31 July 2015

So You're Saying There's A Chance?

Yesterday, Jon Shepherd claimed:
we can be really optimistic and dream of a Yankees collapse, but we know those things rarely happen.
That's easy to say, but I was wondering whether this is in fact the case so I decided to look at how division leaders at the end of July perform for the rest of the season.

From 1998-2014, the team leading its division at the end of July won on average 93.7 games, had nearly a 75% of winning at least 90 games, had an 84% of making it to the playoffs (via 2015 rules) and a 71.4% chance of winning its division. On first glance (and second), this does appear to indicate that it’s highly unlikely that the Yankees are going to collapse.

The problem with just using this method is that it doesn’t take into account whether a division leader is only five games over .500 or fewer (less than a 50% chance of winning their division) or if a division leader is sixteen games over .500 or more (96.6% chance of making it to the playoffs via 2015 rules).  The table below shows how division leaders in this situation usually perform.

Good news, according to this table, the Yankees’ chance of winning the division is only 67.7%. Things are looking up because the Yankees aren’t quite as good as the average division leader. This next table shows how teams have historically done when they have between a 4-8 game lead and are 10 to 15 games above .500.

Ten out of fourteen of these clubs ended up winning their division and that includes teams with only a 4 game lead. All in all, it seems reasonable to argue that the Yankees have a 70% chance of winning the division based solely on history. But a 70% chance of success means that the Yankees have a 30% chance of failure.  What are the chances that the Orioles will actually end up making it to the playoffs based solely on historical data?

Based on how the data looked, I split teams into five groups. These groups are teams that ended July below .500, those between 0 to 5 games above .500, 6 to 9 games above .500, 10 to 15 games above .500 and 16 or more games above .500. This table shows how they’ve performed.

Teams that are below .500 at the end of July have pretty much no chance at being a playoff team even under the current rules. Only 5 teams out of 236 would have made it to the playoffs under the current rules and only 1 of these teams won 90 games. Teams that are under .500 at the end of July should be selling and should realize their situation is hopeless.

Teams in the next category have a 19.2% chance to make it to the playoffs (9% division winners and 10.3% as a wild card) and are underdogs but definitely still in the hunt. These teams include the 2014 Kansas City Royals that went on a huge run and made it all the way to the World Series. More to the point, given that so far there are only four teams that are six games above .500 or more in the AL, it seems that at least one team in this grouping will make it to the playoffs from the AL this year.

Teams in the third category have a 61% chance of making it to the playoffs, primarily as a wild card team. Teams in the fourth category have a slightly higher chance of making it to the playoffs at 70% but are far more likely to end up winning their division. Teams in both of these categories have an excellent chance of making it to the playoffs and should be buyers.

Teams in the fifth category are the elite. Nearly all of them make it to the playoffs and win over 90 games. A whopping 71% of these teams end up winning their division while another 26% make it to the wildcard game. The two teams that wouldn’t have made it to the playoffs under the current set of rules were the 2001 Chicago Cubs and the 2006 Boston Red Sox.

All of this means that on the one hand it’s unlikely that the Yankees will collapse or that the Orioles will make it to the playoffs. A 30% chance is still reasonable but on the whole one would rather be in the Yankees situation.

On the other hand, it does mean that the Yankees have a better chance of collapsing than of the Orioles making it to the playoffs. If it isn't rational to hope that the Yankees will collapse then it doesn't make sense to think the Orioles have a chance. The problem with that reasoning is that someone is going to make it that doesn't have a good shot at the moment and we just don't know who. It's too late in the game to give up now.

All in all, the Orioles’ situation isn’t looking so good. But so far, I’d rather have the Orioles’ record than the Red Sox’ record. Someone in the AL is going to make it to the playoffs despite their bad odds and it may just be the Orioles. The only thing we can do is just wait and hope even if it is a bit irrational.


Jon Shepherd said...

Given underlying performance and other teams in contention, Orioles have about a 5% chance to win division based on projection modeling.

Matt Perez said...

The "problem" with using projection modeling is that people may not necessarily buy the projections especially if they either unable or unwilling to understand the methodology. Instead, they'll focus on outliers in an attempt to discredit the model or else they'll say that these projections normally work but they're inaccurate for their team for reasons a, b and c.

Historical data is much less accurate (because it doesn't differentiate between good and bad teams with the same record or situations) but is harder to argue with. You can't tell me that what happened didn't actually happen.

Jon Shepherd said...

You are modeling both ways. One way is more uncertain than the other.

Mike said...

To me, the Yankees seem to be performing well above their individual talent level and have gotten lucky to be where they're currently at. This is based on nothing but my own observation. According to the peripherals, where should the Yankees be, record-wise?

Matt Perez said...

This method ignores teams talent level and focuses solely on record. This is because everyone thinks that their team is usually better than what their record suggests while their opponents are worse than their records suggest.

Teams in the Yankees' situation win the division 70% of the time indicating that they have a good shot at winning the division regardless of whether they're lucky or not. It also indicates that they haven't clinched it yet.

Scott M said...

Couldn't you factor in opponents' records through 7/31 to adjust for strength of schedule? That might help account for teams playing above or below their talent.

Jon Shepherd said...

Yes, that and more is incorporated in the odds that Baseball Prospectus runs.

Matt Perez said...

Absolutely. But does opponents' record actually measure skill level? Maybe a team got lucky so perhaps I should use something like run differential or base runs? Plus how do you deal with teams that have added or subtracted players? How do you deal with injuries?

That's why some models' attempt to forecast player performance and use that to determine the talent of clubs, their previous opponents and their future opponents.
These models are more precise than mine but they also require the use of projections.

The reason why I think this model has some value is because it doesn't require the use of projections. It states clearly that teams with such and such record historically have had a record good enough to make the playoffs "x" amount of times. It doesn't attempt to judge whether an opposing teams' future opponents are good or bad or whether a team has added talent or what not.

I think this model makes it harder to use wishful thinking to determine playoff odds. It's possible to argue that Fangraphs underestimates the Orioles' chances of making it to the playoffs because of whatever factor. It's impossible to argue that a team has such-and-such record and that those teams have historically made it to the playoffs x% of the time. You can argue that your team will beat the odds for such-and-such reason but at least you have to recognize them.

Given that the best model in the world is useless if it can't be communicated, I feel like the simplicity of it has value.