Nationals Baseball: Statistical Meandering and Bad Stats

Thursday, October 13, 2011

Statistical Meandering and Bad Stats

While watching the games last night I was informed of two things.   The first was the advantage the team winning game 3 gets, after splitting the first two games.   I'm not exactly sure, but I believe they said that 78% of such teams went on to win the series.  Ok, fine.  But is that a tangible advantage or a simple mathematical one? And by that I mean - what would the odds be if wins were just randomly assigned?

There are 10 possible combinations remaining for the games involved.  Making two big assumptions, equal probability of winning and no home field advantage, the odds work out that the Cardinals would win this sereis 68.75% of the time if each of the future games was decided by a coin flip. (25% chance they win in 5, 25% chance they win in 6, 18.75% chance they win in 7).  That is noticeably smaller than 78% so that would suggest there would be a psychological advantage in winning game 3.

Of course that's way too simple. Home field advantage should be factored in.  Unfortunately, that involves going back and looking at each of these series individually.  In the interest of time, I'm not going to do that.  That other assumption though, I can make a run at that. If a team has won 2 of 3 games, it makes more sense to believe that they are the better team.  If that's the case, they would win more than 50% of the games against their opponent.  (Note I'm not saying the Cardinals are better than the Brewers.  I'm saying if we looked at all these series that the winner of game 3 would be objectively better than the loser more often than not).  What does that mean for the chance of winning?

If we simply bump up the percentage chance of winning an individual game to 52%, the odds of winning the series gets bumped up to 71.7%.  Is 52% high?  I don't think so.  That's like saying if these two teams played a 162 game season solely against each other, the better team would win 84 games. To get to a 78% advantage, though, you are looking at a 57% win chance. That's in the better team winning 92 games area.  That does seem high.  These teams don't win much more than that against a mix of all sorts of teams. To win that many solely playing against a good squad seems unlikely. 

There are other things to be considered, too, notably how many such series they are actually looking at and the strength of the starters due to pitch. This isn't anything more than playing with numbers, though.  We're sitting at a chance of winning the series in the low 70s right now.  Off the top of my head, I'd say I'm still not convinced that that 78% is anything more than a blip of a few percentage points caused by small sample size.


The bad stat that I was informed of was made in an attempt to say how unlikely it was that the Cardinal bullpen would hold the lead.  They noted that only YOUR Washington Nationals had more blown saves.   That didn't seem right to me so I looked it up.  It's true.  But you'll also see on that link why that's a bad stat.  The Nats had more save opportunities than all but one other team.  Their save percentage wasn't great but it wasn't near the worst in the league.  Same for the Cardinals.

If you want to evaluate the relief core fairly, well there's a lot of directions you can go. ERA (Cardinals 11th, Nats 4th),  xFIP (7th and 8th),  looking individually at the players involved and their applicable situational stats :
Salas: 2.28 ERA, 3.62 xFIP 
Lynn : 2.22 ERA, 2.21 xFIP
Rzepczynski : held lefties to a .163 / .256 /.221 line with 37 Ks in 118 at bats
Motte : 2.25 ERA, 3.39 xFIP.

Looking at this it's not that suprising the Brewers didn't score again.  You would have expected baserunners, and maybe 1 run was more likely than zero, but the story here isn't "How is this flawed bullpen doing this?",  it's "smart managing of an average bullpen".   But never let facts get in the way of a good narrative.


JDBrew said...

Yeah, that stat is crap. But the world is full of bad stats. Being a commentator would be difficult, it's hard to talk about a game for 3 hours without launching obsceneties at boneheaded plays. And also, it seems like everyone wants to pick on the Nationals. Anyone who pays attention would acknowledge that the Nationals bullpen is not that bad. In fact it's one of the strengths of the team. But what can you do? Yogi Berra said once "50 percent of the lies they tell about me aren't true."

michael k said...

You hear stats like that a lot. "The team that wins game X wins y% of the time", or "the team that scores more runs than the other team through 8.2 innings usually wins the game." Really, a lot of it has to be hogwash. Of course the team that performed better in one part of the game/series is more likely to win. The same is likely to occur even if the game was an exhibition, because over 50% of the time the better team wins that game.

What's important, as you point out, is whether the % boost goes BEYOND both [the effect on the series of a win in general + what the win tells you about the team being better overall]. In other words, if we factor in the fact that a team up 2-1 is more likely to win in random chance, and the fact that a team that wins ANY game is most likely the better team, and we're still left with statistical significance, then the stat is significant.

You attempt to do this in your post, but I offer a different method. We see in other sports that certain games are mathematically more important - in hockey, it's game 2, and in basketball it's game 3 (apparently, I don't follow but a friend tells me this). This is determined by simply calculating the percentage a team that won each game in the series has won the whole series.

This baseball stat is a little different, because it's assuming a two-game split. So, perhaps looking at series were teams are up 2-1, and then looking at series where they were up 2-1 after a two game split, and seeing if there's a difference.

This would require some research, and I'm not about to do it so I can't expect the same from you. But I'm with you in thinking this stat is baloney. Are we to believe that a series that's 2-1 after a 1-1 split is MORE likely to favor the leading team than are series where they got off to a 2-0 lead and then lost a game? Seems like it's more likely to be the other way around.

Another one I'm interested in is the "the team that scores first wins more often" claim you hear so much. Of course, it falls prey to the [neutral effect of a lead on the game + the better team is more likely to score first, anyway] problem above. But I've heard that teams get jittery at the plate when they are losing, especially underdog teams, so I wonder if there's any significance there at all. Of course, you'd have to control for A LOT.

Sorry the post is so long...

John O'Connor said...

I had the same thought as Michael K - is the percentage the same when one team wins the first two games and the other team wins Game 3 (like the Tigers-Rangers series). Is it winning the "pivotal" Game 3, or the inherent value of being up a game in a short series combined with the likelihood that the team winning two of the first three is just better.