ST rankings website

Would be nice to see it with a larger data pool.

Although, Is there even any complete data on the B-series anymore? Heck, is there even any complete data on EVO 2k2 anymore let alone the B-series?

This would’ve been great if we had this from the start.

I think the problems this might face (I’m just posting this to try and find ways to these problems), are the incomplete tournament results. Usually, people list top 3 or top 8, even though there were 20 or 30 people in the tournament. Also, even if people list the whole tournament results (from first place to the person that tied for last), they still don’t quite show who beat who, other than first beat second and second beat third (unless I’m mistaken, this ranking requires that you need to have the stats for all of who beat who, winners and losers, top seed and bottom dweller).

I believe this has a great chance of bringing together international scenes, especially since we are pretty separate, even with the internet. Also, the old data (from 1991-2000) is going to be hard to get, they only list the top 3/8 (You would have to get tournament organizers that were actually at the tournament to remember the people or at least have the brackets, which isn’t likely).

Another question, what tournaments will you be taking into account? Will it be all tournaments? only majors?

Got more questions for later…

Great question. I asked about this and I found results of top 8 of prior evos, as well as the top 64 of another. This leaders to an issue, however, that I will address below.

Here’s the issue with past Evos. What I have is top 8 results (mostly). I can simulate a bracket by doing the following. Let’s say for arguments sake that the top 8 are:

  1. Daigo
  2. Justin Wong
  3. Pony Face
  4. Inkblot
  5. Care Bear
  6. Mike Vick
  7. Donatello
  8. Zangief

I don’t have the full matchups, so what can I do? The algorithm works on a match-by-match basis. Well, one thing I can do is pretend that player 1 beat EVERYONE else, and player 2 beat EVERYONE below him, and etc. I call this the “cascading wins system” So I can enter matches like:

Daigo beats Justin Wong
Daigo beats Pony Face
Daigo beats Inkblot.
… etc …

Then
Justin Wong beats Pony Face
Justin Wong beats Inkblot
Justin Wong beats Care Bear
… etc …

The problem with this is that it totally unfairly rewards Daigo here. In an 8 man tournament, Daigo only had to have 3 wins to make it to the finals. But in this system, he’s being credited with 7 wins! And it’s very unfair to the 7th and 8th place players. Donatello and Zangief are getting hammered by everyone, even Mike Vick.

While it’s possible that Mike Vick beat Zangief, it’s certainly not fair to Zangief to have him record a loss to everyone. He may have had only 1 loss, and that could have been to Daigo!

So I played around with the data, and I don’t like the “cascading wins” system. It leads to really skewed results, waaaaay too favorable to winners, and waaaay too unfavorable to losers.

So another way is what I call “invent the bracket”. We don’t know what the top 8 bracket was, but we know some things. We know Daigo won. We know Zangief took 8th. And we know everyone in between. So we should be able to construct SOME bracket that meets that final top 8 criteria.

In this scenario, we could just plug in a bracket where Inkblot loses to Pony Face, as long as Inkblot and Pony Face take 4th and 3rd respectively. In this system, Daigo gets his 3 wins. They may not have been against the right people EXACTLY, but they were against the right caliber of people. I’ve played around with that, and it works much better than “cascading wins”.

The problem with this is when an angry Inkblot comes out on SRK and posts “WTF?!!?? I never lost to that douchebag pony face! I kicked his ass in casuals. In fact the only reason that bastard placed higher than me is that I lost to Daigo. Yeah I lost to Daigo! I should have taken 2nd. Justin Wong and Pony Face are lucky bastards. This ranking system sucks. You’re just making up data to make me look bad!”.

Now perhaps Inkblot won’t post this. But if he doesn’t, I wouldn’t be suprised if someone else posts something like it. I can explain that the results are more or less the same, and that future performance will ultimately determine his ranking. However, I also know that people can be very attached to their ranking in any game. I want to promote this as a real and true ranking system. So should we try to squeeze those Evo results in? Even if the data isn’t completely correct?

The other option is to not put in those old results, and just take in results that have full brackets. That might be the cleanest approach. But I think it’s terribly unfair to players such as John Choi, and Gian, that have had consistent and high placings in tournaments for years.

So I’m asking the community. What do you think. The options are:

  1. Create brackets that try to fit the results we get. The rankings will on the whole be more correct, but individuals might have slight bumps or dings to their rankings.
  2. Ignore tournament results that don’t have brackets included, and only use future tournament results.

All tournaments. Not just majors. The beauty of this system is that it works with big and small tournaments, linking players by a chain of who has played whom. So if you have an 8 man ranbat, that’s fine! You can win all the small tournaments you want, it won’t inflate your rankings like Apex did.

Number 2 definately.

I mean, yeah, if someone’s got it ALL the Evo results and match details written down somewhere, then great, slap that shit in there.

Even without older tournaments, it’ll quickly become a very useful tool anyway.

Choice “2” is best. You do want the point of this ranking system to be as comprehensive as its accuracy.

Also I know you’ve stated this is shown as “functionality over form”, but if you could change what “mu” and “sigma” are called to something a little more relevant that would help understanding the system as a whole a lot better. That or give each term a little more in-depth explanation.

One of the problems I have with number 2 (even though it is the logical choice) is that it doesn’t give any credit to people that played in the past (not necessarily in the early or mid 90s); then again, if they are still competing, then we can get a good ranking system going. One thing I was hoping for was to have a year-by-year breakdown of the rankings, as well as an overall ranking; it doesn’t seem possible to do the year-by-year ranking (unless tournament organizers can step in), so number 2 is what we have to go with.

I would like to see how guys like Thomas Osaki ranked in ST, although now, I imagine it would be pretty low, due to inactivity. I can help get some international support if you don’t mind (If you want UK SF2 results, you should talk to Reko).

go ruby go!

I do want UK SF2 results. I already have some Japanese results thanks to nohoho.

Zass - Im throwing a midwest tourny in a month. If I scan the finished brackets and send em over to you, will that be added?

I realize this is only for ST for now, but I will keep ALL the brackets for when this is expanded to reach out to (i hope) every fighting game with even a small scene.

This Halo ranking system sounds awesome. It seems much better than the Elo system used for tennis and chess.

Talk to me in PMs keits. I have also written a web interface where you can add your own brackets and players.

there’s some ST OG’s in there that needs to be ranked higher. if you are going to rank somebody especially the OG’s you gotta go way back when ST was at its peak to make an accurate ranking.

When there were rankings here, a big pet peeve of mine was the ‘this person i know is good should be ranked higher’ drama.

In the end, tournaments or should decide who the best is among the good players. What tournaments don’t do a good job of, is differenciating between the bottom 90% of players, or giving them any kind of feedback on how well they are doing compared to the top players.

As for incomplete brackets, why can’t we just list the matches we know and not count the rest? It’s going to be a little off either way, but I don’t like the idea of ignoring valid matches, and I also don’t like the idea of making up data.

This is actually not a bad idea at all. This will work. If you have incomplete brackets for past events, you can send them my way.

Coolness. One question I have, though, is how this system ranks people who’ve never played each other. For example, I’ve won the local MD/VA/DC ST tournaments a few times, but surely I’m not better than, say, K, who shows up only as having lost to Noguchi in a Mikado ranbat, and therefore has a ranking of 0.0. How would your system take into account the fact that I and the people I play with in our tournaments aren’t close to the level of the players in the Mikado ranbat? Would winning a local US tournament be placed on the same level as winning the Mikado ranbat? Would the system place a weight on the tournament itself or on the individual players present?

Also, you should factor in the C3 MD/VA/DC ST results, they’re in the Tournament Results forum.

Edit: Looks like this problem has already come up. For example, GeeseMaster and Gian have comparable rankings (26.43 and 29.29, respectively), and GeeseMaster has 9 fewer recorded matches than Gian. No offense to GeeseMaster, I don’t know him and I’m sure he’s a fine player, but, at the least, Gian should be ranked a heck of a lot higher than anyone in the US who doesn’t have a long and distinguished career of ST domination. The fact that Gian has done well in Mikado and Game Newton ranking battles should mean a lot more than GeeseMaster’s showing at SCFGL and the Winter Showdown. Again, no disrespect to GeeseMaster, the tournaments he’s entered, or the people he’s played against, but it’s objectively true that Gian is way, way better at ST than all but the highest cream of the US ST crop.

I think there needs to be at least a little coaxing of the system to show that certain players who we all know are awesome are ranked well to start with and that winning certain tournaments is more rank-worthy than winning others.

At this early stage the rankings are almost meaningless.

That distinct problem will probably go away once a few more of the past EVO tournaments are added. The system only really needs a few people to cross over to work as a reference points.

For example, Tokido beasted at ST last year. His rank will have gone up, and consequently push up other Japanese players who beat him as and when we add more tournaments. Similarly all of the players who lost to Tokido will have gone down, dragging down anyone who loses to them along with them.

Obviously this needs to happen a number of times before the rift is set in stone. But you get the idea.

I’m all for this, start with whatever info we have and move forward. It will all sort itself out before too long and we’ll have a valuble tool for the community.

hey i know i am not good n all but i feel like i should be on that list i have faced some of those people on that list beating a few of them in tourny or at least giving them a run… i just feel lil left out thats all

Actually, I ask that tournament directors (or an interested player like yourself) pm me, and I can show you how to enter the results yourself. I want to put data entry into the hands of players and TDs, while I focus on development and features.

Julien

well then send zass some tournament data. Saying you should be on the list isn’t going to do anything.