Pointless Analysis Time: Wally the Win Expectancy Worm
Posted by MetaNarrativesFTL
Almost every day of the year, an international sports competition is underway, and the results are churned out day by day. Those following games online can usually use the well regarded newswire "Communicating Unfailingly The Official Fixtures' Final Scores" ("CUTOFFS"). Today's CUTOFFS report included the football matches currently taking place in Quintessence of Dust for the Introduction of Fire, the Round of 16 stages of something called the World Bowl (?) - and opening day from the World Baseball Classic.
Around the world, people have been quick to notice that this time, the CUTOFFS data supplier - based in The Babbage Islands - have been providing the scores in a new format, including full line scores so we can see innings-by-innings match progress. This has caused some consternation, the fiercely conservative Rocket Report bridling at the change, the upstart Tarrentum Falcons Sporting News embracing it. It doesn't really make that big of a difference in the scheme of things: some press sources, such as NordicNation and QSPN almost always provide line scores, others, such as Cassadaigua's official player blogs, generally eschew them.
It's our understanding that CUTOFFS will continue publishing line scores throughout the WBC15, though they might disappear again for WBC16. We also understand that, because CUTOFFS is a central clearinghouse of information, some wires might get crossed on the innings-by-innings results: although we all know that CUTOFFS results are rigorously fact-checked, the individual line scores might vary from actual game results. So be it.
But while we do have the line score data available, it gives an opportunity for a new edition of Pointless Analysis Time. This post has mostly been inspired by a Zwangzugian blogger, who made a very interesting post about a domestic baseball game that helped bring the concept of "Win Expectancy" to my attention. I've been reading up on it, and it's on this basis that I'd like to pay my dues to the Society of Zwangzug Baseball Research, and announce that I have an idea for a mascot our team (remember, the competition to pick the name is still open!): Wally the Win Expectancy Worm.
Wally relaxing after a busy day against Brewdomia
Win Expectancy is a way of determining how likely a team is to win a game based on the game situation. It takes into account runs, outs, and bases. Note that this is based on in-game factors. (I recently read a Rocket Report that guessed that they'd "win 19 of the 20 games against [Virabia]". This isn't the sort of probability Win Expectancy discusses: the issue is whether a team has a 4 run advantage after 3 innings, not whether a team has two championships to its name.)
Win Expectancy can be calculated to take into account every play in a baseball game. On defence, getting outs and stopping runners advancing, and on offence, getting on base, moving into scoring positions, scoring runs and avoiding outs, increase the probability of your team winning. Below, we discuss the WE data from our home game against Brewdomia, but I thought it might be interesting to start with a much simpler look at the other games on show today, and just see how their win expectancy changed on a by-innings basis. (Note: the aforementioned possibility of inaccuracy in the CUTOFFS line scores means the following analysis is not ironclad.)
All apologies for the slightly crude graph designs, but they illustrate some interesting points. For example: Barnettsville's ninth inning rally captured a great comeback win over the Zebras. But how great?
By the ninth innings, Barnettsville's win expectancy had sunk below 10%. Of course, we all remember the story of WBC14, when the entire Quodite team was briefly suspended after gambling heavily that Stars first baseman Hank Everly would end his career with a walk off grand slam (their legal defence that "it was pretty obvious that was going to happen" thankfully stood up in court). But in truth late rallies are very rare: a team with a ninth inning lead has a very high probability of winning, making the Blue Jays' upset win all the more surprising.
Until that remarkable fightback, the win expectancy trended strongly one way. This is often the case. Teams that score the first and second runs of the game have a high probability of winning. But sometimes games seesaw a little more. A real topsy turvy one that put Wally through some calesthenic twists and turns was fought out between Undeadzombiee and Steroga:
This kind of graph is much more unusual. Steroga (who it seems to us are one of the unluckiest nations in international sport) were trending toward a win, but the impact of runs in later innings is illustrated by the huge spikes and troughs as the lead changed hands.
The shape of the graphs depends on the data set you use. You might notice that the one we use includes a slight home advantage, which is why Wally's tail starts slightly above the 50% mark. We've expressed the result in terms of the WE of the home team. Sometimes, it is given as the WE of the +50% team: instead of saying that Steroga's WE was 30.0% (or .300) at the beginning of the eighth inning, you could say that Undeadzombiee's was 69.9%.
The graphs are very simplistic here. A more complex WE chart takes into account each play, and this can be used to show the value of an individual player's contribution. For example, Johnny Stryker hit a 3-RBI home run in the first inning of the Nordics' visit to the hosts (how the Quodite players resisted the urge to bet on this, I don't know). You might be tempted to brush this off: Johnny Stryker hits a lot of home runs, is one really so special? But by our calculation, that home run improved the Nordics' WE from ~56-58% with two runners on, no outs, to almost 75%. In other words, he increased the probability of his team winning by almost 20%! By contrast, Kevin Monroe's ninth inning home run had virtually no impact on the WE - less than 0.1%.
With that in mind, let's turn to our game, and look at the most important plays. We have prepared a much more extensive data set on our own game, but the Undeadzombiee-Steroga clash really tired Wally out, so we're letting him have a rest and will have to do without graphic representation. I've expressed the values in hard terms; this presents a bias towards early innings events. Expressing the changes as proportions of the total WE would probably be fairer, but let's keep things simple for now.
1 & 2. Natalia drives in Amanda, twice (WE changes: +15.5%, +12.5%)
Natalia Kivistö had a fantastic debut: 2 for 3 with a hit by pitch and a stolen base, she hit the first home run of the WBC15 campaign. But that wasn't her biggest play. In terms of win expectancy, her fourth inning RBI single boosted the Quodite WE from 73.3% to 88.8%. That play was a lot bigger than when Masami scored from first on a single in the same inning, and it was bigger than her first inning home run.
Naturally, one might be inclined to distrust a statistic that suggests a hit worth 1 base and 1 RBI was better than a hit worth 4 bases and 2 RBI, but the key thing is that in came later in the game. Quintessence of Dust's lead was cut to 1 by Brewdomia's RBI double, sacrifice fly combo in the fourth, but Natalia's single restored a 3 run lead. Additionally, it came with 2 outs - her home run was the second at bat for the home team. Because a runner on base with 2 outs is likely to be left on, a 2-out RBI is very valuable.
Of course, Win Expectancy be damned, there's no escaping which moment the fans at The Frosty Dome cheered the loudest for.
3. Momoko's web gem at second (WE change: +9.7%)
It's really important to stop rallies in baseball. No football dream would dream of having a defender who only played 5 minutes per game, because their opponent will have many attacks over 90 minutes. In baseball, most of the "attacks" come at once, and the majority of runs are often scored in a single inning, explaining why - unlike the 5 minute defender, the short reliever does exist, and is often in high demand. In the fourth inning, after trailing by 3, Brewdomia began to put runners on base and score runs. Their batter drew a walk, the next hit a double, the next a sacrifice fly. They were within 1 run of the home team.
Then Momoko Koumura justified her Silver Glove at second base in the FABL Free League with a dazzling piece of individual fielding. Shorter than many middle infielders but with a better arm than a number of third basepersons, Momoko has always relied on her footwork to help her out, and it was only a brilliant waltz she danced in the infield dirt to deny a hit that would have advanced a runner into scoring position - and ended up silencing that fourth inning rally, during which they had increased their WE from 14.2% to 38.8%.
Because fielding is a notoriously difficult part of the game to quantify, Win Expectancy should not be completely relied on here. Had it been a routine groundout, it still would have been a big play: the WE change does not reflect the probability of what happened (how many fielders in the world could have scooped and dived like that?) but the impact of what happened on the game situation. Of course, good fielders will probably make a lot of such plays over time.
4. Leadoff hitter caught stealing (WE change: +9.6%)
Let's be clear: there is nothing wrong with stealing bases. So long as you pull it off. (Actually, there occasionally is something wrong, as having a runner on first against a left handed pull hitter makes the second baseperson's role more difficult - but this is negligible.) Everyone agrees that getting caught stealing is a blow, and that the fourth biggest play of the game was a pickoff is no surprise. After singling to short, the Brewdomian leadoff had given their team a great start (in fact, it was the only time in the entire game their win expectancy was above 50%).
Had he stolen effectively, he would have improved his team's WE by about 2.5%. But a sharp pickoff from Hannes Huuskonen, justifying and then some his selection ahead of Radosav, slayed him at second, and that his set his team back almost 10 WE points. That differential illustrates how much more valuable outs are than bases - and why one can be sceptical of the true value of base stealing to a team's overall probability of winning.
5 & 7. Brewdomia's sacrifice fly, RBI double (WE changes: -9.3%, -8.3%)
The biggest play for Brewdomia was their fourth innings sac fly. With a runner on third from the RBI double, their batter hit a deep fly to centre, and Masami Tamagotchi could only make the catch, not the throw. That brought the lead down to 1, illustrating how volatile runs in the early and middle innings can be on the overall WE of a team.
6. Fielding error at shortstop (WE change: +8.4%)
If one great piece of fielding can sharply change a game's profile, another dreadful one certainly can too. Maybe it took a bobble in the dirt - but this isn't an infield that's been played on all Classic. New sand, freshly raked - you could putt a golf shot on it. Well, you could - Brewdomia's shortstop certainly couldn't, as he appeared to possess all the coordination of a hepped-up lab monkey as he let the runner to begin the comeback in the bottom of the fourth.
8. Django's sixth strikeout (WE change: +7.9%)
Brewdomia had another good chance at a rally in the sixth, though they laboured to get their by amassing outs. Still, back to back singles put runners on the corners, and a hit from their batter would have driven in at least 1. But he couldn't get anything on a sinking fastball from Scheinhardt, and two more followed - and then a knee-buckler of a changeup.
Django had a very Django-like game: faintly unimpressive, but did the job. He struck out six and walked four, and gave up nine hits for two earned runs. With the win under his belt he has plenty of time to prepare for what now looks like being a tough encounter against Barnettsville.
9. Masami's outfield double play (WE change: +7.5%)
We always knew she could hit, and we always knew she had a hell of an arm. Perhaps we should be embarrassed for not putting two and two together sooner. In any case, any doubts about Masami Tamagotchi's transfer to the outfield were eradicated as she roped two hits and drew a walk, stole two bases and scored a run from first on a single, and - most crucially - had two outfield assists. Her double play was a highlights reel heartstopper. But as mentioned, such plays can be showy irrelevancies - not this one. It stopped Brewdomia's seventh inning rally and moved the team's WE above 98%.
10. Emily's solo home run (WE change: +7.1%)
Way down at 10 comes the home run, though it was a good 'un, travelling over 400 feet. That's not the issue. It's that it was a hit with a win expectancy already nearing 90%: it had a lesser impact on the game situation than the above plays. Surprising, perhaps, though such an assessment had no bearing on Sanderson's joyous celebrations as Quintessence of Dust's most consistent international hitter rounded the bases in style.