2020-2021 VHL Ratings Package Update
Aug 26, 2021 16:09:46 GMT -5
Philadelphia Flyers, Edmonton Oilers, and 5 more like this
Post by Utah Hockey Club on Aug 26, 2021 16:09:46 GMT -5
Hey All,
After a lot of feedback both in the first ratings thread as well as numerous DMs, I've taken this feedback and updated the ratings package accordingly. This will be the final ratings package for this season. You can see the new ratings HERE and compare with the old ratings HERE. Please don't compare the new ratings with the ratings you see in SIM until Wolf updates SIM. They will be off due to rounding. For example, on my team Nicolas Aube-Kubel was a 67.1 in the GSheet, but due to rounding turned into a 69 in SIM. Likewise, McDavid went from a 87.7 to an 89. Please compare the two GSheets for a more accurate comparison.
The hope here is that this is available to everyone in time for free agency so people can adjust slightly to any changes.
Below are the changes and things to note:
1. The original ratings took every player who played at least one NHL game and applied the same ratings model to them. This resulted in 912 players who had the same rule set applied to them including players who played a handful of games as call ups and this padded the bottom end of the quality of players in the cohort. With the addition of AHL data, any player who played more games in the AHL than NHL, were taken out of the NHL cohort and applied to the AHL model. This results in 774 players in the NHL model and 138 moved into the AHL model. This has results in a depression of all ratings, you'll notice across the board players have a decrease of 1 to 2 OV. Nobody was immune to this. For example, McDavid went from a 87.7 to 86.9. Other players may be more, others less.
2. Any player that played at least one AHL now has a unique rating. AHL players were capped at a max of 65OV. The same is for goalies. For those wondering the best AHL player from that model was Adam Cracknell with a 63.7OV and best goalie was Stuart Skinner with a 65.8OV.
3. With the addition of AHLers there are now 1701 players and 204 goalies with a unique rating. For those number nerds that equates to 30,072 individualized data points across all categories.
4. A few people gave feedback on ratings for players that didn't play full seasons (ie. Tanner Jeannot, Jack Eichel, etc.). This update accounts for that and add in a multiplier to scale back players who failed to play at least 50% of available games. It goes as follows:
1. Greater than 50% games played: 0
2. 35-50% of games played: -1
3. 20-35% of games played: -2
4. 10-20% of games played: -3
5. 0-10% of games played: -4
This is applied to only the skill categories as something like "SZ" isn't affected by how many games you play.
5. The goalies model has been reworked after discussions with other GMs and their worry about goalies that had a high workload were simply rated good for that reason instead of their actual performance such as Martin Jones and Matt Murray. The update to the model attempts to put more weight on SV% and GAA instead of GP and MINS to be a more accurate reflection of a goalies' performance. GP and MINS are still important, but less so than before.
Let me know if you have any questions, but as of now there will be no more work going into this year's ratings. All tweaks moving forward will be put into next year's ratings package. I'll be starting that work early in the new year and doing a thorough audit of the current model to find weaknesses and places to improve, but for now I need a break haha.
docs.google.com/spreadsheets/d/1m28C3PuS3NL5gqp9qJx76jLxOUK9KOuu77U0EFQhBt4/edit
After a lot of feedback both in the first ratings thread as well as numerous DMs, I've taken this feedback and updated the ratings package accordingly. This will be the final ratings package for this season. You can see the new ratings HERE and compare with the old ratings HERE. Please don't compare the new ratings with the ratings you see in SIM until Wolf updates SIM. They will be off due to rounding. For example, on my team Nicolas Aube-Kubel was a 67.1 in the GSheet, but due to rounding turned into a 69 in SIM. Likewise, McDavid went from a 87.7 to an 89. Please compare the two GSheets for a more accurate comparison.
The hope here is that this is available to everyone in time for free agency so people can adjust slightly to any changes.
Below are the changes and things to note:
1. The original ratings took every player who played at least one NHL game and applied the same ratings model to them. This resulted in 912 players who had the same rule set applied to them including players who played a handful of games as call ups and this padded the bottom end of the quality of players in the cohort. With the addition of AHL data, any player who played more games in the AHL than NHL, were taken out of the NHL cohort and applied to the AHL model. This results in 774 players in the NHL model and 138 moved into the AHL model. This has results in a depression of all ratings, you'll notice across the board players have a decrease of 1 to 2 OV. Nobody was immune to this. For example, McDavid went from a 87.7 to 86.9. Other players may be more, others less.
2. Any player that played at least one AHL now has a unique rating. AHL players were capped at a max of 65OV. The same is for goalies. For those wondering the best AHL player from that model was Adam Cracknell with a 63.7OV and best goalie was Stuart Skinner with a 65.8OV.
3. With the addition of AHLers there are now 1701 players and 204 goalies with a unique rating. For those number nerds that equates to 30,072 individualized data points across all categories.
4. A few people gave feedback on ratings for players that didn't play full seasons (ie. Tanner Jeannot, Jack Eichel, etc.). This update accounts for that and add in a multiplier to scale back players who failed to play at least 50% of available games. It goes as follows:
1. Greater than 50% games played: 0
2. 35-50% of games played: -1
3. 20-35% of games played: -2
4. 10-20% of games played: -3
5. 0-10% of games played: -4
This is applied to only the skill categories as something like "SZ" isn't affected by how many games you play.
5. The goalies model has been reworked after discussions with other GMs and their worry about goalies that had a high workload were simply rated good for that reason instead of their actual performance such as Martin Jones and Matt Murray. The update to the model attempts to put more weight on SV% and GAA instead of GP and MINS to be a more accurate reflection of a goalies' performance. GP and MINS are still important, but less so than before.
Let me know if you have any questions, but as of now there will be no more work going into this year's ratings. All tweaks moving forward will be put into next year's ratings package. I'll be starting that work early in the new year and doing a thorough audit of the current model to find weaknesses and places to improve, but for now I need a break haha.
docs.google.com/spreadsheets/d/1m28C3PuS3NL5gqp9qJx76jLxOUK9KOuu77U0EFQhBt4/edit
