A Melee ranking system you're welcome to jump in on

[Stevo]

this seems like an interesting idea

 

[Sausage_Fest]

I thought 100 might have been a little too much at first, but it wasn't that bad after a couple matches. 60 seems fine though, it goes really fast if you are competitive at all with this game.

The database on the website we made is almost ready, and I can have a rough prototype for this entire system up by next weekend. It's gonna be sweet but there is only so much I can do with a real job taking up most of my time.

 

[FabulousGeorge]

Love the idea
Surely there will be the problem that some people are naturally better against certain others.
For example: Player A beats Player B and Player B beats Player C. For the system to work, it will be assumed that Player A will beat and is better than Player C, but due to playstyles or whatever Player C actually beats Player A.
Will this make the system false?
Also, wont it only work if everyone is the best they are going to get? If a player starts off bad, but then improves greatly wont his ranking be lower than it should be due to his early bad results?
I'm not trying to knock it, I really want it to work. I'm just trying to bring some possible problems to light.
Sorry for the incoherency - I'm quite tired.

 

[Manondorf]

Love the idea
Surely there will be the problem that some people are naturally better against certain others.
For example: Player A beats Player B and Player B beats Player C. For the system to work, it will be assumed that Player A will beat and is better than Player C, but due to playstyles or whatever Player C actually beats Player A.
Will this make the system false?
Also, wont it only work if everyone is the best they are going to get? If a player starts off bad, but then improves greatly wont his ranking be lower than it should be due to his early bad results?
I'm not trying to knock it, I really want it to work. I'm just trying to bring some possible problems to light.
Sorry for the incoherency - I'm quite tired.

The best players do not have a style per se. They are dynamic, they adapt and change to suit the situation, so this should not be an issue. A player who is constantly beaten by a play who is regarded as being of lesser skill is a rarity. The system itself is also dynamic. It will be updated as players get better.

I hope this addresses your concerns.

 

[Burnt_Toast]

Love the idea
Surely there will be the problem that some people are naturally better against certain others.
For example: Player A beats Player B and Player B beats Player C. For the system to work, it will be assumed that Player A will beat and is better than Player C, but due to playstyles or whatever Player C actually beats Player A.
Will this make the system false?
Also, wont it only work if everyone is the best they are going to get? If a player starts off bad, but then improves greatly wont his ranking be lower than it should be due to his early bad results?
I'm not trying to knock it, I really want it to work. I'm just trying to bring some possible problems to light.
Sorry for the incoherency - I'm quite tired.

I wanted to add a bit to what Manondorf was saying - Player's A, B, and C in your scenario all have to play several different people to get ranked, not just each other, which helps to wash out results amongst people who play together extremely often and know each other's styles very well. Also, as Manondorf was saying, the system is dynamic, so if you win, your rank improves, exactly the same as if you were to win on a chess server.



****


Rules changes - I know, I said there wouldn't be anymore, but I lied. The KO threshold has been reduced to 50. The number of different people you have to play to get ranked has been increased from 4 to 5, and I've done away with the contest minimum. So basically, just 5 first to 50's against 5 different players and you're ranked.

It's worth mentioning (and I likely should have done so sooner) that, while this isn't a rule, you will need to have at least one victory and one loss to have a truly legit rank. This is because if you have no losses, the math will see your maximum potential rank as infinity, and conversely negative infinity if you have no wins. So if you're just trying to get ranked as quickly as possible, it would be prudent to seek both stronger and weaker players. The strongest person in Melee, however, does not have to worry about this. There's nothing wrong with the system thinking your rank potential is infinite when you actually are the best.


****


In other news, we had a great first day at Don't Go Down There, Jeff. We completed 6 contests including Silent Spectre vs Lucien and I got beaten by a girl for the first time ever in competitive gaming in the closest one of these I've done to date. Our results can be found here starting with contest 000015.

000015 - Silent Spectre vs. Lucien
000016 - Lucien vs. Dunk
000017 - Me vs. Vans
000018 - Smoke 2 Joints vs West Ballz
000019 - Zelgadis vs (Zel I need the name of your oppoenent)
000020 - Mr. Ganondorf vs Sam

 

[HomeMadeWaffles]

i did this with zhu made my fingers hurt lol

other then that i thought it was pretty cool idea

i still kinda dont understand what its for tho >_>

o yea btw this is jman

 

[Burnt_Toast]

J-Man - don't worry, you're not alone. I'm currently working on a post that will summarize a lot of what's been said in this thread and (hopefully) make things a lot more clear to everyone. For now, I'll just say that your contest against Zhu was the most important one completed at the tournament and thank you guys so much for being willing to start from scratch when the data was lost half-way through. You could also read through this post in the interim.



********


We had an even more amazing second day at Don't Go Down There, Jeff. We completed 11 contests for a total of 17 for the weekend and there are now 30 people in this system compared to just 7 before the weekend began.

Here are all of our screenshots (first day of DGDTJ starts at number 000015 and the second starts at 000021) and here is a picture of each player's statistics page. I've also linked to the actual Excel files at the bottom of the first post of this thread.

We also recorded all of Silent Spectre vs. Axe's Pikachu, Silent Spectre vs. Axe's Falco, and J-Man vs. Zhu, so you can expect about 45 awesome matches to appear on youtube sometime soon.

 

[OkamiBW]

So would it be better to play people who are already on this ranking system? For example, should I go and play Signia, Jkun/DJMirror, and Axe since they'll be the easiest for me to play?

Also, could you post a step-by-step (1., 2., 3., etc.) of what we need to do? If not, I'm just going to record a run with Signia starting from making the names to looking at the stats and I'll let you sort through it after I upload it.

 

[Burnt_Toast]

Yeah, playing one of those people you mentioned will connect you to everyone in AZ and Nor-Cal, and then you can play whoever. Ultimately, play whoever you want/can as the ultimate goal is for everybody to be connected.


And here, I made a couple videos showing step by step how to set these up -

Part 1

Part 2

 

[Rostigalen]

You know, this seems quite interesting, and I like the way you set up smashranks.com so far, even though most things are under construction. I'll try to get to play this against one of my pals in the next few weeks - either that, or under B.E.A.S.T (International tournament in Sweden).

 

[Signia]

i did this with zhu made my fingers hurt lol

other then that i thought it was pretty cool idea

i still kinda dont understand what its for tho >_>

o yea btw this is jman

It's a ranking system. Pretty sure it's a way of keeping track of how good everyone is. If you've ever played an arcade fighting game ranking match, or an online fighting game ranking match, it's like that.

 

[john!]

It's a ranking system. Pretty sure it's a way of keeping track of how good everyone is. If you've ever played an arcade fighting game ranking match, or an online fighting game ranking match, it's like that.

crazy utahraptor!

 

[XAQ]

bump bump bump bump bump bump bump

 

[Burnt_Toast]

Thanks for the bump XAQ and sorry for the lack of recent information.

***

The statistics section of the website is almost complete and should be ready to use within a few days. Once completed, anyone will be able to go to the site and enter the results of his/her own contests (there will be a template to fill out KOs, Falls, SDs, etc). Once submitted, one of us will then verify your input against your screenshots and approve the match. At this point, the results will be assimilated into the algorithm.

***

We've got a couple set ups at Winter Game Fest in San Diego. Our goal here is the same as ever - collect as much data from as many traveling smashers as possible. Please come on by and get ranked.

***

Also, the current KO threshold still seems to be a little high (especially for slower paced characters). I'm considering 30, however, I'd like some opinions on this. This means that the average contest would be about 10 or 11 matches.