Does your win expectancy change if the game is still scoreless?
Jeff Sullivan had a variation of this question for me. I found it intriguing enough to take a quick look. This chart is limited to games where the game was tied entering any half inning.
Obviously, the first two half-innings, the game is scoreless entering those half innings. And we don’t see much separation until we get to the start of the 11th half inning (i.e, top of the 6th). From that point onwards, the chance of home team winning increases in a scoreless game.
Why would this be? Almost certainly selection bias. Those games feature a low run environment, either because pitcher parks are disproportionately represented, or great pitchers are disproportionately represented.
Basically, in a small-ball setting, the chance of winning increases for the home team of a tied game.
But, is that ALL it could be? Could maybe managers and players play differently knowing the game is scoreless? Someone else can take it from here.
I would think a big factor in a tied game is the strength of each team’s bullpen because the farther into the game that it is tied the better the team with a better bullpens odds are
Maybe the selection bias is umpires.
The scoreless games probably select for umpires who are more likely to call strikes, and have more impact on the game in general. And isn’t there evidence that umpires tend to call the zone a little friendlier to the home team?
So if 0-0 games are, on average, selected where the umpire already has a higher than normal impact, maybe this amplifies the home team bias the umpires already have leading to an increase in the home team’s chances of winning.
Does this make any sense? I’m not sure why the effect would only start to exhibit itself in the 6th inning, but I believe the answer probably lies in how the effects which already give the home team better odds are altered by a scoreless game.
In scoreless games, home team has three extra outs versus visitors since if they scored, it is no longer scoreless for visitors. Closer to ninth, this is more pronounced.
Home team also has more information during their at bats than the visitors. The closer to the end of the game, the more this helps the home team. Visitors score 2 in top of ninth. Home team knows sacrifice bunts are out. However, with scoreless game, home team willing to sacrifice an out to get one run.
Visiting teams also may not play their best relievers unless they are ahead late in game (though seeing this less).
All of your statements are true, but they don’t explain why there is a higher win expectancy for the home team in a scoreless tie game (0-0) game compared to a non-scoreless tie game (1-1, 2-2, etc).
“In scoreless games, home team has three extra outs versus visitors since if they scored, it is no longer scoreless for visitors. Closer to ninth, this is more pronounced.”
This explains why the peaks are progressively higher for both lines as it gets later in the game, but not why the difference is more pronounced in a scoreless game.
“Home team also has more information during their at bats than the visitors. The closer to the end of the game, the more this helps the home team. Visitors score 2 in top of ninth. Home team knows sacrifice bunts are out. However, with scoreless game, home team willing to sacrifice an out to get one run.”
Both charts are for scenarios entering a scoreless half inning game. If the visiting team scores 2 runs, why would the home team’s strategy change if the score is 1-3 or 2-4 vs 0-2?
“Visiting teams also may not play their best relievers unless they are ahead late in game (though seeing this less).”
Again, this might help explain why the peaks are progressively higher for both lines as it gets later in the game, but not why the difference is more pronounced in a scoreless game.
Excellent, thank you for saying with brevity and clarity what would have taken me twice as long to say.
Just reread article. My head must have been elsewhere when I first read it.
Edit. Im an idiot today.
Our minds don’t always work at 100%. But you were interested enough to comment, and that’s good enough.
Terrific reply!
We should all expect the home team to have a large advantage in the bottom of the ninth, or any later inning, of a scoreless game.
So, if a large fraction of games scoreless at the top ninth are still scoreless going into the bottom of the ninth, then this would mean the home team should have an advantage going into the top of the ninth.
And, if a large fraction of games scoreless at the top of the eighth are still scoreless going into the bottom of the ninth, then this would mean…
There is certainly more than this, but I bet one thing this graph says is that a good fraction of games scoreless at the top of the fifth are still scoreless in the ninth.
My guess would be that since some of the non-scoreless tie games include games where bullpens have been used up (you are not differentiating between 1-1 vs 7-7), that the advantage of being the home team entering the bottom half of an inning is greater, since you can face a worse pitcher one more time compared to a scoreless game where both teams still have their best relievers. This also has an impact on the (less intuitive) advantage that the home team has entering the top half of the inning in a non-scoreless tie game, since many of those games will become a tie game entering the bottom half of the inning, although I’m not sure about this second part. Maybe the home team manager is more willing to use up more good relievers in a high-scoring game.
In other words, maybe there isn’t much of a difference between 0-0 and 1-1. However, maybe there is a big difference between 0-0 and 8-8, even though 8-8 would occur far less frequently than 1-1.
This is my favorite hypothesis so far. Would love to see these charts for 1-1, 2-2, etc.
A potential factor with bullpens is the road team trying to save their best reliever for a save situation, never using him in a tie. So as long as the game is still tied the home team effectively faces a rung down on the bullpen ladder than what the road team is facing.
e.g. Baltimore bringing in Ubaldo before Britton in the AL wild card game
I’m not sure why this is downvoted. I think this gets at a big reason.
We know road teams tend to use relievers sub-optimally in tied, extra-inning games. Britton’s non-game was perhaps a watershed moment for this. My thinking is that if the score is tied, some of these good relievers are more likely to have already gotten into the game for the road team, probably because the starter and relievers didn’t pitch well and got pulled early, leaving the manager fewer options.
In other words, if you look at extra-inning games and divided them into “road manager will use his relievers optimally” and the inverse, the two graphs may look a lot like the ‘([1-9])-\1’ and the ‘0-0’ graphs.
Also FYI if you want to type a regular expression into one of these comments, you have to escape backslashes.
Also my regex misses double-digit scores.
([1-9][0-9]*)-\1
Also I realize I made a critical typo in my comment.
This should say if the game is tied at a high score, so implicitly not a scoreless tie. It’s also possible that the selection bias select more non-scoreless ties for higher run environment, and that means more relievers used, which means more tired relievers, which means the suboptimal relief usage is flattened because usage is driven more by just who is available.
Just to be clear, we’re comparing games that are 0-0 to those that are 1-1, 2-2, 3-3, 4-4, etc.
I wonder if the basic concept of momentum could have anything to do with this. In 0-0 games (or lower scoring games in general), there are less win expectancy swings (momentum shifts). Without this momentum, perhaps home field advantage becomes a slightly bigger factor. When the away team takes a lead, their win expecatancy jumps up by more than the difference from before from home field advantage.
There could also be a league/team difference. My perception could be way off, but I’ve I noticed NL teams just are better at turning a leadoff walk into a run in the bottom of the ninth. I would think they are also more likely to be in a 0-0 game.
Could it be about pinch hitting? E.G, Big Bopper is more likely to have the day (mostly) off and so be available to pinch hit in a 0-0 game. This seems unlikely to cause such a big swing.
One final idea umps are don’t like to end a game by calling a strike. Could that be more so in a that was 0-0 starting the 9th than a 4-4 game at the start of the ninth?
In a scoreless game the win expectancy for either team should asymptote towards something close to 50%. The score will vary more from the true talent of either team. In other words, the sample size will be smaller and so the results will have a wider variance. This doesn’t explain the home/away split, but is a separate point.
I don’t have an answer, but at least we can start by brainstorming things that are different about 0-0 games than, say, 6-6 games. Most have been mentioned already, but in a tied game with a higher score, it’s more likely (or in some cases, guaranteed) that:
– more relievers have been used
– more pinch hitters have been used (especially in the NL)
– the lead has changed more often
– some hitters have had good games
– some pitchers have had poor games
– some fielders have had poor games
– the strike zone is bigger
– the score became tied later in the game
I’d also love to see the same graphs but instead of 0-0 vs. all other tied scores, separate plots for 0-0, 1-1, 2-2, etc.
Also the “momentum” argument – how about something like “home team’s observed chance of winning in the top of the 9th inning in tied games, based on the number of lead changes in the game?” – or based on how late in the game the score became tied?
A visiting team might want to wrap up too-long-extra inning games as they go to the hotel not their home. They’d feel it’s better to be prepared to coming games rather than cling to this one game.
What sample size are we looking at here? It’s interesting that in games entering the bottom of the third tied but not scoreless, the home team hs as won 60% of the time, which is the same win pct. as games tied, non-scoreless games entering the bottom of the sixth.
Go for the win on the road and a tie at home
This has been conventional wisdom in baseball for a long time and a did some quick research last night that indicates that that idea may still be affecting strategic decisions that cause the effect observed here. Of course no team actual tries for a tie. the interpretation of the quote above is that it is wiser for an away team in late innings of close games where they are tied or behind to try a strategy that would result in them getting a lead rather than a safer strategy that would result in them gaining or remaining in a tie. The home team is more likely to be conservative if they are in a late inning tie game. Some of the riskier strategies that an away team might employ would use up players and options that a home team acting conservatively would save for the future. In the case of a game that remains a zero zero tie game the score indicates that the away teams riskier strategies have not worked and the home team by being more conservative and preserving its options have increased their probability of ultimately winning.
In non zero tied games either team could have come from behind to tie the game and used up resources to do so. The visiting is slightly more likely to have come from behind because of the home field advantage they are more likely to be behind at the start of every inning. But the visiting team is much more likely to have used up its resources using risky strategies in a zero zero tie game.
I think it could be that teams with higher expected total runs per game (Rockies, D-Backs, White Sox, etc.) have been below average teams in the recent past. So a high-scoring tie would be disproportionately likely to involve one of them as the home team.
I’d like to see the avg. win pct. of each group.