Purpose
Stoa Analysis is a blog dedicated to studying and analyzing the data available on speechranks.com and local tournament websites. Our main purpose is to calculate probabilities of how individuals will do at debate tournaments in both TP and LD. We also may publish other articles explaining the data available on speechranks, or on other topics, usually in order to make educated guesses about the future.
Methodology
Our methodology can be extremely complex, so for the purposes of this article we will simplify it by explaining the core of the model. If you would like to know more about our methodology please contact us.
The models used for predicting debate winners and losers is based on the following formula:
100/(((100-a)/a)(b/(100-b))+1) = x
Where a is the win percentage of debater a, b is the win percentage of debater b, and x is the probability that a beats b.
For example, if a wins 50% of his/her debates and b wins 75% of his/her debaters the formula would work out as follows:
100/(((100-50)/50)(75/(100-75))+1) = 100/(3+1) = 25
So we can conclude in this scenario that a has a 25% chance of beating b.
This formula is the basis of our predictions. From there, models can be created to adjust for power matching, tournament difficulty, byes etc.
The win percentage of each individual debater is based on their preliminary record as recorded by speechranks.com. For example, if you've won eight debates in prelims, and have lost four debates, your win percentage would be 66.67% for purposes of this model. Because more data comes in as the year progresses, it's likely our tournament predictions will improve in accuracy over time.
Accuracy
We hope to publish reports on how accurate the model performs after each major tournament for which it's utilized. Also coming soon will be a report on how well the formula above predicted each individual debate at the Arkansas Diamond tournament. Stay tuned.
Stoa Analysis is a blog dedicated to studying and analyzing the data available on speechranks.com and local tournament websites. Our main purpose is to calculate probabilities of how individuals will do at debate tournaments in both TP and LD. We also may publish other articles explaining the data available on speechranks, or on other topics, usually in order to make educated guesses about the future.
Methodology
Our methodology can be extremely complex, so for the purposes of this article we will simplify it by explaining the core of the model. If you would like to know more about our methodology please contact us.
The models used for predicting debate winners and losers is based on the following formula:
100/(((100-a)/a)(b/(100-b))+1) = x
Where a is the win percentage of debater a, b is the win percentage of debater b, and x is the probability that a beats b.
For example, if a wins 50% of his/her debates and b wins 75% of his/her debaters the formula would work out as follows:
100/(((100-50)/50)(75/(100-75))+1) = 100/(3+1) = 25
So we can conclude in this scenario that a has a 25% chance of beating b.
This formula is the basis of our predictions. From there, models can be created to adjust for power matching, tournament difficulty, byes etc.
The win percentage of each individual debater is based on their preliminary record as recorded by speechranks.com. For example, if you've won eight debates in prelims, and have lost four debates, your win percentage would be 66.67% for purposes of this model. Because more data comes in as the year progresses, it's likely our tournament predictions will improve in accuracy over time.
Accuracy
We hope to publish reports on how accurate the model performs after each major tournament for which it's utilized. Also coming soon will be a report on how well the formula above predicted each individual debate at the Arkansas Diamond tournament. Stay tuned.
Comments
Post a Comment