Griffin's Power Rankings 2017
Mar 17, 2017 14:15:23 GMT -8
Nationals GM (Preston - Old), Rockies GM (Dan), and 6 more like this
Post by Giants GM (Griffin) on Mar 17, 2017 14:15:23 GMT -8
I had some free time during a school break and on a snow day, so I put together my own rankings/projections for this season. The approach is a bit different, so let me walk through what my rankings are. In general, I calculated the projected probability of each team winning a given matchup. The rosters used are those right now (so as of March 17).
Here is a link to my Excel file: drive.google.com/open?id=0B8qcMvjNAfBJQzZtVnJ4UXR4bnM
And here are the steps:
(1) Ranking teams according to stat category: I ranked each team in each category for our league. I used last years stats for each category as a baseline and made adjustments to each team's rank for this year based on my opinion. So, teams are ranked 1-30 in Runs, Homeruns, and so on.
(2) Scaling probabilities: I scaled the probabilities to be between roughly 10-90% for each category. So the best ranked team has about a 90% chance of winning that category, and the last ranked team has about a 10% chance. This aims to account for some random variation through the year.
(3) Correlation: The categories are not independent, and some are in fact highly correlated. For categories with a correlation above 0.75, I penalized teams that were strong in these categories, and I gave an advantage to teams that were poor in both categories. The reasoning is a bad week would disproportionately affect a normally good team in the categories, while a good week would disproportionately help a typically bad team in each correlated category.
(4) Probability of Win: With some help from one of my friends (a Physics major), I used each category's win probability to compute the chance of each team winning 6, 7, 8, 9, 10, 11, and 12 categories in a given week. This required using Mathematica, so the Excel file is not fully dynamic. To do this, I had Mathematica do a bracket expansion of:
Expand[(A + (1 - A) x) (B + (1 - B) x) (C + (1 - C) x) (D + (1 - D) x) (E + (1 - E) x) (F + (1 - F) x) (G + (1 - G) x) (H + (1 -H) x) (I + (1 - I) x) (K + (1 - K) x) (J + (1 - J) x) (L + (1 -L) x)]
Each letter corresponds to the probability of winning a given category (which I plugged in), and the x's serve as a counter.
I then added up each team's probabilities, with winning 6 categories counting as a win half the time.
(5) Projections: I projected each team's wins totals based on the probabilities and made some broad adjustments for strength of schedule, penalizing teams in tougher divisions while rewarding teams in weaker divisions.
Here are my standings and wins/losses projections for the season:
Here is my raw ranking of teams based on their probability of winning a given week:
Let me know about any thoughts you have, or if you see any glaring mistakes! Looking forward to the start of the season!
Here is a link to my Excel file: drive.google.com/open?id=0B8qcMvjNAfBJQzZtVnJ4UXR4bnM
And here are the steps:
(1) Ranking teams according to stat category: I ranked each team in each category for our league. I used last years stats for each category as a baseline and made adjustments to each team's rank for this year based on my opinion. So, teams are ranked 1-30 in Runs, Homeruns, and so on.
(2) Scaling probabilities: I scaled the probabilities to be between roughly 10-90% for each category. So the best ranked team has about a 90% chance of winning that category, and the last ranked team has about a 10% chance. This aims to account for some random variation through the year.
(3) Correlation: The categories are not independent, and some are in fact highly correlated. For categories with a correlation above 0.75, I penalized teams that were strong in these categories, and I gave an advantage to teams that were poor in both categories. The reasoning is a bad week would disproportionately affect a normally good team in the categories, while a good week would disproportionately help a typically bad team in each correlated category.
(4) Probability of Win: With some help from one of my friends (a Physics major), I used each category's win probability to compute the chance of each team winning 6, 7, 8, 9, 10, 11, and 12 categories in a given week. This required using Mathematica, so the Excel file is not fully dynamic. To do this, I had Mathematica do a bracket expansion of:
Expand[(A + (1 - A) x) (B + (1 - B) x) (C + (1 - C) x) (D + (1 - D) x) (E + (1 - E) x) (F + (1 - F) x) (G + (1 - G) x) (H + (1 -H) x) (I + (1 - I) x) (K + (1 - K) x) (J + (1 - J) x) (L + (1 -L) x)]
Each letter corresponds to the probability of winning a given category (which I plugged in), and the x's serve as a counter.
I then added up each team's probabilities, with winning 6 categories counting as a win half the time.
(5) Projections: I projected each team's wins totals based on the probabilities and made some broad adjustments for strength of schedule, penalizing teams in tougher divisions while rewarding teams in weaker divisions.
Here are my standings and wins/losses projections for the season:
Here is my raw ranking of teams based on their probability of winning a given week:
Let me know about any thoughts you have, or if you see any glaring mistakes! Looking forward to the start of the season!