2019 Playoff Odds
Mar 6, 2019 22:24:00 GMT -8
Nationals GM (Preston - Old), Padres GM (Amy), and 7 more like this
Post by Rockies GM (Dan) on Mar 6, 2019 22:24:00 GMT -8
And we're back again with a new league year, and with that comes more analysis! The schedule is up and with that, I can try and do my best to predict how this year is going to go.
If you don't know by now, I created a program that uses algorithms to compile stats from everyone's Fantrax rosters, then simulates the season 100,000 times.
Looking back on last year
Here's how the preseason average wins stack up against reality:
Expected wins are the x-axis, real wins are the y-axis.
The r2 correlation for that is about .58, which is not fantastic, but there's a clear and obvious trend. Looking at the graph, it did a really good job at figuring out the top teams, and had a hard time differentiating teams in the middle of the pack.
Things already changed in a big way with the algorithms here last year when I started simulating each category in every individual matchup rather than the matchup by itself. That drastically changes methodology, and it will be interesting to see how this year's predictions stack up against reality. I'm hoping it will be more realistic.
I also tweaked how the code takes the total number of accumulated stats and comes up with expected stats. I had been linearly projecting the percentage of ABs counted by looking at how total accumulated ABs affected that number historically. The problem is that led to a negative slope (the fewer ABs, typically the fewer stats you are leaving on your bench, the more ABs, the more bench players and non-counted stats you have). But that was imperfect, and penalized teams who accumulated a lot of depth. Now, I did a linear regression on the historical correlation between ABs counted vs total ABs. All the other expected stats are based off the resulting xAB/totalAB percentage.
With those two big changes, I'm hoping to improve that r2 next year.
Without further ado, let's look at predictions for this year!
NL West
Win Spread
Analysis
This probably doesn't take anyone by surprise. The Giants and Rockies have been two of the best teams in the league for the past few years. The Padres have set themselves up for a possible wild card berth, but a division title over either favorite is unexpected. I'm a little surprised at the fact that the simulation only gives the Giants a 1/3 chance at the division. Until OF free agency, both teams had about 50% odds.
NL Central
Win Spread
Analysis
Well, good news St. Louis - the computer likes your odds at the lottery next year. Also, it's looking like there may be a new (old? familiar?) king atop the NL Central. After a two year reign, the Brewers are not expected to be huge players in the race for the division. Instead it likes the Cubs, with the Pirates as the dark horse. Last year, the Pirates had some miserable luck, missing the playoffs after being ~80% favorites for a berth, the algorithm is betting on that not recurring this year. But the Cubs are a comfortable favorite here.
NL East
Win Spread
Analysis
It's going to take a Herculean effort from any of the three teams in the middle to dethrone the Nationals. The computer predicts the Phillies to fall off this year, following their World Series loss. The biggest riser in this division has to be Atlanta, who went from dead last in the division to an outside shot at a wild card.
AL West
Win Spread
Analysis
This is a brand new AL West. The algorithm has the two worst teams in the division from 2018 now battling for the front. Following a very strong offseason, this division looks like the Angels' to lose. But the Mariners are coming on strong, led by phenom (and player I most wish I had never traded) Juan Soto.
AL Central
Win Spread
Analysis
Did anyone expect this to be any different? The Indians are still the favorite with the White Sox hot on their heels. Biggest difference this year is that the algorithm gives Rob and the Twins an outside shot to thrust themselves back into relevancy.
Let's all pretend that last year's Kansas City prediction never counted. I think I've straightened out the anomaly that led to their being a 70% playoff favorite. This year, the algorithm clearly has them continuing their rebuild.
AL East
Win Spread
Analysis
This division looks no different from 2018. The computer likes Toronto once again in their bid to be a threepeat dynasty. Tampa Bay is predicted to be one of the three best teams in the AL, but has only an outside shot to overtake the Blue Jays for the division. The Yankees are improving, and possibly could make a playoff push this year.
If you don't know by now, I created a program that uses algorithms to compile stats from everyone's Fantrax rosters, then simulates the season 100,000 times.
Looking back on last year
Here's how the preseason average wins stack up against reality:
Expected wins are the x-axis, real wins are the y-axis.
The r2 correlation for that is about .58, which is not fantastic, but there's a clear and obvious trend. Looking at the graph, it did a really good job at figuring out the top teams, and had a hard time differentiating teams in the middle of the pack.
Things already changed in a big way with the algorithms here last year when I started simulating each category in every individual matchup rather than the matchup by itself. That drastically changes methodology, and it will be interesting to see how this year's predictions stack up against reality. I'm hoping it will be more realistic.
I also tweaked how the code takes the total number of accumulated stats and comes up with expected stats. I had been linearly projecting the percentage of ABs counted by looking at how total accumulated ABs affected that number historically. The problem is that led to a negative slope (the fewer ABs, typically the fewer stats you are leaving on your bench, the more ABs, the more bench players and non-counted stats you have). But that was imperfect, and penalized teams who accumulated a lot of depth. Now, I did a linear regression on the historical correlation between ABs counted vs total ABs. All the other expected stats are based off the resulting xAB/totalAB percentage.
With those two big changes, I'm hoping to improve that r2 next year.
Without further ado, let's look at predictions for this year!
NL West
| Team | Division% | Wild Card% | Total Playoff% | Average Wins |
| Arizona Diamondbacks | 0% | 0.24% | 0.24% | 6.99 |
| Colorado Rockies | 65.77% | 33.97% | 99.74% | 17.7 |
| Los Angeles Dodgers | 0% | 0.04% | 0.04% | 6.09 |
| San Diego Padres | 0.93% | 54.7% | 55.63% | 12.94 |
| San Francisco Giants | 33.3% | 65.06% | 98.36% | 16.53 |
Win Spread
Analysis
This probably doesn't take anyone by surprise. The Giants and Rockies have been two of the best teams in the league for the past few years. The Padres have set themselves up for a possible wild card berth, but a division title over either favorite is unexpected. I'm a little surprised at the fact that the simulation only gives the Giants a 1/3 chance at the division. Until OF free agency, both teams had about 50% odds.
NL Central
| Team | Division% | Wild Card% | Total Playoff% | Average Wins |
| Chicago Cubs | 65.43% | 28.5% | 93.93% | 15.47 |
| Cincinnati Reds | 0.12% | 2.42% | 2.54% | 9 |
| Milwaukee Brewers | 4.11% | 26.09% | 30.19% | 11.6 |
| Pittsburgh Pirates | 30.34% | 52.39% | 82.72% | 14.4 |
| St. Louis Cardinals | 0% | 0% | 0% | 1.19 |
Win Spread
Analysis
Well, good news St. Louis - the computer likes your odds at the lottery next year. Also, it's looking like there may be a new (old? familiar?) king atop the NL Central. After a two year reign, the Brewers are not expected to be huge players in the race for the division. Instead it likes the Cubs, with the Pirates as the dark horse. Last year, the Pirates had some miserable luck, missing the playoffs after being ~80% favorites for a berth, the algorithm is betting on that not recurring this year. But the Cubs are a comfortable favorite here.
NL East
| Team | Division% | Wild Card% | Total Playoff% | Average Wins |
| Atlanta Braves | 1.78% | 19.43% | 21.21% | 11.27 |
| Miami Marlins | 0% | 0% | 0% | 3.61 |
| New York Mets | 0.8% | 11.57% | 12.37% | 10.48 |
| Philadelphia Phillies | 0.22% | 3.83% | 4.05% | 9.1 |
| Washington Nationals | 97.2% | 1.77% | 98.97% | 16.69 |
Win Spread
Analysis
It's going to take a Herculean effort from any of the three teams in the middle to dethrone the Nationals. The computer predicts the Phillies to fall off this year, following their World Series loss. The biggest riser in this division has to be Atlanta, who went from dead last in the division to an outside shot at a wild card.
AL West
| Team | Division% | Wild Card% | Total Playoff% | Average Wins |
| Houston Astros | 0% | 0.47% | 0.47% | 7.17 |
| Los Angeles Angels | 84% | 14.25% | 98.24% | 16.48 |
| Oakland Athletics | 3.51% | 34.32% | 37.82% | 11.68 |
| Seattle Mariners | 12.49% | 57.81% | 70.3% | 13.46 |
| Texas Rangers | 0% | 0% | 0% | 1.74 |
Win Spread
Analysis
This is a brand new AL West. The algorithm has the two worst teams in the division from 2018 now battling for the front. Following a very strong offseason, this division looks like the Angels' to lose. But the Mariners are coming on strong, led by phenom (and player I most wish I had never traded) Juan Soto.
AL Central
| Team | Division% | Wild Card% | Total Playoff% | Average Wins |
| Chicago White Sox | 21.22% | 35.71% | 56.93% | 12.62 |
| Cleveland Indians | 71.01% | 18.15% | 89.17% | 14.46 |
| Detroit Tigers | 0.05% | 0.29% | 0.34% | 7.05 |
| Kansas City Royals | 0% | 0% | 0% | 3.57 |
| Minnesota Twins | 7.71% | 21.8% | 29.52% | 11.48 |
Win Spread
Analysis
Did anyone expect this to be any different? The Indians are still the favorite with the White Sox hot on their heels. Biggest difference this year is that the algorithm gives Rob and the Twins an outside shot to thrust themselves back into relevancy.
Let's all pretend that last year's Kansas City prediction never counted. I think I've straightened out the anomaly that led to their being a 70% playoff favorite. This year, the algorithm clearly has them continuing their rebuild.
AL East
| Team | Division% | Wild Card% | Total Playoff% | Average Wins |
| Baltimore Orioles | 0% | 0.01% | 0.01% | 5.46 |
| Boston Red Sox | 0% | 0% | 0% | 3.1 |
| New York Yankees | 0.29% | 22.21% | 22.5% | 11 |
| Tampa Bay Rays | 22.05% | 72.92% | 94.97% | 15.32 |
| Toronto Blue Jays | 77.66% | 22.06% | 99.72% | 17.32 |
Win Spread
Analysis
This division looks no different from 2018. The computer likes Toronto once again in their bid to be a threepeat dynasty. Tampa Bay is predicted to be one of the three best teams in the AL, but has only an outside shot to overtake the Blue Jays for the division. The Yankees are improving, and possibly could make a playoff push this year.
