Tom Tango has a long history of finding new ways of looking at baseball. Tango, along with Mitchel Lichtman and Andrew Dolphin wrote “The Book: Playing the Percentages of Baseball”. The trio introduced the wOBP stat and Tango individually created the Fielding Independent Pitching Stat. Needless to say, the guy has a pretty solid resume it comes to thinking about baseball. In his new work, Tango puts new spin on an old debate: the pitcher win.

Instead of rehashing the old debate about wins, Tango has tried to find a solution amidst all the heat and noise. Instead of banishing the “W” from baseball, he suggests we reformulate how the win is calculated.

The Tango-win takes aim at what happened Monday night: after 15 exhausting innings, the Reds had finally taken a 2 run lead. Logan Ondrusek, despite almost blowing the game by giving up a run, is credited with the win because he was able to get three outs before giving up 2 runs. Both the believers in the win and the card-carrying members of the SABR community can agree this is an absurd result.

To remedy this, the Tango-win asks one simple question: which pitcher contributed the most to help their team win?

Pitchers are graded on the following scale:

For the winning team, each pitcher gets 1 point for every out they record and lose 4 points for every run allowed.

For the losing team, each pitcher again receives 1 point for each out, but loses 6 points per run scored.

The pitcher with the most points on the winning team gets the win, the pitcher who has the lowest total for the losing team gets the loss. If there is a tie, the pitcher that enters the game first is tagged with the “L”.

Here are both the Reds current and Tango-Wins.

Starters W  L Tango-W Tango-L
Cueto 4 3 5 2
Leake 2 3 2 2
Cingrani 2 3 3 0
Simon 6 2 5 1
Bailey 3 3 2 3
Relievers W L Tango-W Tango-L
Broxton 1 0 0 0
Ondrusek 1 2 0 4
Partch 0 0 0 0
Chapman 0 1 0 1
Bell 0 0 0 2
LeCure 1 1 1 1
Marshall 0 0 0 2
Parra 0 0 0 1
Christini 0 1 0 0
Hoover 1 4 0 3

Let’s take a look at some of the games where the Tango system and the traditional W-L system disagree:

Monday, March 31st, 2014

Reds: 0 
WLBs: 1 
(Normal L: Cueto)

Its opening day in Cincinnati! And the Tango is already presented with its first major challenge: what happens when the offense fails get a run across the plate? Johnny Cueto pitches 8 innings and gives up one run. Under the normal Win-Loss system, Cueto takes the L here.

Under the Tango, Cueto gets credit for the seven innings he pitched (+21) but loses points for the one run he gives up (-6). Cueto ends up with a line of +15. Manny Parra takes the L because he entered the game and only recorded two outs (2 points). Parra did give up a hit and a walk, but its rough to deal him the loss when he didn’t ever have a chance to win.

[Update: As Tom Tango points out in the comment section, you cannot get a loss if you don’t give up a run. The loss would go to Cueto in this situation.]

Wednesday April 2, 2014

WLBs: 0 
Reds: 1 
(Normal W: Hoover)

Its game 2! The Tango does a bit better in this game. Tony Cingrani throws 7 shutout innings (+21), but the Reds didn’t manage to cross the plate until the ninth. Under normal rules, Hoover gets the W. Notching twenty-one outs against the Cardinals seems like a greater accomplishment than three.

Tuesday, April 8, 2014.

Reds: 5
WLBs: 7
(L- Ondrusek)

Homer Bailey went five innings (+15) and gave up four runs (-24) for a Tango total of -9 points. Logan Ondrusek did all he could to try to get Bailey off the hook by giving up two runs (-12), but he couldn’t quite do it because he went two innings (+6). JJ Hoover also tried to get his buddy Bailey off the hook but only managed to get a Tango-Total of -3 (1 run, 1 innings).

Now, this game is tough for me because when Bailey left the game, it was tied. The Reds win percentage was 46%. By the time Logan Ondrusek entered the game, the Reds were up 6-5 and had a 62% win percentage. Yet after Ondrusek gave up two runs the Reds had a 29% win percentage. Bailey also gave up almost 40 points of win percentage when he blew a 4-0 lead, but the Reds had pretty much a coinflip in the game. When Ondrusek left the game, the Reds were facing a serious uphill struggle. This one could go either way — it was a photo finish in this race to the bottom.

Saturday, April 19, 2014.

Reds: 4
Stupid Cubs: 8
(L: Cingrani)

Tony Cingrani gave up three runs (-18) over five innings (+15) for a Tango total of (-3). As you can see by the final score, the Reds were still in the game even after Cingrani was pulled. Logan Ondrusek “relieves” Cingrani and promptly gets bombed for three runs (-18) and only records 2 outs (+2). The Reds might have not had much of a chance when Cingrani left the game (they were trailing 3-0), but they certainly didn’t have a shot after Ondrusek doubled the score.

Tango says: Ondru, you lose.

Monday April 21, 2014

Reds: 5
Pirates: 6
(L- Hoover)

Mike Leake went seven (+21) but gave up four runs (-24) in this game. He gets bailed out because even though Leake wasn’t great in this game, the bullpen had a contest regarding who could be worse: Manny Parra 2 outs (+2) and 1 run (-6); JJ Hoover 1 whole inning (+3), 1 run (-6).

This one is another tough call. Parra gave up the tying run in the bottom of eight, yet Hoover gave up a run in the bottom of 9. Hoover was only bailed out because it looked like he was getting a 4 out save. Turns out, that 4 out save only lasted three, and the Pirates got the win.

Wednesday April 30, 2014

Stupid Cubs: 9
Reds: 4

Tony Cingrani had another weak start only going four innings (+12) but he gave up three runs (-18) for a Tango total of -6. When he left the game, the Reds and Cubs had basically a 50/50 win percentage split. Nick Christiani enters the game and gives up two runs (-12) over two innings (+6). Yet bad as this performance was, Sean Marshall bails out his teammate by only recording two outs (+2) while giving up four run (-24). Under the Tango-win, Marshall gets to pick up the check for this one.

Tango-ing With Tango

The Tango system, although better than the normal win, still suffers from the fundamental problem associated with pitcher wins: it still relies on how many runs the offense scores. It does not make much sense to say that giving up a run is only -4 points if the team wins the game but -6 if the team loses. The Reds lost game 1 because they got shut out, not because of a particular pitching performance.

The Tango system bails out starters in these situations, but hangs the loss on relievers because they record fewer outs. This is a pretty constant trend in the table above: starting pitchers, by virtue of pitching more innings rack up more points than relievers.

Consider the following example:

If a “winning” starter goes six innings (+18) and gives up three runs (-12), then a reliever needs to throw 2.1 shutout innings in order to take the win. While we criticize the existing W-L system for not dealing with modern bullpen mechanics, having a reliever appear in 3 consecutive innings seems less reasonable.

Now what happens if the team loses the game? The same pitcher that went six inning (+18) and gave up three runs (-18) has a score of 0. If any of the relievers gave up 1 run without recording 6 outs (once again, this rarely happens), then the bullpen takes the loss.

So the Tango system doesn’t deal particularly well with “middle” performances (giving up 3-4 runs in a start) because the starting pitcher will probably get the W if the team wins, while if a reliever gives up say, 1-2 runs, the loss will probably be hung on the reliever. The margin for error associated with relief performances is much smaller than starters under the Tango-win system.

Considering that league-average ERA is 4.88, it is not uncommon for  relief pitchers to give up a run or two.

All of this is to say that league average starters (ERA of 4.50, or a “quality start”) benefit greatly from this system while starters who are below league average suffer. The Tango system is probably better than our current system in determining who influenced the outcome of the game, but it is fairly unsatisfying. Really, the problem is trying to tie one player’s performance down as the “winning” or “losing” performance. Twenty-seven outs is just too long and complicated to pin it all on one player.

At the end of the day, the idea of giving pitchers wins and losses is so broken that perhaps we should abandon the idea that pitchers lose games and just give the win to whoever has the highest Tango-Score on the winning team.

14 Responses

  1. zaglamir

    This is something I’ve been toying with a lot lately. Perhaps the nicest solution I’ve found so far is this (though it’s only a counting stat for starting pitchers):

    Going only by score when the pitcher leaves the game (remove the bullpen aspect), assuming they go more than 4.2 innings(more than half game). Figure out his season Run Support/Inning Pitched:

    RSIP = (Runs Scored by Pitcher’s Team During his Innings Pitched for season)/(# of Innings Pitched in season)

    Compare Earned Runs vs RSIP*IP for each game, to determine Win/Loss. This will give you a win counter that is more independent of slumps. There would be four outcomes: W, L, ND(tie), and Incomplete (didn’t make it half the game).

    To make it totally offense independent, could use the league average run support per inning.

    Essentially, this pits the two pitchers against one another, seeing who was better compared to league average against the opposing offense that day. It would leave in the aspect of facing the Astros contributes more Wins than facing the Tigers, because the offense your facing contributes to your ER total, and your offense doesn’t hinder you.

    I wish I had time to pull up all of the data from the past 20 years and look at whether this “adjusted win” correlates from season to season with overall performance. I just don’t have the time.

    (Or perhaps, one of you folks would like to point out a flaw in this idea that I’m overlooking? It seems like it should work, as a counting stat, but I’m unsure if I’m missing something obvious. In theory, it measures whether you were above or below your average for that day, independent of whether your team was in a slump or not).

    • Michael Maffie

      If I understand correctly, you are suggesting we use league average ERA (4.88), then compare that number to how each pitcher performed on any given day. Whichever pitcher perfumed better gets the “w”.

      This does isolate the offenses from each other, but is essentially the same thing as the Tango without the changing point penalty for wins and losses (Runs/IP is comprised of Runs*[Outs/3]).

      One of my main objections to the win (in general – not your modification) is that it doesn’t tell us anything that other stats don’t tell us (FIP, ERA, RA, xFIP, SIERA etc). In fact, it can be misleading because we add all the variance associated with a team’s offense into the mix.

      But, if we gotta award pitchers the win, finding a way to break the offense away from it would be a good first step.

      • zaglamir

        Essentially, yes. The only thing that I would suggest different to what you’ve summed up is that we’re not using League Average ERA, so much as a teams average runs per inning. The Reds offense is not equivalent to the Tigers offense. So, instead of using League Average (I did add a line in the first post saying that might be worth investigating, so I understand the confusion), we’d be seeing what a pitcher SHOULD have gotten (W/L) if his offense had performed at their average level. For this stat, it would be possible for both pitchers to get a “Win.”

        Example: Leake holds the Cardinals scoreless through 6, Wacha holds the Reds to 2 runs through 6. On average, the Cardinals will have scored more than 2 runs with 6 innings, thus Wacha gets an “adjWin.” On average, the Reds will have scored (slightly more than) 0 runs per 6 innings, thus Leake also gets an “adjWin.”

      • Michael Maffie

        Sorry about the confusion — I get what you are saying now.

        I think that would be rather shocking to the reader because the “adjWin” vs “normal wins” — Im guessing — would be pretty big. The variance in runs scored is pretty big, although it does tend to converge around 4-5 runs per game.

        One problem (and this would be difficult to fix) is that team runs scored will float throughout the season. Rosters change, trades are made, injuries hit teams later in the season, players wear down. Im guessing it takes over half a season for that number to stabilize. It probably would be more useful in a retrospective way than during the season.

      • zaglamir

        Agreed with the problem. That’s where I’ve been stuck. I just don’t know how to make it a stat that isn’t “retroactive.”

        Also the variance on runs per game isn’t as large as you’d think (otherwise, I would have dropped this idea immediately):

        Assuming a Poisson Distribution (roughly correct here), sigma is only 2.0. That means 67% of games fall between 2 and 6 runs. That’s a wide range, but not as huge as I would have thought.

      • Michael Maffie

        I had seen that article when I was writing the quality start article but had forgotten about it. Thanks for posting.

      • zaglamir

        A thought I have had is to find a way to calculate: “How many runs should this, LINEUP score.” Some combination of OPS, LOB%, etc and then use that as a standard against the Earned Runs by the pitcher could be measured.

        The real take-away is: I want to be less of a nuclear physicist and more of a guy who looks at baseball stats all day.

  2. cfd3000

    I actually don’t have a problem with wins and losses as a long term measure of pitcher performance. Meet me at the bar and we can discuss that properly though – no way there’s enough space to have that conversation in full in this format. Nobody gives any real credence to W/L totals for relievers so why waste energy trying to fix it for relievers? And in the long run (meaning over years or even a career) W/L totals do tend to correlate with our general regard for starting pitchers. If for no other reason than utility for comparisons with bygone days I say leave it alone. It’s fascinating, for example, to discuss who will be the next 300 game winner or if anyone ever will. Don’t take that away. One of the beauties of following this great game is the rolling link to the past that allows for comparisons of players from different eras. And we’ll continue to assess CYA candidates, trade options, potential call ups, rotation merits and the like based on more insightful stats like whip, era, FIP and xFIP, K% and BB%, etc. etc.

  3. tangotiger

    Please note that there is an exception to the rule: you can’t get an L if you don’t give up a run.

    Also note that this is a work-in-progress, so numbers are changing slightly.

    Another request I saw was to charge a W to either the SP, the bullpen, or the Offense. As others have noted, no one really cares about a W for a specific reliever.

    • Michael Maffie

      I updated the article to reflect the 0 runs exception.

      I like the idea of aggregating the bullpen performance.

  4. Groujo

    Intriguing idea. The best part of it is that it eliminates the TIMING issues. It has always seemed to me that the biggest injustice in the current system is that a pitcher can lose the win just because his team falls momentarily behind after he leaves the game. [e.g., Cueto pitches 8 shutout innings, and leaves the game up 1-0. Chapman allows a run in the 9th and we win in the bottom of the 9th. Chapman gets the win????] Any change that eliminates the issue of when your offense scores will be a better indicator of performance and which pitcher actually “won” a game.

  5. hof13

    I haven’t had a chance to read what has been written was playing around with numbers. I took the Tango score given and divided by 0.8 times (innings pitched) to get a modified Tango score. It yielded the following results.

    game 1 above: Cueto 2.67, Parra 3.75 (Loss Cueto)

    game 2: Cingrani 3.75, Hoover 3.75 (winning tie goes to most innings, Cingrani)

    game 3: Homer -2.25, Ondrusek -3.75, Hoover -3.75 (losing tie fewest innings, Hoover)

    game 4: Cingrani -.75, Ondrusek -30 (Ondrusek by a tall margin)

    game 5: Leake -.54, Parra -7.5, Hoover -3.75 (Parra )

  6. yoobee

    One question: it seems like the number of wins and Tango wins should be equal, but they are not in the chart above. Why is that?

  7. Lucie

    Hello there, You have done an excellent job.

    I’ll certanly digg it and personally recommend to my friends.

    I am confident they will be benefited from this website.