RPI Description
Disclaimer: Ive been told that I write too much, and write at too complex a level. I believe in giving people all the information they could want so theyre not left second-guessing something thats an omission rather than a mistake. Well handle all of the mechanics read only if you care about the gory details of our new tiebreaker system /disclaimer
For the upcoming season, TWL is planning to change their tiebreaker system to include an RPI-like component. This will hopefully partition teams somewhat better than in the past, where round differential and number of forfeits was an indirect measure of strength of schedule. The old tiebreaker scheme was:
1. Points
2. Wins
3. Head to Head
4. Round-Win Differential
5. Least Number of Forfeit Wins (Strength of Schedule)
6. Extra Head-to-Head match
The new scheme will be:
1. Points
2. Wins
3. If only two teams are tied, Head-to-Head if applicable. If more than two teams are tied, RPI score.
4. If only two teams are tied, RPI score. If more than two teams are tied, Head-to-Head*
5. Round-Win Differential
6. Extra Head-to-Head match
*Head-to-head, when more than two teams are tied, is a special case. If, after RPI ranking you are directly one rank below a team that you have beaten, you will be ranked above them. If you are more than one rank below a team that you have beaten, your head-to-head win will not apply.
Theoretically, the RPI score will not allow for too many ties. It is a very loose version of the college basketball function, and includes no exceptions for things like quality wins. The basic RPI formula is:
The Formula
RPI = 0.25*(your win%) + 0.5*(your opp. win% excluding matches with you) + 0.25*(your opp.s opp win%)
The idea of Opponents win% excluding matches with you is to divorce it from your own winning percentage to prevent double counting. It means that if you beat a 3-4 team, their contribution to your Opp. Win% stat is 3-3 (since they were 3-3 in all matches except with you), which is 0.500. This similarly carries over to Opponents Opponents Win%.
There are five difficulties which come up on our format that the NCAA doesnt have to worry about. These difficulties, and how we address them, are:
- Low sample number: A 30 game season does better statistics than an 8 game season. This system was attempted during Season 2, and it was found to be a fairly reasonable system worth a shot here.
- Forfeits: Since only win percentage is counted, should a forfeit win be equated the same as a real win. We are ruling that forfeits do count in building a teams win/loss percentage, but the team who gets the forfeit gets no benefit in the Opp. Win% metric, since their forfeit win is automatically subtracted out. The only way it really affects statistics is when a team that should have won the match loses (or vice versa), and with roughly 50 60 matches to compensate we expect it wont have much effect.
- Draws: Right now, these functionally count as losses in terms of winning percentage. This is one were still shaky on we could make it that draws just arent counted in terms of total games played for counting win/loss.
- Teams dropping league: All wins will count for your own winning percentage, but teams that do not play at least 4 games in the league will not be counted towards the other two stats (Opp. Win% and Opps Opp Win%). Some opinions were voiced early on about 3 or 5 games, so I stuck with 4 as the middle ground. This too is open to debate.
- Teams playing the same team twice: This is most often just a problem with the West Coasters. Technically, neither of your matches against a team should count against the Opp. Win% stat, meaning that if you beat a 3-4 team twice, they should be counted as 3-2 in BOTH cases for the Opp. Win% stat. Really, this is only a problem with the automatic Excel sheet, which isnt smart enough to handle that, and will be handled manually throughout the season.
The Excel Calculation Sheet
The sheet can be found as RPI Demo Sheet.xls in the Enemy Territory Downloads section of the Teamwarfare website. It only covers through Week 7 as it was written primarily to demonstrate the idea to the admins at the time. This sheet includes five worksheets all with a separate purpose:
- Team Names & IDs just puts teams into a simple numerical format for lookup functions. Referencing a number is better than possibly misspelling a name and messing with search functions.
- Schedule Results is where the weekly results are recorded. Cells in yellow are for manual entry and in blue are for automatic calculation.
- Team & Opp Wins is for summarizing all of the total statistics from the Schedule Results page for simplicity putting it in one contiguous table.
- RPI actually generates the RPI rating based on the stats from Team & Opp Wins
- Sorted was a manual copy and sort of the key stats from RPI - just a way for people to see how it broke up the teams to show how the tiebreakers would work. Thankfully, there were no ties, though Im sure Raiders Coven would have had something to say about finishing 0.0001 points behind t&a on the basis of the Draws rule above.
The idea would be that we would host this sheet throughout the season on ET-Center, or make it otherwise publicly available, so that teams could check to make sure that the stats were recorded correctly, and so that teams could follow the stat and keep track of their relative position.
Bear in mind that this sheet was written pretty much in one pass and was not directly cross-checked for errors. I believe I solved the main ones, but wouldnt assert that its perfect yet. Its more a proof-of-concept sheet, which is admittedly some pretty dense Excel. Any help in correcting errors would of course be appreciated, but I admittedly have an Excel style all my own thanks to a previous boss.
Post edited by AlmightyTDawg at 9/16/2004 11:39:44 PM
Disclaimer: Ive been told that I write too much, and write at too complex a level. I believe in giving people all the information they could want so theyre not left second-guessing something thats an omission rather than a mistake. Well handle all of the mechanics read only if you care about the gory details of our new tiebreaker system /disclaimer
For the upcoming season, TWL is planning to change their tiebreaker system to include an RPI-like component. This will hopefully partition teams somewhat better than in the past, where round differential and number of forfeits was an indirect measure of strength of schedule. The old tiebreaker scheme was:
1. Points
2. Wins
3. Head to Head
4. Round-Win Differential
5. Least Number of Forfeit Wins (Strength of Schedule)
6. Extra Head-to-Head match
The new scheme will be:
1. Points
2. Wins
3. If only two teams are tied, Head-to-Head if applicable. If more than two teams are tied, RPI score.
4. If only two teams are tied, RPI score. If more than two teams are tied, Head-to-Head*
5. Round-Win Differential
6. Extra Head-to-Head match
*Head-to-head, when more than two teams are tied, is a special case. If, after RPI ranking you are directly one rank below a team that you have beaten, you will be ranked above them. If you are more than one rank below a team that you have beaten, your head-to-head win will not apply.
Theoretically, the RPI score will not allow for too many ties. It is a very loose version of the college basketball function, and includes no exceptions for things like quality wins. The basic RPI formula is:
The Formula
RPI = 0.25*(your win%) + 0.5*(your opp. win% excluding matches with you) + 0.25*(your opp.s opp win%)
The idea of Opponents win% excluding matches with you is to divorce it from your own winning percentage to prevent double counting. It means that if you beat a 3-4 team, their contribution to your Opp. Win% stat is 3-3 (since they were 3-3 in all matches except with you), which is 0.500. This similarly carries over to Opponents Opponents Win%.
There are five difficulties which come up on our format that the NCAA doesnt have to worry about. These difficulties, and how we address them, are:
- Low sample number: A 30 game season does better statistics than an 8 game season. This system was attempted during Season 2, and it was found to be a fairly reasonable system worth a shot here.
- Forfeits: Since only win percentage is counted, should a forfeit win be equated the same as a real win. We are ruling that forfeits do count in building a teams win/loss percentage, but the team who gets the forfeit gets no benefit in the Opp. Win% metric, since their forfeit win is automatically subtracted out. The only way it really affects statistics is when a team that should have won the match loses (or vice versa), and with roughly 50 60 matches to compensate we expect it wont have much effect.
- Draws: Right now, these functionally count as losses in terms of winning percentage. This is one were still shaky on we could make it that draws just arent counted in terms of total games played for counting win/loss.
- Teams dropping league: All wins will count for your own winning percentage, but teams that do not play at least 4 games in the league will not be counted towards the other two stats (Opp. Win% and Opps Opp Win%). Some opinions were voiced early on about 3 or 5 games, so I stuck with 4 as the middle ground. This too is open to debate.
- Teams playing the same team twice: This is most often just a problem with the West Coasters. Technically, neither of your matches against a team should count against the Opp. Win% stat, meaning that if you beat a 3-4 team twice, they should be counted as 3-2 in BOTH cases for the Opp. Win% stat. Really, this is only a problem with the automatic Excel sheet, which isnt smart enough to handle that, and will be handled manually throughout the season.
The Excel Calculation Sheet
The sheet can be found as RPI Demo Sheet.xls in the Enemy Territory Downloads section of the Teamwarfare website. It only covers through Week 7 as it was written primarily to demonstrate the idea to the admins at the time. This sheet includes five worksheets all with a separate purpose:
- Team Names & IDs just puts teams into a simple numerical format for lookup functions. Referencing a number is better than possibly misspelling a name and messing with search functions.
- Schedule Results is where the weekly results are recorded. Cells in yellow are for manual entry and in blue are for automatic calculation.
- Team & Opp Wins is for summarizing all of the total statistics from the Schedule Results page for simplicity putting it in one contiguous table.
- RPI actually generates the RPI rating based on the stats from Team & Opp Wins
- Sorted was a manual copy and sort of the key stats from RPI - just a way for people to see how it broke up the teams to show how the tiebreakers would work. Thankfully, there were no ties, though Im sure Raiders Coven would have had something to say about finishing 0.0001 points behind t&a on the basis of the Draws rule above.
The idea would be that we would host this sheet throughout the season on ET-Center, or make it otherwise publicly available, so that teams could check to make sure that the stats were recorded correctly, and so that teams could follow the stat and keep track of their relative position.
Bear in mind that this sheet was written pretty much in one pass and was not directly cross-checked for errors. I believe I solved the main ones, but wouldnt assert that its perfect yet. Its more a proof-of-concept sheet, which is admittedly some pretty dense Excel. Any help in correcting errors would of course be appreciated, but I admittedly have an Excel style all my own thanks to a previous boss.
Post edited by AlmightyTDawg at 9/16/2004 11:39:44 PM