit from trading a "great" player for a couple "barely above average" ones, simply because what they have currently fits together so poorly. This is an extreme case, but I can think of at least one situation off the top of my head where this is certainly true.

This is true, too...but it's also true that when someone does that kind of thing around here (and around other online leagues!), they often get yelled at because the yellers have a specific view of market value that doesn't match the guy making the trade.

*grimaces sheepishly*

[Ted]

There are teams out there that could benefit from trading a "great" player for a couple "barely above average" ones, simply because what they have currently fits together so poorly. This is an extreme case, but I can think of at least one situation off the top of my head where this is certainly true.

This is true, too...but it's also true that when someone does that kind of thing around here (and around other online leagues!), they often get yelled at because the yellers have a specific view of market value that doesn't match the guy making the trade.

You know, I've been thinking about this. Here's why the person gets yelled at. Just because trading a great player for two "just above average" ones might make your team better, doesn't mean it's a good move. Here's an analogy.

You have a $100 bill. But you can't spend it because the vendor can't break $100. Two $20 bills would be more valuable to you. Do you trade your $100 bill for two 20's? No. That's stupid. What you do is find a guy with two 50's, and trade with him. Then maybe you trade one of those 50's for two 20's. Or if you're lucky the guy with two 20's also has a 10 to throw in.

That's why people get yelled at for making those kinds of need based deals. They're dumb. Taking a small loss to get more workable parts is fine. Taking a huge loss (getting $40 for $100) is stupid.

[RonCo]

We can create lots of scenarios off that example where making a deal for two twenties now would be in your best interest (you need to pay back a $40 loan to Guido the Killer at 4:00, and it’s 3:55), but that’s not a particularly useful scenario anyway, because in general what we’re talking about is variable utility on either side of the deal. So the player I value at $100 today, someone else might value at $160 (and both be perfectly right). So he gives me two players I value at $70 each for that $100. He, of course, is dealing from a position of strength, so for him those players are worth essentially zero dollars.

In that scenario, the guy I trade with gains $160 (his evaluation of my player minus the two guys he values at $0), and I gain $40 (two $70 players in return for one I value at $100). Hence, two winners.

The problem we get in people yelling at GMs who make such deals is that they either (1) do not see a reason the GM traded the $100 player (don’t understand the strategy behind it), or (2) value that $100 player closer to the $160 that the receiver values them at (or higher), hence sees the return as negative.

The argument that “you could have gotten more!” is almost always true if you spend more time. But a salesman, as they say, is always closing, and the final answer on whether it’s a good trade or not is whether the trade increased your horde in the way you want it to.

The main point your initial conversation was focused on is right—different teams value the same resource in different fashions, hence allowing trades that are possible to be good for both sides.

[RonCo]

Another example of totally different valuation…

A big revenue/budget team has $10M dollars he’s going to lose to the ether at the end of the season. That $10M has zero $ value to that big budget team. That same $10M has potentially more than $10M in marginal value to a struggling team, especially if they have a bad contract that they need to move in order to accelerate the process of getting their finances in order (which we both agree is a precursor to making a good run).

The value that low-revenue team should assign to that $10M is:

1) $10M
2) Adjusted up for the marginal value in wins/revenue of the player(s) I can get for that $10M (which assumes I’ll acquire players wisely)

So, if I package a prospect I value at $5M with my bad contract, and give that to the big budget team--understanding that the big budget team is just going to cut the guy and suffer “the loss” in return for the prospect--we can both win.

The big revenue team trades $10M it values at zero for a prospect it values at whatever it values it at…the small revenue team trades a prospect it values at $5M for $10M+marginal utility.

Both sides win, and the league goes apeshit crazy because the big budget team got “something for nothing!”…which is, of course, not true. What they got was something for a resource someone else could use better than they could.

[RonCo]

The point here is that you cannot assess the parameters of a trade without understanding the value structure of the teams in question (their horde, or if not their horde, their strategy...which may not be hording at all--in which case, yelling at the guy means you're upset that he has a strategy that doesn't line up with your views on hording).

[aaronweiner]

The point here is that you cannot assess the parameters of a trade without understanding the value structure of the teams in question (their horde, or if not their horde, their strategy...which may not be hording at all--in which case, yelling at the guy means you're upset that he has a strategy that doesn't line up with your views on hording).

Nonsense. There's a subjective and an objective logic to every trade.

In the example you gave with the team getting "something for nothing," you correctly point out that such a trade isn't, in fact, something for nothing, and no one is strongarming that team into making such a deal. They could choose the deficit and the prospect.

Similarly, just because two people have different valuation systems doesn't mean they can't be absurdly wrong about that valuation. If I actually got $100 from you for two $20s, I'd be feeling pretty darn good. The fact that those $20s will buy more in rural Thailand doesn't mean that they'll buy more than the $100.

[RonCo]

Yes, a subjective logic (based on feeling/personal opinion) is often the reason people yell at others even when an objective logic would say there's reason to make the decision.

This is the root of the phrase that includes "walk a mile in the other guy's shoes," I suppose.

I suppose I should have said "you cannot objectively assess the parameters of a trade without understanding the value structures of the teams in question."

[RonCo]

Similarly, just because two people have different valuation systems doesn't mean they can't be absurdly wrong about that valuation.

This is true too. On both sides.

[aaronweiner]

Is there any threshold at which you're wrong about that, or do you think, for example, that if I dealt my entire farm for a backup catcher that I might be perfectly right?

[RonCo]

In the example you gave with the team getting "something for nothing," you correctly point out that such a trade isn't, in fact, something for nothing, and no one is strongarming that team into making such a deal. They could choose the deficit and the prospect.

yes, and sometimes they would be right to do that, and other times not. The objective answer for whether a decision is right depends on all the parameters, including what the teams are attempting to accomplish at the time (what their value structure and strategies are).

The subjective answer is, of course, in the opinion and gut reaction of the beholder.

[RonCo]

Is there any threshold at which you're wrong about that, or do you think, for example, that if I dealt my entire farm for a backup catcher that I might be perfectly right?

The general statement that either side can be absurdly wrong about something is fundamentally sound, I think.

So by definition, I can be wrong on any immediate assessment of any trade. As a rule I think I'm more often wrong when I use subjective logic than I am when I use objective logic, but .. um ... maybe I'm wrong there, too.

I'll back down here, though, because I don't want to create another flame war in the process of talking process and frameworks.