About three days ago my measured probability that the existence of Win Probability" could become a contentious topic was around .01%, but once again this board has made manifest an argument from the seemingly empty aether.
Someone really needs to chart the changing Existence of Win Probability As Contentious Topic (EWPACT) by post number. That would be fascinating.
Exactly so. It’s far more interesting than things like Pitcher Wins or RBIs which don’t tell you much about how good a player is (rather it tells you more how good a player’s team around him is). I like to know how probably a win is for a team at some point of time in a game and how improbable a comeback was.
Imagine this scenario for me, if you will:
You and I decide to meet up at Camden Yards to watch a game together. The visiting team is the Yankees, but we both root for the Orioles. We settle into our seats with a beer and some hot dogs, and the game begins. It’s a pretty tight affair for the first couple of innings, but then Baltimore goes up by a run in the third off a big home run. In the top of the fourth, the Yankees put a guy on first, he steals second, and then comes home on a ground ball up the middle, so it’s tied 1-1 in the middle of the fourth.
The Orioles go 1-2-3 in the bottom of the fourth, and then in the top of the fifth things start to go bad for Baltimore. The Yanks get a lead-off double, followed by a strike-out and a walk. Then there’s a ground ball to third, but the fielder boots it, everyone’s safe, and it’s bases loaded with one out. The next batter connects with a hanging slider and deposits it onto Eutaw Street. Now it’s 5-1 Yankees.
The Orioles make a pitching change, and during the break in the game I turn to you, blow my cheeks out, and say, “Well, that sucked. What sort of chance do you think they’ve got now?”
Do you sit there and say, “Well, surely there’s no point discussing this unless we’re intending to wager some money on it. After all, we’ll know the result in about 90 minutes anyway, so what’s the point of speculating?”
Would it not interest you even the slightest bit to know that there have been 3,923 games since 1957 in which the visiting side was up by four runs with one out in the top of the fifth inning, and the visiting side has gone on to win 3,544 of those games, or 90.34%? And can you not grasp that, even if this number does not in any way guarantee the outcome of the particular game we’re watching, it might be a pretty damn good guide to the relative chances of the two teams?
If you’re completely indifferent to these questions,. and can’t see how they might add something to a person’s enjoyment of the game of baseball, well…I don’t exactly know what to say about that, except that you and I are very different.
Well, no, you’re advocating ignorance. I don’t understand that. I am honestly curious as to why you think we should have less information.
I don’t understand this either, and in this case I am honestly struggling to believe you are not joking.
As to whether fans are constantly interested as to the likelihood of a team winning the game, of course they are. The primary point of interest in a sporting event is who will win. Fans care about “what the players are actually manage to accomplish on the field” because that determines who will win. Fans talk about comebacks and blown leads because those terms imply an unlikely event that beat the odds. The most famous games in MLB history, like the Bobby Thompson game, are generally games where a team came back from a poor position (or blew a good position, depending how you look at it.)
This is obviously wrong. I don’t bet on sports at all, and yet the probability of who’s going to win is something I think about as I watch a sport, because that’s why I’m watching it. If that didn’t matter to me, why bother watching? I could find out who won by reading the news.
Much of the drama of watching a baseball games is the implicit notion of win probability. I want my team to score runs and prevent them because doing those things makes it likelier they will win.
Today the Blue Jays fell behind 8-0 to the Marlins. As I type this, they have mounted a spectacular comeback, and the game is tied 11-11. The concept of win probability is a big part of the drama of this game. If they can win, this will be an incredibly thrilling game, much more so than had they won 11-0, because coming back from 8-0 is so amazingly rare. I don’t know what that WP is, and I honetly wish I did because I’m curious; surely it’s not even five percent, is it? Is it one percent? I dunno, but this is really cool, and it’s cool BECAUSE it’s unlikely. I have no money riding on this game, I’m just a baseball fan.
As to “why not just simulate the games,” in no way does that silly idea proceed from the idea of wanting more information about the history of baseball. I’ll ask you the same question; why is ignorance preferable to knowing something?
I do not, and never will, understand people who have a hostility to knowing things. I mean, do you complain that they still count errors even though it’s an amazingly flawed statistic?
I propose that (EWPACT*Post Number)/ Characters per Post Length might be a more meaningful metric.
According to Greg Stoll’s Win Expectancy calculator, a visiting team that’s up 8-0 in the middle of the third inning, which was the situation today, has a 98.1% chance of winning the game (356 wins out of 363, historically), meaning that the Blue Jays had just under 2% chance when they started batting in the bottom of the third.
It looks like they’re probably going to lose the game now, but it was pretty amazing that they got back to even after that start.
Will this hijack ever end?
Not likely.
So how bout them Yankees?
I was semi-excited when I saw that this thread had so many replies. Ugh.
C’mon, who doesn’t love a good extra innings game?
With my Phillies finding so many creative ways to lose in this weird year, it’s pretty rare for Win Probability to ever enter the discussion.
ETA: I was typing while all the anti-hijack commentary just above was being written. So I’m not trying to be in your face. The timing just worked out that way.
That is a very nice summary. I’m on the “WP is (largely) meaningless” team so let me respond.
Bottom line up front: WP is not absolutely meaningless. However, the error bars on it, as applied to a particular game in progress are very large. A fact completely ignored by all telecasters and IMO many fans.
The error bars are much larger than the confident numbers coming out of the computer & the sportscasters’ pie holes. Instead of saying it’s e.g. 87.5% they ought to be saying it’s 85% +/-15%. That would be an accurate indication of the actual predictive value everyone is crowing about enjoying.
Not nearly as compelling a narrative but much more accurate as a matter of statistics.
The easiest synopsis of the argument against comes down to this:
WP as computed and telecast, assumes you’re looking at:
- Two completely average teams.
- In completely average physical condition (injuries, etc).
- Staffed by completely average players and managers.
- Who will today individually deliver a completely average performance.
- Who will today collectively deliver a completely average performance.
- In a ballpark of completely average playing nature (e.g. “hitter’s park” vs “pitchers park” vs “catcher’s park”).
- Under completely average sun & weather conditions.
- As umpired by completely average umpires.
As well, all the historical differences between rules, balls, player skills, strategic techniques (e.g. the shift, the DH), park size & design, mound height, etc., from 1880-whenever to today are fully averaged out. They aren’t playing in 2020; they’re playing some weird 1950-ish averaged sort of game.
Once you apply all those caveats to the computed WP, you begin to understand just how massively artificial it is. As simple algebra it’s unassailable. As an actual description of e.g. the 2019 Yankees vs the 2019 Marlins in NYC on a particular day in Aug 2019 it’s a joke.
What is missing from their confident calculation is far more impactful on the particular game at hand than what they are including.
No sane person would argue against the straw men in Jeff’s 2nd-3rd paragraph. The actual ebb and flow of success and failure on the field and most especially on the scoreboard surely affects the likelihood of each team winning as the game unfolds. And is fun to watch. The see-saw games are by far the most engaging. Precisely because humans love to predict things. No matter how crappy their prediction heuristics are, they still make confident predictions. My point, and I believe that of @Acsenray is that WP as actually computed does a really shitty job of capturing that. It’s a shitty prediction, but we love predictions. So many people fall for it.
Here's another related argument:
The WP of all of MLB history is 50%. Before the game starts that’s all we can say; by the end there’ll be one loser & one winner & MLB’s lifetime record will still be 50%.
Once we add in the fact one team is home and one is visitors, the WP shifts slightly. Google tells me it’s 54% in favor of the home team before the game starts.
Now if I could get a bookie to offer me payoffs (less vig of course) reflecting 46%/54% visitor/home likelihoods on every game every day I’d be a rich man.
They don’t offer those odds. Instead they say “Hmm, Yankees at home vs. Marlins; I like a 26%/74% likelihood a lot better”. Using some mix of math & intuition to set their lines by considering the particulars of this individual game at this moment in the teams’ histories.
WP ignores all that; it’s 46%/54% and that’s all the averaging calc in the computer can see or think. It is blind to the entire rest of the particular factors affecting this particular game.
Now where WP *might* have some small value is as a change-o-meter. If they started the game with a more conventional bookie-style WP for the two teams and then simply talked about the changes as the game progresses that would be more valid (though still not completely so), and might be interesting once the fans got educated.
E.g. after single followed by nice double puts men on 2nd & 3rd with 1 out in the top of the 7th when behind by 1 run they could say something like “Well Bob, that double bumped their WP by 2.5%; things are looking up for the visitors”. And when the next batter hits into a double play: “Well Bob, that ends the inning and their WP has slumped 4% from just before this last batter came up and 1.5% from when their half-inning started. Time is running out for the visitors”.
That captures the concept that game play & the evolving current game state affects actual winning likelihood. But …
Even using this “Deltas only, not absolute values” approach there is still all the problems with averaging.
Against the Yankees at home I’ll take Red Sox down by 1 at the middle of the 7th over Marlins down by 1 at the middle of the 7th any day. But the WP and therefore WP Delta says the situations are equal. I think we can agree they are not equal. Which means the Deltas need error bars too, or at least a way to tweak them towards the non-average particulars of the instant game.
My bottom line:
Statistics in general are a fuzzy squishy thing. Every number needs an error bar. And every error bar needs an error bar. Statistics are highly predictive of processes with low inherent variability like lottery draws and widget-making machinery. They are less predictive of processes with high variability. Human and human group performance has high variability. Humans like simple-to-grasp predictions. At some point you end up with more bathwater than baby. Most baseball fans don’t seem to grasp all the above. So they end up cooing at mostly bathwater.
You could move this hijack to another thread, that would be cool.
You know what you could do, instead of complaining? Write some posts of your own about the game you watched yesterday, or about how your team is doing, or about how the coronavirus is still fucking with the schedule. It’s possible to have more than one conversation about baseball in a thread about baseball.
I watched the Yankee game last night. Nobody blathered on endlessly about boring shit. It was cool.
SI’s Tom Verducci posted an article yesterday, in which he discussed that a big part of the reason why MLB is trying so hard to keep the season going is the monetary payout (from the TV networks) for running the postseason.
He also notes that there are discussions underway to potentially hold the postseason in a bubble, and play all the games at neutral sites. As the COVID outbreaks on the Marlins and Cardinals were directly related to traveling, some within MLB want to take that out of the equation.
https://www.si.com/mlb/2020/08/12/baseball-playoffs-world-series-bubble
The Cardinals are supposed to finally resume play on Friday, with a game in Chicago against the White Sox. However, Jon Heyman is reporting that another Cardinals coach has tested positive for COVID-19, so the series start may be pushed to a doubleheader on Saturday.
Cards better get used to playing doubleheaders.
The thread isn’t about “baseball,” it’s about the current season, and the hijack is not. It’s really a completely different conversation that has overwhelmed the thread.