(This article is by guest contributor Carter Schrum)
For a long time, many, if not all tournaments have used use a tie-break method based primarily or solely on speaker points for calculating breaks and determining final rankings. This has been a sore point for many debaters who feel that speaker points are too subjective and that other factors such as strength of schedule should be given greater weight in tie breaks. We have developed a new formula that could largely solve such disparities in modern tabulation processes by giving proportional weight to speaker points and strength of schedule that could better reflect the performance of competitors at a given tournament. This formula is simply an average of three variables: S1 (explained below), W% (Win percentage of a team), and lastly S2 (strength of schedule or Avg W% of opponents).
The formula (S1+W%+S2)/3
This formula seeks to balance the effect speaker points have on overall rank via the use of a key variable which we call "Speaker Point Significance," or simply "Sig" for short. This variable is simply the overall percentage time at a given tournament that the team with the higher speaker points won the round. In our experience, this number tends to sit around 80%-85%. We then take the combined avg speaks of a given team (Usually between 110-163 pts) and multiply it by the "Sig" variable and divide by two to get the first variable used in the formula, S1. This number usually sits between 35 and 75. Here's an example of how we'd calculate S1... Say team: Novice/Novice had a combined speaker point average of 150 and the speaker point significance for the tournament was 81%. The first step would be to multiply 150 * 0.81 = 121.5. We'd then divide this number by two (NOTE: This step is skipped in LD as there is only one debater per team.) to get a final S1 of 60.75.
Once we have found S1, the rest is easy. We simply add the win percentage (W%) of said team for the tournament being ranked as well as the strength of schedule (S2) for a given team to S1 and then divide the sum by three. Given that some teams will have disproportionately high speaks/strength of schedule. It may be necessary to adjust rankings so that a 4-2 doesn't rank higher than a 3-3, etc. This simple yet witty formula could be a much fairer way of determining who breaks at a tournament as well as breaking ties in final rankings. We have already run our system for both Parli and TP debate at the Arkansas Diamond Tournament where it would have ranked some of final standings differently. In fact, it would have sent a different team to TP finals. If you wish to see how you would have done (some would say "should" have done) at Diamond, then links to the full results will be given below (NOTE: we did not adjust these particular ranks to account for lower record teams that managed to overtake higher record teams in overall score. If you so fell the need, feel free to do so...)
And there you have it. A simple formula to fix all those apparent flaws in current Stoa tie-breaks. Another benefit to this system is that there have never been two teams that attain the exact same score, meaning second layer tie breaks (which can be obscure factors such as alphabetical order of last name) would become almost entirely unnecessary and very, very rarely if ever, used. If you believe, as we do, that is a better metric of overall performance and thus a better tie break system, then tell your tournament's tab director about it. Introduce it to your club leader as an alternative way of how to decide breaks and list final rankings. Imagine how much less heartbreak there would be at tournaments such as NITOC where tab just doesn't seem to make sense!
(NOTE: This is in not a prediction model. It is simply for the purposes of measuring individual performances of teams at a given tournament and ranking/breaking based off that measurement.)
Diamond TP Rankings: https://docs.google.com/spreadsheets/d/1PhLtv5gGoieqMbj6KbBDEksDD09Bely14LHATp2xuQ8/edit#gid=0
Diamond Parli Rankings: https://docs.google.com/spreadsheets/d/1r9_WifQZjTonOg6b2rwd1vvedXrpx31uQdr9w_80G0w/edit#gid=0
For a long time, many, if not all tournaments have used use a tie-break method based primarily or solely on speaker points for calculating breaks and determining final rankings. This has been a sore point for many debaters who feel that speaker points are too subjective and that other factors such as strength of schedule should be given greater weight in tie breaks. We have developed a new formula that could largely solve such disparities in modern tabulation processes by giving proportional weight to speaker points and strength of schedule that could better reflect the performance of competitors at a given tournament. This formula is simply an average of three variables: S1 (explained below), W% (Win percentage of a team), and lastly S2 (strength of schedule or Avg W% of opponents).
The formula (S1+W%+S2)/3
This formula seeks to balance the effect speaker points have on overall rank via the use of a key variable which we call "Speaker Point Significance," or simply "Sig" for short. This variable is simply the overall percentage time at a given tournament that the team with the higher speaker points won the round. In our experience, this number tends to sit around 80%-85%. We then take the combined avg speaks of a given team (Usually between 110-163 pts) and multiply it by the "Sig" variable and divide by two to get the first variable used in the formula, S1. This number usually sits between 35 and 75. Here's an example of how we'd calculate S1... Say team: Novice/Novice had a combined speaker point average of 150 and the speaker point significance for the tournament was 81%. The first step would be to multiply 150 * 0.81 = 121.5. We'd then divide this number by two (NOTE: This step is skipped in LD as there is only one debater per team.) to get a final S1 of 60.75.
Once we have found S1, the rest is easy. We simply add the win percentage (W%) of said team for the tournament being ranked as well as the strength of schedule (S2) for a given team to S1 and then divide the sum by three. Given that some teams will have disproportionately high speaks/strength of schedule. It may be necessary to adjust rankings so that a 4-2 doesn't rank higher than a 3-3, etc. This simple yet witty formula could be a much fairer way of determining who breaks at a tournament as well as breaking ties in final rankings. We have already run our system for both Parli and TP debate at the Arkansas Diamond Tournament where it would have ranked some of final standings differently. In fact, it would have sent a different team to TP finals. If you wish to see how you would have done (some would say "should" have done) at Diamond, then links to the full results will be given below (NOTE: we did not adjust these particular ranks to account for lower record teams that managed to overtake higher record teams in overall score. If you so fell the need, feel free to do so...)
And there you have it. A simple formula to fix all those apparent flaws in current Stoa tie-breaks. Another benefit to this system is that there have never been two teams that attain the exact same score, meaning second layer tie breaks (which can be obscure factors such as alphabetical order of last name) would become almost entirely unnecessary and very, very rarely if ever, used. If you believe, as we do, that is a better metric of overall performance and thus a better tie break system, then tell your tournament's tab director about it. Introduce it to your club leader as an alternative way of how to decide breaks and list final rankings. Imagine how much less heartbreak there would be at tournaments such as NITOC where tab just doesn't seem to make sense!
(NOTE: This is in not a prediction model. It is simply for the purposes of measuring individual performances of teams at a given tournament and ranking/breaking based off that measurement.)
Diamond TP Rankings: https://docs.google.com/spreadsheets/d/1PhLtv5gGoieqMbj6KbBDEksDD09Bely14LHATp2xuQ8/edit#gid=0
Diamond Parli Rankings: https://docs.google.com/spreadsheets/d/1r9_WifQZjTonOg6b2rwd1vvedXrpx31uQdr9w_80G0w/edit#gid=0
Comments
Post a Comment