-
Welcome to Smogon! Take a moment to read the Introduction to Smogon for a run-down on everything Smogon, and make sure you take some time to read the global rules.
-
Congrats to the winners of the 2023 Smog Awards!
OMPL VIII - Discussion Thread
- Thread starter Ransei
- Start date
- Status
(no offense intended)
Hey there nerds and... yeh you're all nerds.
The tl;dr of this post is: I calculated ELOs and price per wins for everyone up until this point. If you want to read about why or how read the rest of this post. The main thing to know about the ELO is that it is not directly equivalent to PS ladder ELO. I did not set a minimum so many people are below 1000, and it scales somewhat differently. I recommend you compare ELOs to each other rather than to any outside usage of the system. If you don't want to bother reading the rest of this post, check out my spreadsheets linked here:
CLICK THIS TO CHECK OUT THE SPREADSHEET
note that there are two sheets in there, so if you wanna see ELOs you have to go to the second page
I've been crunching some numbers for y'all the past couple days. I usually make a post at the end of OMPL analyzing the Price/Win of every player as a method of determining which players were the best deals of the draft. Not which teams had the best drafts though, because every team spends just about the same amount so... whoever had the most overall wins had the best price per draft. Kinda useless for that.
I decided to get a head start on that this year, and simultaneously
came to the realization that its really not a good way to measure player skill or performance. Not that we can really objectively measure either of those in a tour like this: stuff like hax and a small sample size mess with any data analysis you try to do. HOWEVER, I decided to look for a better way to analyze this.
I started with a tours truism: W/L cannot directly measure performance. There are several reasons given for this, some of which can't be helped (hax again, and other factors that are at least somewhat luck-based like matchup, out of battle contributions, Gmansour20 being cute af). The one I decided to try and do something about was the fact that not all opponents are equal. Not every win is as impressive, nor every loss as demeaning.
I thought about it briefly, and realized there was a way to measure player skill taking into account the opponents performance right in front of me, something we've all used: our ladder rankings system.
I used an ELO calculator (here, which I found linked on this Smogon page explaining ratings) to calculate every players ELO. I used a weighting factor of 100 and did NOT floor the ELO at 1000. This means that the ELO numbers themselves are not particularly equatable to the ones you will be used to from PS! ladders. PS uses a much more complex process, which makes sense in the context of a ladder, but a) doesn't make as much sense in a tour setting for various reasons and b) is hard and I b1) lack the technical knowhow and capability and b2) don't wanna. There is no decay, and I don't plan to include any.
I was hoping to be able to get a GXE set up for each player but that proved a little beyond me unfortunately. If anyone could show me how to calculate it (as in: show me what to plug the numbers into, not actually explain the math) I would love to do it.
Once the tour itself is over I will make a larger post talking about who performed the best as a whole, how it compares to player prices, stuff like that. For now, if you are interested I recommend checking out the Google Doc above to see how everyone stacks up.
The tl;dr of this post is: I calculated ELOs and price per wins for everyone up until this point. If you want to read about why or how read the rest of this post. The main thing to know about the ELO is that it is not directly equivalent to PS ladder ELO. I did not set a minimum so many people are below 1000, and it scales somewhat differently. I recommend you compare ELOs to each other rather than to any outside usage of the system. If you don't want to bother reading the rest of this post, check out my spreadsheets linked here:
CLICK THIS TO CHECK OUT THE SPREADSHEET
note that there are two sheets in there, so if you wanna see ELOs you have to go to the second page
I've been crunching some numbers for y'all the past couple days. I usually make a post at the end of OMPL analyzing the Price/Win of every player as a method of determining which players were the best deals of the draft. Not which teams had the best drafts though, because every team spends just about the same amount so... whoever had the most overall wins had the best price per draft. Kinda useless for that.
I decided to get a head start on that this year, and simultaneously
Hey Ferb, isn't simultaneously on our list of S words seldom used by kids?
came to the realization that its really not a good way to measure player skill or performance. Not that we can really objectively measure either of those in a tour like this: stuff like hax and a small sample size mess with any data analysis you try to do. HOWEVER, I decided to look for a better way to analyze this.
I started with a tours truism: W/L cannot directly measure performance. There are several reasons given for this, some of which can't be helped (hax again, and other factors that are at least somewhat luck-based like matchup, out of battle contributions, Gmansour20 being cute af). The one I decided to try and do something about was the fact that not all opponents are equal. Not every win is as impressive, nor every loss as demeaning.
I thought about it briefly, and realized there was a way to measure player skill taking into account the opponents performance right in front of me, something we've all used: our ladder rankings system.
I used an ELO calculator (here, which I found linked on this Smogon page explaining ratings) to calculate every players ELO. I used a weighting factor of 100 and did NOT floor the ELO at 1000. This means that the ELO numbers themselves are not particularly equatable to the ones you will be used to from PS! ladders. PS uses a much more complex process, which makes sense in the context of a ladder, but a) doesn't make as much sense in a tour setting for various reasons and b) is hard and I b1) lack the technical knowhow and capability and b2) don't wanna. There is no decay, and I don't plan to include any.
I was hoping to be able to get a GXE set up for each player but that proved a little beyond me unfortunately. If anyone could show me how to calculate it (as in: show me what to plug the numbers into, not actually explain the math) I would love to do it.
Once the tour itself is over I will make a larger post talking about who performed the best as a whole, how it compares to player prices, stuff like that. For now, if you are interested I recommend checking out the Google Doc above to see how everyone stacks up.
- Status