My friend Tom Tango has more ideas than he can keep up with. One of those ideas is something he calls the Indis — to be pronounced IN-dees. The idea is that the name would sound like the nickname everyone used for Indiana Jones. The logo for the Indis is the Indiana Jones hat. I have told Tom that The Indis would look better from a purely superficial perspective if he called them the Indees or even the Indys. So far, I have not convinced him.

Indis is shorts for Individualized Won-Loss records, and it is something that he came up with years ago. He described it like so: “I have a simple method to convert Wins Above Replacement (WAR) into an individualized Won-Loss record for each player, such that the sum of the players’ individual Won-Loss records on a team will match that team’s Won-Loss record.”

Simple, right? I can give you a basic (and only a basic — I’m not very smart) explanation for how he does it.

Each player gets what Tom calls “game spaces.” I like to call them opportunity spaces because that’s what we are talking about — these represent the opportunity that this player has to contribute to the team. We all understand that baseball is a rare sport because the manager has only a limited ability to put players in what you might call a winning position. In the NBA, if you are down one, you can and will absolutely set up the play for James or Curry or Harden. At the goal line, time running out, you will call for Brady to throw or Gurley to run or Brown to get open in the end zone.

But down a run with two outs in the ninth, you can’t just have Trout or Harper or Stanton hit. It has to be their turn. You can’t just start Kershaw or Scherzer every time out either. So each player only has a limited opportunity to help the team. Joey Votto doesn’t get substantially more of a chance to help the team than Adam Duvall or Scott Schebler. Corey Kluber only got three more starts this year than Josh Tomlin.

So each player gets opportunity spaces. Tom divides them up like so.

Position players get 4/7 of the games — adding up to 93 opportunity spaces for all position players.

Pitchers get 3/7 of the games — meaning they have the other 69 opportunity spaces to divvy up.

Tom gives each player his individualized spaces based, for now anyway, entirely on playing time. There are other very interesting ways to divide the opportunity spaces, and we can talk about those later. But let’s explain where the Indis are right now.

Let’s look at four everyday players from last year: Joey Votto, Giancarlo Stanton, Aaron Judge and Jose Altuve.

Votto had the most opportunity spaces of the four with 11.7. Ths is because Votto played all 162 games and got 707 plate appearances. Giancarlo Stanton, who played in a career-high 159 games, was just behind with 11.5 opportunity spaces.

Judge and Altuve both played fewer innings, so their opportunity spaces were a bit less at 10.4 and 10.2 respectively.

So what next? Tom uses WAR to determine how much value the player was able to squeeze into his opportunity space. Votto was so good he stuffed 10.8 wins in his 11.7 opportunity spaces.

That makes his record a sparkling: 10.8-0.9.

Yes, it is weird to see decimal points in won-loss records, but it’s hard to work around it. You certainly could round the numbers, and that works pretty well when talking about a whole player’s career. But for one year, when you round up or down, it doesn’t always work. In Votto’s case because of how close his totals are to the round number, it’s fine, he would be a sensible 11-1. But a player who went 11.4-0.5 would also be 11-1, and that’s not quite the same thing.*

**Bill James in his Win Shares formula fixes this problem by multiplying the numbers by three, which would make Votto 32-3. This gives you the advantage of not having to round off the numbers and having them look much more impressive. But Tom is not crazy about that idea because one of the things he loves about the Indis is that the numbers actually add up to a team’s win-loss total. You can say, somewhat reasonably, that Votto was responsible for 10.8 of the Reds’ 68 wins while only contributing .9 to their 94 losses. *

OK, that’s Votto. Stanton was even better. He stuffed 11.1 wins into his 11.5 opportunity spaces.

This makes Stanton: 11.1-0.4 (or, multiplying by three, 33-1).

Now, there’s something about Stanton’s record that needs to be discussed: The Marlins won more games than the Reds, 77-68. This means that there are more wins for the Marlins to allocate. The simple but powerful idea here (and behind WIn Shares) is that the reason a team wins more games is BECAUSE the players offered more value. This will become clearer (and a bit more controversial) when we look at Altuve and Judge.

Jose Altuve, as noted above, had 10.2 opportunity spaces. But here’s where it gets spacy; he actually “won” 11.3 games. He was so good he contributed MORE WINS THAN HIS OPPORTUNITY ALLOWED. The greatest seasons and careers are like that. A few examples:

In 1923, Babe Ruth was so good he fit 19 wins in 12 opportunity spaces. Barry Bonds in 2004 had 16 wins in only 10 opportunity spaces. Willie Mays was so absurdly good for his career that he had 210 wins in 183 opportunity spaces.

So, how do you denote that sort of crazy goodness? That’s tricky and I’ll be interested to hear your points of view. One way to do it is to use the negative number:

Altuve: 11.3-(-1.1)

Ruth: 19-(-7)

Bonds: 16-(-6)

Mays: 210-(-27)

Another way to do it is to simply replace the hyphen between the wins and losses with a plus sign.

Altuve: 11.3+1.1

Ruth: 19+7

Bonds: 16+6

Mays: 210+27

Again, your comments are welcome below. But the point is that a player can be so good that they basically go above and beyond their space is a very cool concept. In their cases, Altuve and Judge BOTH went above and beyond.

Altuve: 11.3+1.1

Judge: 10.6+0.3

Now, here’s where it gets interesting. Judge and Altuve had essentially the same WAR; Judge actually had a higher fWAR. And yet, Altuve has the better Indi. Why? Well, you know why: Because the Astros won 10 more games. The Astros had more wins to divvy up among players. Now the question is: Should Judge be docked or Altuve given extra credit because of how many games the team won? This was at the heart of Bill James’ criticism of WAR because it does not attempt to look at specific wins and losses. It is a context neutral statistic, and the Masters of WAR have made a powerful and convincing case that this is exactly how it should be.

But isn’t there room also for the Indis? Tom has latched onto a phrase that does a nice job of explaining what something like the Indis can determine: “Extracted value.” If you believe that Altuve and Judge put up the same value — a fair assumption — it is also worth knowing that the Astros EXTRACTED more value from Altuve than the Yankees did from Judge.

We can see this in all walks of life. Let’s say you have two authors who write equally good books; I’m not sure how you would judge this, but let’s say there was a WAR formula for books and each book was 8.7 wins above replacement book. They’re both wonderful, funny, charming, fascinating.*

**One of these books is about Harry Houdini, by the way.*

The first book gets a better cover, is edited better, has a better title, gets a few huge publicity breaks and sells a million copies.* The second book has a boring cover, is sloppily edited, is called “Blah blah blah blah,” publicity just doesn’t connect and it dies on the vine, selling only a few thousand copies.

**This is the Houdini book.*

I do want the WAR formula to tell me about the quality of the book, absolutely. But I also want something to tell me that, yeah, the first book was much more successful than the second. You can argue, “But the authors had little to nothing to do with the difference, one just got lucky.” Maybe so. But reality is reality. Let’s go to the Indis!

The Indis say that Astros won more games and got more out of Altuve’s performance than the Yankees did out of Judge. I mentioned above that the opportunity spaces are now divided up by playing time, but Tom says they could in the future be divided by something else, like RE24 (Run Expectancy) or WPA (Win Probability Added). That could give an even fuller picture. The truth is that Judge DID hit much better in low leverage situations (.299/.442/.673) than in high leverage situations (.219/.361/.500). He DID hit 28 of his 52 home runs with no one on base. We’re not placing fault here, but because he did those things, the Yankees won fewer games than they would have if his hitting had been better spread out.

And that’s what the Indis show.

We’ll talk next about pitcher Indis which are fascinating in a whole other way, but in the meantime I want to repeat the request that you offer some of your thoughts in the comments. Tom has too many things going on to spend a lot of time on the Indis now, but I really think that this type of won-loss idea — added to the beauty of WAR — would be hugely helpful in judging baseball players each season and over their careers. I think it’s kind of raw now, and I’m not entirely sure how you get around some of the awkwardness of negative losses and decimal points. But I think we should encourage Tom and others to keep on working with this concept. It’s pretty fun.

This seems like an interesting way of understanding the relationship of individual player performance to team wins but I’m still resistant to the idea that we should evaluate a player based on what his teammates did.

Could you run the numbers on a more extreme split than Altuve/Judge? For example, the next two names on the Fangraphs WAR leaderboard are Anthony Rendon and Giancarlo Stanton, both at exactly 6.9 WAR, with Stanton playing 12 more games than Rendon. How would this stat describe the difference between the two of them given that Rendon’s team won 97 games and Stanton’s team won 77?

I want to let the comments pile up but the Indis but you asked about Stanton and Rendon.

Stanton: 11.1-0.4

Rendon: 8.9-1.0

Indis is a worthless name, and win-loss record is a bad way to express it. Call this Team Value Adjusted WAR (TV-WAR), subtract losses from wins in the formulas leaving the value expressed as one number, and you really will have something like what those of us who like and understand WAR as a predictor of raw talent and future performance, but don’t think it is a proper measure of value when discussing past value (i.e. evaluating players for the MVP award), have been asking for. Precisely because it’s a personal performance metric with a team performance adjustment expressive of real extracted value.

One reason my version will get resistance: it will, of course, mean that half of players are in the negative numbers (it’s also TV-LAR), and no one will like the looks of that. But just as you get used to it when you look at plus/minus on a lousy hockey team, you will get used to it when you look at all those negative numbers on a lousy baseball team.

The first thing that strikes me is that even setting aside the whole “negative wins” thing, I dislike the notion of a W-L record that doesn’t look any like a real W-L record.

I read James’ Win Shares when it first came out, and never cared for the idea that he was letting a player’s value be bound by his team’s record, and so when WAR came along, I, like lots of others, happily jumped ship. Maybe we’re wrong, maybe it Beta vs VHS all over again.

The recent return of this debate brings to mind the Heisenberg Uncertainty Principle: we can measure how good a layer is or how much he helped his team, but we can’t do both at the same time.

The leaderboard is here:

https://twitter.com/tangotiger/status/929346695862915072

And the “extracted value” term came at the tail-end of this podcast:

http://tangotiger.com/index.php/site/comments/war-podcast

I like it!. I prefer the plus sign to the double minus.

And I am happy having both stats.

I’m a Sabremetrics (and Joe Pos) fanboy, but no expert, especially in statistics. For any of these models, I would be interested in not just the point estimates, but the error around the estimations, when trying to judge superiority/inferiority.

Stanton hits 59 HR with a total of 29 players on base.

Judge hits 52 HR with a total of 33 players on base.

Judge leads in OPS with RISP – 1.015 to .892

Judge leads in OPS with RISP and 2 outs – .890 to .722

Judge leads in OPS “Late and Close” – .760 to .745

Stanton leads in OPS “High Leverage” – .982 to .861

*

Stanton gets 70 “High Leverage” PA against the NL East and has a 1.061 OPS.

Judge gets 53 “High Leverage” PA against the AL East and has a .739 OPS.

Stanton gets 60 “High Leverage” PA in other games and an .893 OPS.

Judge gets 69 “High Leverage” PA in other games and a .955 OPS.

*

“the Yankees won fewer games than they would have if his hitting had been better spread out. And that’s what the Indis show.”

*

That is simply not true. The Indis show that against a much weaker schedule the Marlins won 77 games and Stanton got a proportional amount of that credit relative to his overall play. Judge got a proportional amount of credit relative to his overall play of the 91 Yankee Wins.

If a tool like this were used for determining the MVP, it seems that we would be limiting the MVP to teams that won a lot of games. How would Ernie Banks’ 58-59 seasons stack up looking at WAR vs. Indis?

He’d look fantastic in 1959:

http://tangotiger.net/wonloss/index4.php?yearid=1959

And in 1958, he’d be in the company of Mays/Mantle, ahead of Aaron:

http://tangotiger.net/wonloss/index4.php?yearid=1958

I guess I’m just not sure how useful this stat (or, specifically, the component of the stat which takes into account team win totals) is. We already have a fairly useful stat to take into account how many games a team won: wins. Using wins to modify player WAR totals in a way that’s not fundamentally related to a player’s performance just doesn’t seem useful.

*

To look at the books analogy for a second: we already have a stat that captures the disparity in how many copies each book sold, and again, that stat is “how many copies did each book sell”. Now, we can try to use statistical analysis to understand *why* those books sold as many copies as they did, and try to estimate the quality of the book, and the effect of publicity, and the effect of advertising, and ultimately to combine all of those into an explanation of the total sales. And the way to do that is to attempt to model all of those things. So, similarly, when you’re trying to evaluate the quality of the book and its effect on sales, I think the best goal is to try to model that as best you can, and not conflate it with other things. I agree that it’s good to know how many copies the book sold in total is useful. That doesn’t mean that it has to be imported into your opinion of the actual literary quality of the book. We already *know* how many copies the book sold in total. Isolating the quality of the book is fundamentally exactly what we should be trying to do with BookWAR. And the same is true of the baseball WAR that we’re talking about here. That would be my concern, basically.

What the Indis does is “undo” the flattening of a player’s W-L record into the single-dimension WAR, and restores it to its two-dimensional value.

There are some interesting ideas here certainly and it’s very important to get back to actual wins versus hypothetical ones, lest we wind up like the statistician claiming that he hit the target before he fired the shot. Context matters and Indis seems to have some ways to adjust for this.

I’d still like to see more information about how this is calculated and am wary of increasingly complex models, as adding an additional variable includes (at least) an additional assumption that is often hidden from the people consuming the information.

I may up a tree here, but the object of the season for each team is to maximize wins, starting at zero and increasing each time the team outscores its opponent. Or, put another way, losses are simply lost opportunities to win. In that sense, I’m not sure there’s a need for a player’s win-loss record. In fact, I’m not entirely sure how to interpret it: does it really make sense to say that Joey Votto’s play resulted in eleven wins and one loss? Rather, doesn’t it make sense simply to add the wins and losses together and say that Votto’s performance contributed ten wins to the Reds’ total. Likewise, Altuve’s performance resulted in 12 wins (or 12.4, or whatever) for the Astros. I think this has the advantage of being intuitively more accessible. I know it’s a minor point, but it might help help the stat to gain acceptance. Just a (possibly dumb) thought…

This is off topic – still digesting the stat – but I see that Joe mentions attaches a negative connotation that Judge hit 28 of 52 home with no one on base. I see or hear this quite often in regards to players who hit more home runs with the bases empty. But it ignores the fact that the majority of home runs leaguewide are hit with the bases empty. In 2017, 59% of home runs were solo shots. Some of that was due to opportunity as 56.7% of all plate appearances came with the bases empty. But players still hit more home runs with the bases empty regardless of the PAs. The league hit 3.43 HR/100 PA with the bases empty, 3.12 with men on.

Judge only had 52.8% of his PAs with the bases empty, and hit 53.8% of his home runs in those situations. Perhaps a hair better than the league but pretty much neutral in comparison to normal ratios. It is not a negative that he hit more HR with the bases empty.

Also, in terms of digesting the idea, do the “losses” from a team perspective add up to the team’s losses as well? In other words, since Altuve has 1.1 more “wins” than opportunities,do the rest of the Astros combine for 62.1 “losses’?

Yes, it all adds up.

I like the idea, but I think there are a couple of flaws in the logic.

1. This Indis calculation is based on the assumption that a replacement-level player contributes zero wins. This is not true. Hypothetical teams of replacement-level players win anywhere from 50-69 games in a season, based upon different studies I have read, some of which I thought were on this website. Therefore a player with 10.8 WAR is contributing more than 10.8 “wins.” In other words, should a replacement-level player with opportunity of 10 games be given an Indi of 0-10? That seems wrong to me.

2. Bill James didn’t multiply by three to avoid fractions. That would mean that each win share was a multiple of 1/3. He handled the rounding elsewhere. I thought that he multiplied by 3 mostly for cosmetic and presentation purposes. However, I haven’t read his Win Share theory in very many years. Joe, you must have known you were wrong before you posted the article, because 10.8-0.9 multiplied by three does not result in a pair of whole numbers.

This stat is a nice idea, but I think it needs more work. If I am missing anything in the logic, then please let me know.

“1. This Indis calculation is based on the assumption that a replacement-level player contributes zero wins.”

This is not true. A team of replacement level players will have 48 wins and 114 losses using The Indis.

***

As for the multiplying by 3, Bill indeed does round at the end.

If a team has players that are exceeding their opportunities to contribute, is it reasonable to assume that there are people on the team bringing the team down? There must be a reason the Yankees did not extract all those wins out of Judge. If I’m a GM, Ill put as much effort into tracking the bottom end of the scale. The Astros must have fewer players with poor W-L record.

It’s a interesting article, and an interesting way of looking at things. But. Let’s take Votto’s last five years and…some average guy on the best recent five years of the Dodgers. I could be way out, but it seems that using Indis, those two guys would be closer together than they would be by any of the traditional WAR-style metrics. What does this tell us, other than Votto has given his best years to a relatively terrible team, and some other guy was cheap or friendly or healthy or whatever? Is it a boon to our understanding of the game?

Or that might not be the reason for this stat, and I might have skimmed the relevant bit. My apologies if so, and my apologies if you’ve already had to answer variations on this dumb question a hundred times already. I think any individual stat that factors in what the rest of the team did ought to be really careful – yes, I’m some dumbass on the internet and y’all earn your living writing and thinking about baseball, so I am almost certainly in the wrong here.

I think there’s a fundamental misexplanation if you’ve come to this conclusion. This is Joey Votto’s career, from best to worst:

11 – 1 : 2017

10 + 0 : 2010

10 – 1 : 2015

10 – 2 : 2013

10 – 2 : 2011

8 + 1 : 2012

8 – 1 : 2009

7 – 4 : 2016

6 – 3 : 2008

3 – 1 : 2014

0 – 1 : 2007

How would mid-season trades effect the W-L record matchups? Let’s say Votto is 7 of the Reds’ 36 wins before the trade deadline and then is traded to, who cares, the Padres who have 24 wins, or the Astros who have 49 wins. How would Votto’s Indis score be affected? How would the elegance of the totals matching the team’s W-L be impacted? What about trades for prospects who don’t play in MLB that season?

The accounting is (currently) based on the seasonal performance.

Nothing however is stopping us from applying the logic on a game by game basis. Once you do that however, you will likely not like the implications being made. Basically, we’re not ready for it.

As for prospects: we’re simply accounting for actual wins by actual performance.

“Tom uses WAR to determine how much value the player was able to squeeze into his opportunity space. Votto was so good he stuffed 10.8 wins in his 11.7 opportunity spaces.”

For us dummies, this is missing a step. What is the calculation here?

Boy, do I love this stuff. I have a general comment about these great articles and comments about WAR, dWAR, Indis, et al.

I’d like to start by saying that the object of playing a baseball game is to win it. In my view, the advanced analytics are an attempt to identify the inputs that contribute to wins. The inputs, any and all stats derived from outcomes of played games, are the independent variables. The outputs are wins. They are the dependent variable. The dependent variable (wins) is what we’re trying to explain (actual performance) or predict (future performance). Do not think of wins as less important because it is called the dependent variable. Similarly, do not think of stats as more important because they are called the independent variables. The most important variable is wins. I submit that any attempt to “adgust” actual win totals is messing with the output (independent variable) and is fallacious.

Sorry, last sentence should read “I submit that any attempt to “adjust” actual win totals is messing with the output (DEPENDENT VARIABLE) and is fallacious.”

I’m surely missing something, but WAR alone does not add up to a team’s actual wins. How does Indis resolve this issue? Sounds to me like you’re taking WAR and somehow converting it to Indis, and this adds up to actual wins?

Sorry I am late. I did not see the pitching Indis article. I like the stat but have some unease. I think the unease is that a team with great pitching will inflate the numbers of that team’s hitters, and a team with poor pitching will do the opposite. I understand park effects etc but there will still be differences based on pitching. For example, for the Dodgers, a home run when Clayton Kershaw is pitching would be worth much more towards a win than a home run when their worst starter is pitching. And over a season, a player who hits 30 HR for the Dodgers will get a lot more credit (or “Indis”) than a player who hits 30 HR, even in the exact number of plate appearances, for the Padres.

I get that there is friction between measuring a player’s performance vs. a player’s contribution, could there be a hybrid of split of those measurements, like JAWS combines career and peak performance?