Before we get to the GCC predictions, I need to give a couple of disclaimers due to a bit of an anomaly. If I don't, you'll probably think that this post is basically this meme come to life:
(Yes, it is Dylan George writing this if you didn't know)
So here's the issue: I've not been to a lot of tournaments (two, to be exact). At those tournaments, KC Kickoff and Arkansas Diamond, I've won a total of 11 rounds and lost 1. Therefore my model treats me as if I have a 92% win rate because I've won 92% of my debates. This is vastly higher than any other debater at GCC. So am I really the best or have I just gotten lucky? I don't know, but I'm pretty sure it's the latter. Until we have a larger sample size however, the model is treating my past results as predictive. Maybe you think the model is super inaccurate in giving me such a high chance of making finals/winning. Maybe you're right, I don't know. But I am convinced that the rest of the model's predictions of everyone's chances of going 6-0, 5-1 etc. is accurate. We will see just how accurate after the tournament. The main point I want you to take away is that this mode is NOT my opinion. The model was predetermined and I am simply relaying the results, which I believe are accurate overall. Outliers (like my record) don't justify discarding the whole model.
There's one more issue though. I've developed what I believe to be a very good system of determining everyone's chances at finals. Determining everyone's chances of winning however is much harder for multiple reasons. First, it's calculating someone's chances against an unknown debater, which is nearly impossible. Second, there are multiple judges, which may mean that the better debater will win more often than my model currently anticipates. Third, if someone of a low-skill level makes finals, it's reasonable to assume that they've gotten better and therefore would be more likely to win finals than the model would anticipate. All of this makes predicting finals hard. So I'm presented with two options:
1. Calculate everyone's chances against the average finalist. This will mean that the better-skilled debaters will be counted as more likely to win finals if they were to make it than lower-skilled debaters. The main problem with this system is lower-skilled debaters may be in finals because they have gotten better and not because they were lucky and therefore should not be counted as a huge underdog. Although it can almost be proven that some lower-skilled debaters have made it before by luck, while others haven't.
2. Assume everyone has a 50% chance to win once making finals. This isn't unreasonable, given the fact that the two people who make finals will likely both be very skilled, so much so that the difference between them would be insignificant. However, there have been cases in which there is a clear favorite going into finals and if a certain debater were to make finals, it might be improper not to consider him/her an underdog.
Both of these methods have flaws associated with them. Personally, I'm just satisfied to know everyone's chances at making finals, but I feel it would be insufficient to not include everyone's chances of winning. So while I used the first method for predicting Escalade, I will be using the second method for predicting GCC, although I will include the results of both in my Excel sheet below. The reason I'm changing the method for this tournament is that the first method would essentially have me at an 80% chance at winning finals if I were to make it and a 40% chance of winning overall because of my record. I just don't feel justified in giving myself an 80% chance to win finals if I were to make it. Maybe I'm dead wrong, maybe I shouldn't change my method and maybe I really am just that good. Maybe I'm completely unjustified in switching my winner calculations based off my personal opinion. I doubt it though. And like I said before, I'll include both in the Excel sheet. So for the below prediction, I will assume everyone would have a 50% chance of winning finals if they were to make it.
And now, finally, what you've been waiting for, the GCC LD predictions of winning:
Dylan George - 26%
Jonathan Arens - 9%
Jacob Priestner - 7%
Adam Wilkins - 7%
Mattea Miller - 6%
David Zilkie - 5%
Cedar Roach - 4%
Alyssa Sloneker - 4%
Nathan Boone - 4%
Josh Buxton - 4%
Elizabeth Miller - 3%
Naomi Wiedmaier - 2%
Charles Miller - 2%
Jonathan Zilkie - 2%
Samuel Chapman - 2%
All other debaters are at 1% or less.
Full results can be viewed here:
https://docs.google.com/spreadsheets/d/1bzLLEgJ49fJkWD4OLhkFRmR_BuuS-F5oTCwUefQfL4c/edit?usp=sharing
(Yes, it is Dylan George writing this if you didn't know)
So here's the issue: I've not been to a lot of tournaments (two, to be exact). At those tournaments, KC Kickoff and Arkansas Diamond, I've won a total of 11 rounds and lost 1. Therefore my model treats me as if I have a 92% win rate because I've won 92% of my debates. This is vastly higher than any other debater at GCC. So am I really the best or have I just gotten lucky? I don't know, but I'm pretty sure it's the latter. Until we have a larger sample size however, the model is treating my past results as predictive. Maybe you think the model is super inaccurate in giving me such a high chance of making finals/winning. Maybe you're right, I don't know. But I am convinced that the rest of the model's predictions of everyone's chances of going 6-0, 5-1 etc. is accurate. We will see just how accurate after the tournament. The main point I want you to take away is that this mode is NOT my opinion. The model was predetermined and I am simply relaying the results, which I believe are accurate overall. Outliers (like my record) don't justify discarding the whole model.
There's one more issue though. I've developed what I believe to be a very good system of determining everyone's chances at finals. Determining everyone's chances of winning however is much harder for multiple reasons. First, it's calculating someone's chances against an unknown debater, which is nearly impossible. Second, there are multiple judges, which may mean that the better debater will win more often than my model currently anticipates. Third, if someone of a low-skill level makes finals, it's reasonable to assume that they've gotten better and therefore would be more likely to win finals than the model would anticipate. All of this makes predicting finals hard. So I'm presented with two options:
1. Calculate everyone's chances against the average finalist. This will mean that the better-skilled debaters will be counted as more likely to win finals if they were to make it than lower-skilled debaters. The main problem with this system is lower-skilled debaters may be in finals because they have gotten better and not because they were lucky and therefore should not be counted as a huge underdog. Although it can almost be proven that some lower-skilled debaters have made it before by luck, while others haven't.
2. Assume everyone has a 50% chance to win once making finals. This isn't unreasonable, given the fact that the two people who make finals will likely both be very skilled, so much so that the difference between them would be insignificant. However, there have been cases in which there is a clear favorite going into finals and if a certain debater were to make finals, it might be improper not to consider him/her an underdog.
Both of these methods have flaws associated with them. Personally, I'm just satisfied to know everyone's chances at making finals, but I feel it would be insufficient to not include everyone's chances of winning. So while I used the first method for predicting Escalade, I will be using the second method for predicting GCC, although I will include the results of both in my Excel sheet below. The reason I'm changing the method for this tournament is that the first method would essentially have me at an 80% chance at winning finals if I were to make it and a 40% chance of winning overall because of my record. I just don't feel justified in giving myself an 80% chance to win finals if I were to make it. Maybe I'm dead wrong, maybe I shouldn't change my method and maybe I really am just that good. Maybe I'm completely unjustified in switching my winner calculations based off my personal opinion. I doubt it though. And like I said before, I'll include both in the Excel sheet. So for the below prediction, I will assume everyone would have a 50% chance of winning finals if they were to make it.
And now, finally, what you've been waiting for, the GCC LD predictions of winning:
Dylan George - 26%
Jonathan Arens - 9%
Jacob Priestner - 7%
Adam Wilkins - 7%
Mattea Miller - 6%
David Zilkie - 5%
Cedar Roach - 4%
Alyssa Sloneker - 4%
Nathan Boone - 4%
Josh Buxton - 4%
Elizabeth Miller - 3%
Naomi Wiedmaier - 2%
Charles Miller - 2%
Jonathan Zilkie - 2%
Samuel Chapman - 2%
All other debaters are at 1% or less.
Full results can be viewed here:
https://docs.google.com/spreadsheets/d/1bzLLEgJ49fJkWD4OLhkFRmR_BuuS-F5oTCwUefQfL4c/edit?usp=sharing
Comments
Post a Comment