Official: SSBPD unsupported; source code released.

[Mahie]

It's the only "correct", international, and language independent way to express dates. And it's the official one in my country. And it can be sorted.

Year first, really ? Wouldn't you go D, M, Y ? As in today being 06/04/12 ?

 

[FoxLisk]

Bam! You can now get tournaments seeded - go to TO Tools and upload a file that includes (at least) a player list, it will find whichever players it can and return you a copy of your file with their skill set accordingly.

Also, redid some of the copy, as it was out of date/********

 

[Nintendude]

Wow, that is EPIC!

 

[FoxLisk]

Wow, that is EPIC!

Thanks!

I actually was chatting with this about Nintendude, but to get it out there for everyone:

The system currently assigns player skill very naively. It just squishes the actual ratings down into the 1-10 scale linearly. This is all well and good in terms of "fairness" and "accurate representation of the data," but is unlikely to really be very useful for having a tournament that is "seeded well" and "fun to be at" and "not rigged." Basically, in the few tournaments that I've run through it, you get a disproportionate section of the entrants rated 1-3, because the top couple players stretch the higher end out a lot.

So uh, i'm thinking about how to solve that. The skill ratings will improve over time. I'll bug my friends who know statistics and listen to what y'all have to say and stuff and we'll settle on a better method of squishing the wide range of elo scores down to the narrow range of tio-acceptable scores.

 

[Dees]

When I search for myself it says there are no matches. But in the list of people I may have meant it has my name exactly as I typed it in. Does that happen with other people too?

 

[Bones0]

Yeah, it happens for me. It's just so if there is someone with a similar name they can find themselves. Like if there were me, "Bones", and someone named "Bonez", but someone searched "Bones", both names would come up so they can go "Oh, I want the one spelled with a 'z.'" Personally, I just go to the full player list and CTRL+F for names when I'm looking for an exact name.

 

[Strong Badam]

Duck, from Michigan, was not at 'E52 Battle To End Hunger' which was recently added to his history. I believe that's a different person named Perfect Duck, from Kansas.

yeah that's blake

 

[OkamiBW]

Can't we just release a new version of TIO with the grand finals glitch fixed and skill ratings of 1 through 9999?

 

[FoxLisk]

Can't we just release a new version of TIO with the grand finals glitch fixed and skill ratings of 1 through 9999?

I am unable to do that. Tio is created and maintained by Neal Dt, so it's up to him to handle things like that. There's been some word of being able to add functionality on to Tio via plugins, but it's unclear that that will be possible anytime soon.

 

[Inty17]

So I finally looked at my ELO rating on your website, and I saw that my match with KayS at G-rev XVII was recorded incorrectly. It says he beat me 2-0, but I beat him 2-0 win. Could someone correct this? I know it's not a big deal, but I hope someone fixes it.

 

[Fortress | Sveet]

Foxlisk, could you get a list of the matched players and then evenly distribute them between 1 and 10?

 

[FoxLisk]

Foxlisk, could you get a list of the matched players and then evenly distribute them between 1 and 10?

I think this will spread out lower rated players at the cost of squishing together higher rated ones. It's basically saying "the top ~10 players are evenly skilled" which is rarely close to true.

This is probably a good solution for generating pools and probably not so good for generating brackets. I haven't done much testing but those are my thoughts on that method.

It would be reasonable to have different skill assignation metrics based on pools and brackets, If people would like that, actually.

 

[Zivilyn Bane]

I think this will spread out lower rated players at the cost of squishing together higher rated ones. It's basically saying "the top ~10 players are evenly skilled" which is rarely close to true.

This is probably a good solution for generating pools and probably not so good for generating brackets. I haven't done much testing but those are my thoughts on that method.

It would be reasonable to have different skill assignation metrics based on pools and brackets, If people would like that, actually.

What about just a simple sliding scale like:
Elo 2000+ = skill lvl 10
1900+ = 9
1800+ = 8
Ect
Ect
1100 and below = 1

 

[Nintendude]

I think that skill levels 10, 9, 8, and 7 should be saved for the top-4 seeds, as they do in a lot of sporting events (such a March Madness, where they seed the four 1 seeds). At the very least, skill levels 10 and 9 should be reserved and if there are no upsets, the two 10s should face in winners finals.

 

[Fortress | Sveet]

I think this will spread out lower rated players at the cost of squishing together higher rated ones. It's basically saying "the top ~10 players are evenly skilled" which is rarely close to true.

This is probably a good solution for generating pools and probably not so good for generating brackets. I haven't done much testing but those are my thoughts on that method.

It would be reasonable to have different skill assignation metrics based on pools and brackets, If people would like that, actually.

1. Kels- 1793
2. Dart- 1700
3. DarkAtma- 1688
4. Tink- 1602
5. Matt R- 1544
6. Duck- 1528
7. Tom R- 1519
8. Rat- 1473
8. ORLY- 1473
10. Sveet- 1419
11. Moose- 1407
12. Juggleguy- 1327

If these guys all went to the tourney, then the pools would be seeded accordingly.

Spoiler

Pool1:
Kels
Rat
ORLY

Pool2:
Dart
Tom
Sveet

Pool3:
DarkAtma
Duck
Moose

Pool4:
Tink
Matt
Juggleguy

btw, Little England is at 1399. He deserves higher, so would you please upload your files chris?

The numbers are out of date since I posted this a while ago, but it should explain my point pretty well


edit- to do this programatically, just take however many pools there are and make that many players at each skill rank. For example, in this one, put Kels, Dart, Atma and Tink at skill 10 and Matt, Duck, Tom and Rat at skill 9, etc.

 

[Kal]

I think the best way to handle it is to simply set the highest ranking player in your tournament as "10," and then scale accordingly (rounding values to the nearest integer). This might require some modification of Tio, though.

 

[Fortress | Sveet]

R u saying something like

step: (maxElo - minElo)/10

So that everyone between maxElo and (maxElo-step) gets a 9 (u said max gets 10 right?)


Seems like a decent way to do it. Standard deviation stuffs possible?

 

[Kal]

I would ask someone like Ferrish to come up with a more accurate distribution (I'm sure normal or logistical is better), but I was just suggesting something as easy as the following.

Suppose, in our list of entrants, player A with skill a has the highest rank. Then we assign player A a rank of 10. For a player B with skill b, we assign him rank (b/a)*10, rounded to the nearest integer.

I'm sure there's a better way to do it, e.g. taking the lowest ranking player and highest ranking player and distributing everyone in between logistically or normally, but I'm not very savvy when it comes to this sort of thing. Seriously, we should just consult Ferrish. >_>

 

[PGH Carroll]

Can someone with the power fix a players name.

There's about is about three "Dr. Stienbock"s

Theyre just all spelled differently lol.

It should be "Dr. Stienbock" and while your at it go head and put him down for Pennsylvania and doctor Mario.

 

[Bones0]

1. Two of the Shell Shocked tournaments that were uploaded show up as having 0 entrants. Is this from something I did, a processing error, or just a normal precursor to a proper upload?

2. Two of the tournaments I entered have my matches being played on two different dates even though both tournaments were 1-day events.

3. The chart that has a line graph of my past Elo has seemingly random dates. I just thought it was strange. It isn't really a huge deal, but I'm assuming whatever is causing the problem with #2 is likely related to #3

4. This is super nitpicky, but there is no pop-up text for the flag images when you go to the Character page. For instance, when you hover over a US flag it usually says what state the player is from. None of the other flags say anything either. Hopefully that's just a minor HTML issue.

 

[FoxLisk]

1. Two of the Shell Shocked tournaments that were uploaded show up as having 0 entrants. Is this from something I did, a processing error, or just a normal precursor to a proper upload?

This is an error that is probably something really arcane and weird that you did. I'll look into it tomorrow and let you know/try to account for it in the future.

2. Two of the tournaments I entered have my matches being played on two different dates even though both tournaments were 1-day events.

this is just an artifact of me handling dates foolishly, dont worry about it. it doesnt affect anything.

3. The chart that has a line graph of my past Elo has seemingly random dates. I just thought it was strange. It isn't really a huge deal, but I'm assuming whatever is causing the problem with #2 is likely related to #3

those dates are the same as the actual tournament dates but off by one month. presumably I messed up the date formatting somewhere. I'll take care of that this weekend too. Thanks for pointing it out.

4. This is super nitpicky, but there is no pop-up text for the flag images when you go to the Character page. For instance, when you hover over a US flag it usually says what state the player is from. None of the other flags say anything either. Hopefully that's just a minor HTML issue.

oops. my bad. will handle that this weekend too.

 

[Bones0]

This is an error that is probably something really arcane and weird that you did. I'll look into it tomorrow and let you know/try to account for it in the future.

Oh wait, CTL was the one who uploaded those. I uploaded Shell Shocked X.
*facepalm*

 

[Zivilyn Bane]

Look at the balance.

Ten different character mains in the top 14. Not too shabby.

 

[Bones0]

If a mod could split THIS player, that'd be great. :D Just take the matches from "Velocity's Fest 1" and change his name to "DJ-MD" (Marth main from MD). Idk who the other DJ is though.

 

[OkamiBW]

So I finally looked at my ELO rating on your website, and I saw that my match with KayS at G-rev XVII was recorded incorrectly. It says he beat me 2-0, but I beat him 2-0 win. Could someone correct this? I know it's not a big deal, but I hope someone fixes it.

Did somebody acknowledge this at all?

 

[FoxoF]

Hi,

I haven't read this thread in its entirety, so I'm not sure if this issue has come up yet, but...

Are there any adjustments made for time?

As in, are more recent wins more valuable that very old wins?

If not, can someone explain to me why this shouldn't be the case?

I'm just trying to get my head round the concept, because, in reality, aren't more recent wins more valuable that wins you got in 2004?

Thanks guys

 

[Strong Badam]

More recent wins overall mean more to your rating because of chronology, however the values of the wins don't change afaik. Given enough data, even if someone was terrible years ago and is good now, their rating will reflect that.

 

[Kal]

This summarizes it. Any way to account for skill as a function of time is going to be arbitrary, e.g. you are going to choose some scaling value. Elo already accounts for time, provided everyone continues to play at the same frequency. All you truly need to account for the issue, then, is to hide inactive players, with some arbitrary threshold chosen to distinguish between active and inactive players.

 

[Cosmo!]

This is an interesting project.

I remember speculating that Elo would fall flat on its face if it was going to be applied to a local power ranking, but when you track down hundreds of tournaments and try to build an international system it doesn't actually look so bad!

The only downside is it does currently lack a ton of old data and I feel it was started too late in the game's lifespan. Great site / project though, congrats

 

[OkamiBW]

So I finally looked at my ELO rating on your website, and I saw that my match with KayS at G-rev XVII was recorded incorrectly. It says he beat me 2-0, but I beat him 2-0 win. Could someone correct this? I know it's not a big deal, but I hope someone fixes it.

Poor guy. Guess we're just going to ignore him. Maybe if I make my post in bright pink people would notice the quote. Or heck, I'll make the quote in pink.



Notice his post please! :x

 

[Zivilyn Bane]

Inty I can fix that for you tonight. I try to get some verification with either a YouTube vid or the TO. Do you have either? Or OkamiBW, were you the TO for that event?

 

[OkamiBW]

Inty I can fix that for you tonight. I try to get some verification with either a YouTube vid or the TO. Do you have either? Or OkamiBW, were you the TO for that event?

Nah, I wasn't. I was just bugged by everyone ignoring him. *shrugs*

Though, I did TO this unseeded tourney: http://www.smashboards.com/showthread.php?t=320724

Spoiler

Check out these results, lol:
1: Mew2King
2: Rubyiris
3: Silly Kyle
4: OkamiBW
5: Neil
5: KiraFlax (brawl player, lol)
7: Axe
7: RaveRemix (another brawl player)

 

[Strong Badam]

ehh idk about no entry fee tourneys. usually there's not much incentive to try your hardest. axe went YL lol and m2k went falco/peach

 

[Kal]

I don't think that's a fair excuse to dismiss results, Strong Bad. I think we need a dichotomy where we either ignore all results or allow all results.

 

[Dekar289]

uh yes actually, it is a fair excuse
no entry fee = no pot = no try
there's enough sandbagging in regular tournies

 

[FoxLisk]

I don't think that's a fair excuse to dismiss results, Strong Bad. I think we need a dichotomy where we either ignore all results or allow all results.

This is definitely false. But I think I probably agree that no-entry tournaments are acceptable as long as they're played with a standard ruleset.

@Okami & Inty17

sorry to ignore you. There's currently no convenient way to change the result of a set through the UI, so it requires erasing the tournament, fixing the file, and reuploading it. I will add something so the mods can change individual set results at some point, and then these things will be easy to fix. There have been a bunch of requests like this so I've just decided not to do a bunch of them individually and wait until I get that up. Sorry again; it'll be taken care of before long.

 

[Strong Badam]

I think free entry tourneys are fine, but specifically free entry tourneys like this one that is before a regional/major, similar to the tourney that was held thursday near APEX, are a bit different. Top players are trying to save their best for the next days, and/or messing around with secondaries. It is not an accurate representation of some players and as such shouldn't be included.

 

[FoxLisk]

I don't think free/not free is the correct cutoff. I think if you host a tournament because you want some sort of organization for friendlies, it shouldn't be counted, but if you have a free/$1/whatever tournament that is still meaningful - say a PR tournament or something - that is valid.

Anyway, just pushed a big change to how set editing is done. Should be very easy for us to switch winners of sets now. i'll go handle the ones that have been mentioned in the thread right now.

 

[FoxLisk]

If you specify a player skill greater than 10 in Tio, it will seed based on those values. However, if you go to the Players tab and select a player who's skill is set to a number greater than 10, Tio will crash on you.

So I can offer much better seeding (basically I can assign N players skill from 1 to N, in order of their rating) at the cost of you not being able to select them under Players*. That means that you can't change their name, nickname, skill, or location anymore in the seeded version.

How do people feel about this tradeoff?
Should I just make it a separate option? What do people think about this?

*The Events -> Entrants tab is safe, but you can't modify any values there.

 

[ajp_anton]

Year first, really ? Wouldn't you go D, M, Y ? As in today being 06/04/12 ?

That's the way most countries do it, but there are a few, including Sweden, that have YYMMDD as the official order. However it's about as official as the metric system is in the UK, meaning people still use mostly DDMMYY, but on official papers the YYMMDD is used.