Official: SSBPD unsupported; source code released.

[Mahie]

I agree with the above. The beauty of the Elo system is that it leaves no room to bias, whereas adding decay would mean that you assume that all players lose skill in the same manner over time. Not everyone is going to come back as strong, not everyone will feel the rust the same way.

Hiding their results unless you specifically seek them , and letting them correct themselves on their own if the players are to reenter tournaments, is the best way to go, imo.

Another point is that it works kind of like a history book. You can check up Ken's rating and think " Damn, Ken was pretty good, he dominated as some point ", and then maybe look up the rating of other players during Ken's Era, compare to what they are now, that kind of stuff.
If Ken's rating just keeps going down, he'll eventually be a nobody and you won't have traces of how dominant he used to be.

 

[Fortress | Sveet]

Sorry for not seeing it foxlisk. I have 20 credithours of classes with no blow offs, work 20 hours a week and have my hands in other side projects too. I usually just skim posts that aren't directly aimed at me. I know that's how the system was in the past and it never came to my attention that it changed.

I understand ur decision and I have to agree. Assuming all players decay in skill at the same rate is too strong of an assumption.

 

[OkamiBW]

I have 20 credithours of classes with no blow offs

I'm not the only one?! :o
(Though I'm only doing 16 hours of work a week...

Spoiler

+overtime

)

And for decay, it'd just be based off of time since the tourney. Though, I realize the problem is that some players get worse faster/slower than others. I know Pokemon Online has a decay. And there's something where if you don't play for a certain amount of time, your ranking starts to decay, but then when you play once, your decay goes away in addition to gaining/losing points like usual. Not sure how useful it would be for this ELO rating though.

If Ken's rating just keeps going down, he'll eventually be a nobody and you won't have traces of how dominant he used to be.

It just depends on what the ELO rating stands for. If it's supposed to be an accurate "predictor" of future tournament results, then a history book isn't exactly what you want it to be. Nolan Ryan's got some legit stats in baseball, but if he were to come back today at the age of 65, he's not going to do nearly as well as he did before (not to mention the metagame has evolved in baseball since he played). Alternatively, if it's supposed to be a history book, decay doesn't matter.

 

[Strong Badam]

I think we can just not be ******** and think Elo will do everything for us. Elo will be useful for tournaments, more useful than biased "this guy's about this good" seeding. Ken isn't even rated that high in the system yet. If he turns out to be rated high from old tournaments, so be it. If he comes out of retirement, gets a good seeding, then does poorly, his rating will go down and his seeding for the next tourney will be lower. That's how Elo works. It's not like you can put a number on how much worse you become from being away from the game anyhow.
FoxLisk and the rest of the MBR that's working on this project has already stated that Elo decay will not be included and we've explained why. Drop it and move onto something that will be more worthwhile to discuss.

 

[Kal]

http://www.youtube.com/watch?v=HMFHMBi8BcA

 

[Fortress | Sveet]

I did some research into what TrueSkill actually is. Instead of a singular value for each player, as Elo has it, TrueSkill has a value for skill level and another for uncertainty. As time since the last ranking increases the uncertainty value increases but the skill level value remains the same.

 

[Zivilyn Bane]

I love you Strong Bad.

 

[Bones0]

TrueSkill is what they used for Halo 3/Reach and it's horrible for ranking. Basically what happens is as long as you show improvement your rating will stay flexible. Once you reach a 50% win ratio, the system decides that you have reached your max skill and it becomes less flexible. The problem with this is people plateau and blow up all the time. If you wanted to get a quick ranking of all players over a short period of time, TrueSkill would be great. Over time though, it only loses accuracy. Players who plateaued early have low ratings even if they improve significantly because they have to win so many games in order to raise their rating. Veteran players who achieved obscenely high ratings early in the game's history can repeatedly perform poorly without it affecting their rating. This is the exact opposite of what we want from a ratings system. You don't want to set a player's rating based on a few performances and adjust it a little over a long period of time. You want to get a general idea of a player's skill and then fine-tune their rating based on their present performances instead of their past performances.

This is all just based on what I know from my experience playing Halo 3 and Reach. Idk the details about how they created and maintained ratings, but I know they used TrueSkill and I know that it SUCKS for rankings when improvement and decay in skill is common and often drastic.

 

[Fortress | Sveet]

Bones I agree that it allows for quick rankings, but i disagree that it is harder to gain/lose rank or that the accuracy is less over time. Would this be your perception because of how the matchmaking system works? I could see an issue in Halo matchmaking that once the system was certain of your skill, you would be matched with teammates of that skill thus bogging down your own contribution to your personal wins and losses. One could possibly idle for hundreds of games without seeing much decrease in their ranking since there are 4 other people possibly winning the game.

The Glicko rating system, which TrueSkill is built on, works very similarly to Elo. Points are gained and lost relative to your opponent's skill. The only difference is that it takes into account how certain the system is of its own ranking. Being able to account for how certain your claims are is an invaluable tool. Glicko family rating systems are used by the Australian Chess Federation and by online chess servers such as chess.com

 

[Kal]

Eh, I wouldn't worry about who uses what systems. Elo is used by USCF and FIDE, with FIDE using a normal distribution and USCF using a logistical one (the same one being used by SSBPD).

 

[Tee ay eye]

my ELO went up like 30 points in the past day but idk why. i don't think any new tournaments were even added, let alone those involving myself.

i uploaded 3 tournaments, but none of them were processed/confirmed yet

 

[Paju]

my ELO went up like 30 points in the past day but idk why. i don't think any new tournaments were even added, let alone those involving myself.

i uploaded 3 tournaments, but none of them were processed/confirmed yet

The latest uploads just probably got reprocessed.

 

[Bones0]

Bones I agree that it allows for quick rankings, but i disagree that it is harder to gain/lose rank or that the accuracy is less over time.

As far as I know, the difficulty in gaining/losing your rank is the entire basis of TrueSkill. You have the k value which determines how fast or slow you rank up, and the more games you play, the slower you rank up. If you wanted to use TrueSkill to rank something like intelligence, it would probably work very well because someone isn't going to gain a burst of intelligence no matter how hard they work at it. When applied to Melee and other games, however, people who improve are dragged down by their original ranking. TrueSkill essentially determines your skill level then attaches a ball and chain to you so if you want to improve your ranking you have to perform better than your current ranking repeatedly. Your rating should have as little to do with past performances as possible. TrueSkill reduces the impact improvement has on your rating while Elo increases it. If you place 24th consistently in a 48-man bracket with TrueSkill, getting one match further than normal won't change your rating hardly at all. With Elo, getting one match further means you beat someone rated higher than you, meaning your improvement won't vary based on what level you are at. Going from 6-2 to 7-2 will yield about the same change as going from 1-2 to 2-2 because your rating is determined solely by the players you beat and their relative skill, not on your tournament performance relative to past tournaments.

 

[FoxLisk]

But the point of those confidence-interval influenced ratings systems is to avoid that. If you go 6-2 five times in a row and then go 7-2 once, it's realistically more likely that you had a good day and/or your opponent had a bad day than that you got better. if you go 7-2 again next time your rating will start to catch up. It's not a bad thing.

 

[Strong Badam]

my ELO went up like 30 points in the past day but idk why. i don't think any new tournaments were even added, let alone those involving myself.

i uploaded 3 tournaments, but none of them were processed/confirmed yet

It's possible that older tournaments were added involving players that you beat whose ratings changed retroactively.

 

[Nintendude]

The whole case-sensitivity thing is getting kind of annoying. I understand that tio is not case sensitive, but is there a way to automatically post-process the data so it automatically merges players whose tags only differ by capitalization?

Is this a possibility, FoxLisk?

 

[Fortress | Sveet]

I think he said that TIO is case sensative and there can be players with the same name but different capitalizations and those players could even play each other.

 

[Nintendude]

Well, yeah. The idea is that the ssbpd automatically merges players (perhaps when the data is reprocessed). I doubt there would be 2 players with identical tags in the same tio file.

 

[GOTM]

I doubt there would be 2 players with identical tags in the same tio file.

Doesn't mean it can't happen. Something like this should be up to the TO and the players to sort out so that the correct name is entered into Tio. However, if you want the software to handle it, just compare the lower case versions of the names. Banning the ability to have same names/different case though seems a lot harsher than just making sure every TO is aware of who he is inputting.

 

[Nintendude]

Doesn't mean it can't happen. Something like this should be up to the TO and the players to sort out so that the correct name is entered into Tio. However, if you want the software to handle it, just compare the lower case versions of the names. Banning the ability to have same names/different case though seems a lot harsher than just making sure every TO is aware of who he is inputting.

Nothing is being "banned" and I think it's unreasonable for a TO to know the case of all of the letters in someone's name as it appears in the ssbpd. Overall, there will be much less hassle for everyone if names differing by case get merged automatically. I'd much rather manually sort a few matches every 100 tio files than have to merge profiles all the time.

 

[Fortress | Sveet]

I still think my suggestion before is good: don't automatically insert a player into the DB. Search thru the db for closest matching players and allow the uploader to confirm or create a new player. This solves the case problem as well as giving a solution for duplicate names.

 

[Nintendude]

That's definitely a better idea if FoxLisk is able to make that possible.

 

[Zivilyn Bane]

I still think my suggestion before is good: don't automatically insert a player into the DB. Search thru the db for closest matching players and allow the uploader to confirm or create a new player. This solves the case problem as well as giving a solution for duplicate names.

Wouldn't this take as much time as it takes us to rename the players in Tio before processing? Only that would require participation on the uploaders part, which would most likely make uploaders less interested in uploading.

As far as uploading goes, I like the way it is now because even if somebody doesn't want to do it, it only takes like 5 seconds so they can do it anyways. But if we make the system all of a sudden require time and effort on the part of the uploader, I think we would see a sharp decline in the number of people uploading.

As for myself, I'm fine doing some legwork, although I definitely wish it could be easier or less time consuming. It gets pretty repetitive/boring going through every single file before I process just to change the names to match that of the DB.

 

[Bones0]

It should only get easier to manage tournaments in the future, hopefully. TOs/players will know to make sure tags are put in correctly, and TOs can even encourage people who try to enter with generic names "Billy" to add something to make it unique ("BillyF", "Billy1990", etc.).

 

[Fortress | Sveet]

Well the TO is the person with the most information about the tournament. The more detached the person filling out the player information, the more likely there will be mistakes. Allowing the TO to do it also allows room for growth in the project. For example, if player's characters were eventually added to the system for an automatic Character Ranking List, we could suggest characters for the top 8 as well. The data for the last few CRLs were gathered by myself, by hand. I can tell you that in many cases I had to simply assume the players in question played their mains since not all TOs responded to my PMs. If the TOs were given the option to provide additional information such as this, ssbpd would be able to grow as ideas are presented.

As for time cost, it doesn't have to take more than marginal extra time. There are 3 types of interactions a user has with a program. The one that takes the most time is mouse interaction. Clicking on one button isn't that hard, but more than that gets very tedious. The next easiest is keyboard. If you have keyboard shortcuts, experienced users can speed through the interface very accurately. The absolute fastest is simply eye interaction. Viewing the screen is so easy most people forget to even mention it. By maximizing only eye interaction, the time spent on the confirmation page will be minimal.

In this picture, i have tried to express visually somewhat how i view the confirmation page being. In the vast majority of cases, it is expected that the player's name matches the name in the database besides capitalization. Because of that, the closest matching name is preselect so no work is needed if the name matches. Currently I visualise using the mouse for the exceptions, which would only become tedious if there are many on the same page. I'm sure keyboard shortcuts could be added to improve efficiency.

tl;dr By creating an efficient UI, the amount of discouragement added by this additional page should would be minimal, and compared to the benefits in terms of accuracy and growth potential, I think it would be worth it.

 

[Bones0]

Sveet's interface definitely seems like it'd work great. Even if you think it's too much of a hassle for a TO to skim down the list and mark of which player was actually in attendance, it could still be implemented for the moderators to use to speed up the process of separating players with similar names. Either way, it'd definitely be WAY faster than going through Tio to try and rename players with the proper nickname, especially with case sensitivity enabled.

 

[Fortress | Sveet]

Bones saying my idea is good makes me feel like i just won my second consecutive MLG

[collapse=context]

[/collapse]

 

[Mahone]

is that leffen wearing drpp's pants?

 

[OkamiBW]

Btw, the Kosuke in the Arizona tournaments should be titled under SuperNoko. I'm not sure if Tai knew that, but he prefers to go into the SSBPD as that name.

http://www.ssbpd.com/player/Kosuke
http://www.ssbpd.com/player/SuperNoko

I also still need to upload a tournament or two from my other laptop. :/

 

[Morin0]

Bones saying my idea is good makes me feel like i just won my second consecutive MLG

[collapse=context]

[/collapse]

Spoiler

Unrelated but

great GIF to add in a bee, lol.

 

[Zivilyn Bane]

I like it too Sveet. I'm just a little worried about people actually using it. What's to stop them from just hitting complete and then tournaments getting rated for the wrong player? To me it's still a lot easier to fix the Tio file before we upload than it would be to fix it afterwards, especially since it would be really easy for a wrong entry to go unnoticed.

 

[Fortress | Sveet]

Well I think admins should still review the submissions as well; they could be shown the exact same page, except with the TO's selections. Going forward, the purpose of admins should be verification of data not gathering/manipulation of data. Not to mention, of all smashers, TOs are generally the most responsible and least lazy. Its not all of them, but I will vouch for most of them.

In the short run this would also mean an interface for the files beyond downloading them and resubmitting them. I know foxlisk has the ability to, but i figure he has other things hes working on and he doesn't want to code a temporary solution just to scrap it later when the final solution comes around. The sooner a final solution is discussed, the sooner he agrees on one, and the sooner it would be implemented.

 

[Kal]

I like how my state exemplifies the problem of multiple players having the same name. I think I should look for work in Canada.

 

[~Twitch~]

is that leffen wearing drpp's pants?

bahahahaa!

aight i'll team with you at civil war.

 

[KrIsP!]

Just wanna say something after seeing sveet's picture including 3 different vash's. What if people could add their SWF profile to their ssbpd account. Even as a link. Just to help differentiate between smashingsmasher98 and smashingsmasherWC. It would take some coding I know, just a feature I think would be nifty...something for later on maybe since I'm sure there's lots of work to be done.

 

[Fortress | Sveet]

While I'm not foxlisk, I full support users claiming their profile and filling in their information

 

[FoxLisk]

yeah i've thought about that. haven't really looked into how to do it right yet, may or may not require support from whoever maintains the boards. i'm also not really sure what exactly people would want to do that for? It wouldn't make sense to actually give people control over their results and history and stuff, and I can't think of anything private that a person should be able to see about themselves but not someone else.

The only real benefit i can see from it is allowing people to potentially provide like a pic or a bio or whatnot, which would maybe move things towards being more of a wiki-style effort, but im not at all sure that's the right direction to go in.

anyway i have definitely considered it somewhat, and will give real thought to it some time in the future.

 

[Fortress | Sveet]

Do w/e you want, as long as you keep being a BAWS

 

[Juggleguy]

Ziv, when you get done with the tio file for The Big House, would you mind emailing me back a copy of it?

 

[Fortress | Sveet]

I think John's games from IMDB actually belong to John!.