Exactly how bad were the 2010 Mariners? A lesson in the flaws of pythagorean projections.


We all know the Mariners were bad, but exactly how bad were they? If you read Geoff Baker’s blog on the Seattle Times, you’d be led to believe that this team was the worst team that was ever put together. While the offense turned out to be historically bad, the starting pitching was surprising good, at least for the first half of the year.

As we all know, the Mariners finished the year a pathetic 61-101. The also only scored 513 runs, which is the lowest total since the DH was added. Defensively they gave up 698 runs. Using the Pythagorean projection formula, the Mariner’s “should” have been 59-103. This suggests that M’s were even worse than their record (a scary thought indeed), and I expect you have seen this point being made on other sites throughout the offseason.

Before we move on, I need to address one issue. I know this is going to make those who love their sabermetrics cringe, but unfortunately the Pythagorean projection system (pythag) is seriously flawed. Teams that win with good pitching and average hitting always win more than their pythag suggests, while team’s who win with great hitting and average pitching always have less wins then their pythag predicts. So using a team’s pythag to assess if they were better or worse than their record isn’t a viable method.

This is because of the effect of blowouts. Since the Pythagorean projection system only takes into account runs scored and runs allowed over the entire season, losses by more than 3 runs can actually count as 2 losses for the team’s pythag. In the same manner, wins by more than 3 runs can actually count as 2 wins for a team’s pythag. I’ve never seen a team get 2 losses added to their record when only playing 1 game. Proponents of Pythagorean projections claim that by the end of the season it should all even out, but in practice it usually doesn’t a couple teams each season.

Lets use an example to illustrate this point, the 2009 Mariners. This is a team that won 85 games, but had a pythag of just 78 wins. A difference of 1-2 win between the pythag and the actual win total isn’t a big deal. Like all real data, there’s always a little variance. But here we have a difference of 7 wins, which is quite significant. This is because the 2009 Mariners suffered quite a few more blowout losses than they had blowout wins.

Those that swear by Pythagorean projections will say that this means that a team isn’t as good as their record because they’re getting dominated too often. While that might work in other sports, I just don’t buy that argument in Baseball. Roster construction plays too much of a roll here. This was a team with a below average offense, but had good pitching. A team that wins in this way will rarely rack up enough runs to win in blowout fashion, but that doesn’t make them a bad team. The Angel’s won a world series this way. (The Angels also won a lot more than their pythag predicted that year as well)

Conversely, teams that win with a ton of offense and average (or worse) pitching are always on the positive side of the ledger when it comes to their pythag. The Mariners of the 90’s are a great example, as are the Yankees in almost any year. Since this article is already way too long, I wont go into the specifics for these teams. You’ll simply have to take my word for it, or look it up on your own. Now, back to the 2010 Mariners:

If you account for these blowout losses by allowing them to only account for what would be the maximum so they only get credit for 1 loss or 1 win per game, essentially capping the run difference allowed per individual game, (golfers will recognize this from the way handicaps are calculated, capping the strokes per hole) the 2010 Mariners adjusted pythag comes out to be 64-98. While this record is still bad, it’s better than their actual record and suggests that the team wasn’t as bad as their record shows, though only slightly so.(For the record doing this for the 2009 Mariners leads to a pythag of 83-79, much closer to their actual record)

Another way to look at a team is to use their hitting lines, and their opponents hitting lines, and predict how many run “should” have scored. This tends to normalize for pitcher who scatters a lot of base runners over many innings, but doesn’t give up many runs (Miguel Bautista circa 2007) vs. pitchers who dominate mostly but always have 1 bad inning (Freddie Garcia, Doug Fister).

This projection says the M’s should have scored 546 runs, 33 more than they actually did. That should come as no surprise to anyone who watched them strand almost every base runner they had all season. On the pitching side they should have only given up 668 runs, which is 30 less than they actually did. These are some interesting numbers, and I’ll have to get into them in another post. The adjusted pythag using these numbers is a 67-95. Again this shows that they actually weren’t as bad as their record shows.

Baseball prospectus also adjusts these stats based on strength of schedule. Exactly how this is done is a topic for another post. Based on this adjustment, it turns that their pythag record should have been 69-93. This means that the M’s ran into more than their fair share of the opposing aces and didn’t get their share of #4 and #5 starters. (If you disagree with this don’t blame me, these aren’t my stats. I just found them interesting.)

So, what does all this mean? Well, for my money it says that the M’s weren’t quite as bad as we thought. Instead of 101 losses bad they were closer to 95 losses bad. Don’t get me wrong, that’s still awful, but perhaps not as awful as our raw emotions led us to believe as the season ended, and certainly not the 115-118 losses bad that Baker claimed they were.

I’m interested to hear what the rest of you have to say on this subject. Please post your responses in the comments thread below.