• Welcome to Smashboards, the world's largest Super Smash Brothers community! Over 250,000 Smash Bros. fans from around the world have come to discuss these great games in over 19 million posts!

    You are currently viewing our boards as a visitor. Click here to sign up right now and start on your path in the Smash community!

The Official Michigan Brawl Power Rankings: Season 11 Images Up!

D

Deleted member

Guest
Ori_bro said:
I've heard Bible Black and Boku no Pico have some of the greatest plots ever imagined.

:phone:
Click to expand...
You heard right.

I happen to be an expert on this subject.


The last time I left the Midwest boards I saw a thread about if people can be asexual or something. That's all I needed.
 
Rabbi Nevins

Rabbi Nevins

Smash Lord
Joined
Aug 11, 2009
Messages
1,373
Location
East Lansing, MI
aisight said:
For anyone still interested in experimenting with a mathematical player rankings system, I wrote a program that makes it really easy for you to do so.
http://allisbrawl.com/blogpost.aspx?id=143935
Click to expand...
This is awesome thanks ^_^

If nobody is willing to then I could gather up what data I can and put it in. I'm curious to see what the ranking from last season would be. However if someone on the PR panel would help that would be good, considering you all have already (presumably) gathered all the win/loss data

:phone:
 
Zankoku

Zankoku

Never Knows Best
Administrator
BRoomer
Joined
Nov 8, 2006
Messages
22,906
Location
Milpitas, CA
NNID
SSBM_PLAYER
Keep in mind that you'll want to do things in chronological order - enter in the brackets one full round at a time, winners' brackets followed by losers' brackets, then the grand finals for last.
 
zyth

zyth

Smash Lord
Joined
Jul 16, 2008
Messages
1,474
Location
Southfield, MI
wow that looks like a promising program. downloaded it and did some dummy results and liked what i saw. this should be used for regional rankings, or state
 
Rabbi Nevins

Rabbi Nevins

Smash Lord
Joined
Aug 11, 2009
Messages
1,373
Location
East Lansing, MI
I don't think this could be used in its current state for official rankings. FZk pointed out to me that it would get very complicated dealing with out of state players at local tournaments. You'd have to do the entire region and possible more to get a full sample. Like say I beat MJG (as expected) and he has no record of results in the program for Michigan, to the program it thinks mjg is a scrub because he has no previous results. So basically we would have to only consider local results and no matches against anyone outside of our state. Otherwise we would have to include all of that persons statistics as well

:phone:
 
zyth

zyth

Smash Lord
Joined
Jul 16, 2008
Messages
1,474
Location
Southfield, MI
hmm yea i see what your sayin, we could just do this for state tho just to see how ppl in our state compare to others
 
Zankoku

Zankoku

Never Knows Best
Administrator
BRoomer
Joined
Nov 8, 2006
Messages
22,906
Location
Milpitas, CA
NNID
SSBM_PLAYER
Rabbi Nevins said:
I don't think this could be used in its current state for official rankings. FZk pointed out to me that it would get very complicated dealing with out of state players at local tournaments. You'd have to do the entire region and possible more to get a full sample. Like say I beat MJG (as expected) and he has no record of results in the program for Michigan, to the program it thinks mjg is a scrub because he has no previous results. So basically we would have to only consider local results and no matches against anyone outside of our state. Otherwise we would have to include all of that persons statistics as well

:phone:
Click to expand...
First, here's a link to the Glicko rating system as described by Wikipedia.
http://en.wikipedia.org/wiki/Glicko_rating_system

Now that I'm feeling less dead, I'll address your totally valid concern.

Yes, the first time anyone is entered into the system, there will be no real data on exactly how good he is. I wouldn't say that puts him at the level of a scrub, though... by default, at a mu of 1500 and a sigma of 500, the best the system knows is that MJG's skill level has a 99% chance of lying somewhere between 0 and 3000. The "score" you see displayed on the leaderboards is actually the calculated conservative estimate of the player's lowest possible skill level, with 99% confidence.

Also, your new score is based on a combination of of your current mu, your opponent's mu, and your current sigma - greater difference in mu leads to greater change, and lower sigma leads to less change. Since nobody's sigma will go higher than the starting default, MJG's 500 sigma means his score will change in a single tournament faster than most others - I should know this, I used to keep track of both Vegas and SoCal tournaments, and though this, TKD immediately started from a conservative estimate of 0 to being top 10 in a single regional tournament.

The other thing you should note is that any mathematical ratings system will have the same initial issue with "new" strong players entering into the "local" system. The idea is that anyone new will have a fairly "average" starting mu and a ginormous sigma that will quickly shift their mu toward where they actually belong.

Regardless, I agree that the only way you can ever extend the scope of the ratings system beyond Michigan tournaments is by calculating a much larger range of tournaments, though it of course takes much more work. This is the same reason I chose to limit my recording to SoCal and Las Vegas at the time; I had my locals' results, and SoCal alone didn't add too many more tournaments to process in parallel, plus we went to a regional over there every two months or so.
 
Rabbi Nevins

Rabbi Nevins

Smash Lord
Joined
Aug 11, 2009
Messages
1,373
Location
East Lansing, MI
aisight said:
First, here's a link to the Glicko rating system as described by Wikipedia.
http://en.wikipedia.org/wiki/Glicko_rating_system

Now that I'm feeling less dead, I'll address your totally valid concern.

Yes, the first time anyone is entered into the system, there will be no real data on exactly how good he is. I wouldn't say that puts him at the level of a scrub, though... by default, at a mu of 1500 and a sigma of 500, the best the system knows is that MJG's skill level has a 99% chance of lying somewhere between 0 and 3000. The "score" you see displayed on the leaderboards is actually the calculated conservative estimate of the player's lowest possible skill level, with 99% confidence.

Also, your new score is based on a combination of of your current mu, your opponent's mu, and your current sigma - greater difference in mu leads to greater change, and lower sigma leads to less change. Since nobody's sigma will go higher than the starting default, MJG's 500 sigma means his score will change in a single tournament faster than most others - I should know this, I used to keep track of both Vegas and SoCal tournaments, and though this, TKD immediately started from a conservative estimate of 0 to being top 10 in a single regional tournament.

The other thing you should note is that any mathematical ratings system will have the same initial issue with "new" strong players entering into the "local" system. The idea is that anyone new will have a fairly "average" starting mu and a ginormous sigma that will quickly shift their mu toward where they actually belong.

Regardless, I agree that the only way you can ever extend the scope of the ratings system beyond Michigan tournaments is by calculating a much larger range of tournaments, though it of course takes much more work. This is the same reason I chose to limit my recording to SoCal and Las Vegas at the time; I had my locals' results, and SoCal alone didn't add too many more tournaments to process in parallel, plus we went to a regional over there every two months or so.
Click to expand...
Yeah I realized that I was flawed in my thinking shortly after I posted. I figured the player would obviously grow faster for less tournaments but more victories. Im reading through the wiki atm. Im excited to see how these results turn out, seems like a very interesting algorithm. It would be interesting to see a more formal outline of the TrueSkill algorithm as well.
 
S

Sailor Princess Twenty-Two.

Smash Master
Joined
Mar 2, 2011
Messages
3,114
Location
Ann Arbor, MI
zyth said:
Gio and tutu only remove me from the chat for their own amusement
Click to expand...
LMAOOOOO

Kicking zyth is a lot of fun.

I need to be added back into the group, I left it cause it was legit ****** my phone with like 1000 messages at a time. Also I'm the main reason why that group is active, if I hadn't added everyone i ever felt like talking to on every passing whim I had to do so, it would be a very boring chat indeed.

Also **** L_Cancel for somehow being invincible in that chat even tho I made it.
 
You must log in or register to reply here.
Top Bottom