keyboard_arrow_uptop

Today, we’re launching the first iteration of our Playoff Odds Report for 2013.

Before we get underway, some notes. PECOTA does not hate your favorite team. PECOTA is a collection of algorithms, written in computer code and run by an unfeeling machine. It cannot hate, or love. It can do only what it is told to do, nothing more or less.

I’m the one that hates your favorite team.

No, but seriously. These odds are based on preliminary depth charts, with teams still having many roster moves to make before the season starts. Thus, they aren’t etched in stone, they’re etched in sand. And these are PECOTA’s estimates of what would happen if the season were played thousands of times. PECOTA is not infallible, and the season is played but once. It is a prediction, not destiny.

Which brings us to the changes we’ve made to the methodology of the Playoff Odds. We recognize that PECOTA is not infallible, which is why we use a Monte Carlo sim with random numbers instead of just ranking teams by  the output of the Depth Charts. But in examining the Playoff Odds Report, we found that the randomness was pretty symmetrical. So what we’ve done is skew the randomness toward the mean a bit more, so that we no longer have playoff percentages above 90 percent before spring training games are even underway. What this also means is that you’ll see a smaller spread of SimWin and SimLoss than what you see in the Depth Charts and other products.

As you might have noticed, we’ve added two new columns to the Playoff Odds Report, just to the left of the one- and seven-day deltas: Adjusted Playoff Percentage and World Series Win Percentage.

Adjusted Playoff Percentage, which we abbreviate as Playoff Pct (Adj), is scaled more like Playoff Percentage used to be when there was only one Wild Card in each league. It’s the chance that a team will make it to the Division Series round (in other words, the odds that it will win the division plus the odds that it will win the Wild Card and win the play-in game), and it’s figured as so:

Div Pct + (WC Pct * AdjWinPct)

To determine the adjusted WinPct, we use the odds ratio method to figure out the odds of a team winning a matchup against the other Wild Card team. To calculate the winning percentage of the other Wild Card team, we take an average of all other teams in that league, weighted by their odds of winning the Wild Card. This way, the better your team is, the weaker your expected Wild Card opponent.

(We have to apply a slight adjustment to the odds ratio method to get everything to work out correctly; instead of assuming that the league average is .500, we figure out what the league average would be if we removed the team in question. Otherwise we end up overestimating the Wild Card advancement rate.)

The benefit of the Adjusted Playoff Percentage is that it allows us to compare teams across divisions and leagues more readily because it places a greater emphasis on a team’s chances of winning the division (which in the current playoff format is much more valuable than winning the Wild Card). If you have two teams with otherwise equal Playoff Percentages, the team that is more likely to win its division instead of a Wild Card berth is in a better situation, and our Adjusted Playoff Percentage reflects that.

We also have our WS Win Pct, which tells us the odds of a team winning the World Series given its Adj. Playoff Pct and its expected win percentage. What we first do is come up with an expected win percentage for the Division Series round, using the team’s expected win percentage and the odds ratio method outlined above. Then we use the binomial probability mass function to figure out the odds of winning a five-game series. We multiply the Adjusted Playoff Percentage by that to figure the odds of advancing to the Championship Series round, and then start the process all over again (except we use the probability mass function for a seven-game series this time). Then we repeat the process one more time to get the odds of winning the World Series.

This World Series percentage is useful, again, because it helps put a team’s playoff chances in perspective. Because we use a league-quality adjustment, an AL team with the same Adjusted Playoff Percentage is favored in the World Series Win Percentage.

(The astute among you may notice that these odds will be uniformly lower than the odds of winning the World Series given out by sportsbooks. This is not because we hate your team, but because sportsbooks include something called the vigorish. When adding up the implied probabilities from one sportsbook, the total came out to nearly 140 percent. Because of a pernicious reality-based bias we have, our World Series odds instead add up to 100 percent.)

You can view the Playoff Odds Report through this link and also from the dropdown list accessible by mousing over the "Standings" button on the navbar at the top of any BP page. We will continue to update the odds up until the start of the season (as well as daily during the season), as the Depth Charts continue to be updated to reflect roster moves and the like. But for now, we hope that you find these useful, and that the new additions help to further illuminate the playoff races in the early going.

You need to be logged in to comment. Login or Subscribe
apilgrim
2/15
It is pretty amazing the Angels won the World Series in about 1 of every 10 simulations.
Lindemann
2/15
I think I'm actually going to miss seeing that some team has a 96 percent chance to go to the playoffs on the first day of the season, then watching that steadily decline as they played below expectations. (Or the experience like last year of watching the Nationals' playoff odds climb.) It's like watching conventional wisdom be proved wrong by reality in real time.
chrisw
2/15
Are these expectancies accounting for the actual MLB schedule for each team or is it based on expectations in a vacuum?
cwyers
2/15
We use the actual MLB schedule.
TGisriel
2/15
The projections indicate that the Orioles are the 3rd worst team in the American League, winning more that only the Astros and the Twins. As I recall, last year the PECOTA projections showed the Orioles to be the worst team in the American League. We know how that turned out. I'll take the over for the Orioles.
gweedoh565
2/15
Just remember that it's Colin that hates the Orioles, not PECOTA.
thegeneral13
2/15
You could argue they are more talented than teams projected to be within spitting distance in the win column, like KC, CHI, CLE, and SEA, but divisional strength is a headwind and there's very little difference in the forecasts for those teams anyway. So if you want to be PO'd about being projected last within that (effectively equal) cluster, then fine. But I can't see an argument to project them ahead of anyone else in the AL that is based on talent and not what happened last year.
mshopoff
2/15
0.4%? So you're saying there's a chance!
MFBabyFeets
2/15
So the O's are the 3rd worst team in in the AL and The Angel's are going to win the division with the A's finishing 3rd. What are the odds on Scioscia getting fired before the All Star ? I think there's good money to be had there. The Angels rotation is still weak and you shouldn't expect 30 starts from Hanson and Weaver. The Angel's followed last season's offseason formula signing the big sexy hitter and getting an above average SP but not a great one. I am expecting similar results to last years team... underachievement.
Wozzyck
2/16
PECOTA seems to direct most of its hatred towards any pitcher that happened to pitch in AA in 2010. You might want to check your database for errors as it seems to be a pervasive problem (see the player cards for Felix Doubront, Julio Teheran, Zach Britton, Martin Perez, etc.).
mcquown
2/16
This is the good thing about stats, instead of going by perceptions. I haven't run a query to check them all (so I'm just responding with perceptions, I realize), but just scanning the FIP leader list for 50+ IP pitchers in AA in 2010 (http://www.baseballprospectus.com/sortable/index.php?cid=1380439), I see quite a few names who got fairly rosy projections for 2013, including: Brandon Beachy Jake McGee Cory Luebke Mike Minor
Wozzyck
2/16
Yes, but look at the stats from those players on their player cards from the year/level in question. All negative WARPs, low beyond reason. Is this contributing to their PECOTA projections? Well, I have no idea as it's proprietary info, but the four you mentioned at least have some nontrivial amount of MLB time in the bank prior to 2012 which perhaps mitigates the problem for them in the eyes of PECOTA. (Britton does too, but I'm willing to concede that a crummy projection is warranted in that case.)
mcquown
2/16
Mea culpa, I misunderstood your point.
bornyank1
2/16
Thanks for pointing this out, we'll take a look.
mcquown
2/16
We had some bad data that was generated during testing on our "research" server leak into production, fouling 3 of the various year-levels: 2010 Rookie League 2010 AA 2012 Rookie League Thanks for pointing it out. These stats were quickly restored to their previously rational values quickly on the "research" server, but just now on production (and additional checks were added to the process to avoid a recurrence). None of them were used in PECOTA computations, though I fully understand your concerns now.
bdbelknap
2/16
Write all you want. Everyone knows that PECOTA is a malevolent spirit. Nate Silver cut his index finger while cooking dinner. As he typed the initial algorithms of PECOTA, he bleed on his computer. This gave life to a vicious monster that hates everyone's favorite teams and seeks to crush all spring hopes.
rweiler
2/20
Maybe so, but it doesn't hate the Dodgers. Amazing what a $200m+ payroll can do to sooth a vicious monster.
edman8585
2/17
Colin, if you give them a better than 1/7 shot at winning the World Series, you don't hate my team.