Author Topic: Possible Overall Rankings ideas (Read 9671 times)

douglas · « **on:** April 24, 2006, 07:29:04 am »

On a slightly random note, with the current points system, would it not be possible to see who had the lowest score overall? Like a TSC Grand Champion *shot*

Obviously games with less competitors/stages would be weighted less so it wouldn't be reflective of people's different skills, but I reckon it'd be interesting anyway.

Rolken · « **Reply #1 on:** April 24, 2006, 03:32:09 pm »

It would be possible. I've held off for the reason you cited; it'd be difficult to make one that ranked people accurately.

Shadow Wedge · « **Reply #2 on:** April 24, 2006, 06:28:40 pm »

Well, you could use the low scores to create high scores:
If you take a person's score in a game and subtract it from the highest possible score, you'd get some really big number for the very good people, and a lower number for the less good people. If you add all of these up, you'd get a good score weighting. This has a drawback though: a game with a lot of charts will naturally have a higher worst possible score than a game with fewer charts. So a possible solution to that is perhaps dividing the obtained score in a game by the number of charts.

douglas · « **Reply #3 on:** April 24, 2006, 07:06:02 pm »

You mean add up a player's percentage ratings for each game? That's fairer on a per-game basis, but it takes more effort to get highly ranked in a game with loads of charts and that wouldn't be reflected under that scoring system. Perhaps a fairer basis for comparison would be between games of the same type; maybe a 16 bit champion, an 8-bit champion, a GBA/DS champion and a 3D champion? Then weight the "TSC Championship" on those categories.

I doubt there's a "fair" way to do it that'd suit everyone; it was just a random thought.

Shadow Wedge · « **Reply #4 on:** April 27, 2006, 11:37:22 am »

Quote from: douglas on April 24, 2006, 07:06:02 pm

You mean add up a player's percentage ratings for each game?

It's not exactly adding up the percentage ratings. To get the percentage, you divide by the total number of points. What I'm suggesting is dividing by the total number of charts. So if there are say 5 charts with 5 records on each of them, and I have 5 points, then there's a max of 25 points. So my percentage would be 80%, but the number added to the overall score would be (25 - 5) / 5 = 4.

For a better example, let's compare the top people on SPA and SA2(B):
On SPA, Rolken is #1 with 38 out of a possible 163 points. There are 51 charts. This would make Rolken's overall score addition (163 - 38) / 51 = 2.45. Now, this may seem rather low, but there's only an average of 3.20 submission per chart, whereas in my first example there are 5 submissions per chart (wow, cyberscore much eh?).

In SA2(B), SuperSonic101 is #1 with 928 out of a possible 11874 points. There are ~~a shitload of~~ 286 charts: 131 from times, 7 from races, 22 for bosses, yay we ignore freestyle :D, 64 for rings, 62 for score. This makes SS101's overall score contribution (11874 - 928) / 286 = 38.27. Now, this is because there are 41.52 records per chart.

Actually, the maximum number of points you can get is (number of records - 0) / (number of charts) = average number of records per chart. So the higher the average number of records for the game is, the better your score can be. And your score is based on the number of points you have, which is based on how good your records are.

Now, just to compare: the maximum number of points you can have to get the same overall score in SA2(B) as Rolken has from SPA is 11173, which is 5.9%. So it's up to Rolken to see if it's fair to have that much of a discrepency between two games, one of which is vastly unused and the other heavily used.

Edit: It can easily be calculated by: percentage * max points / number of charts, which makes it a lot easier/cheaper to calculate.

yse · « **Reply #5 on:** May 14, 2006, 03:39:51 am »

Per recent discussion in the chat, I like SWedge's idea as mentioned two posts above.

However, SA2B being worth 16 times SPA is probably a bit much.

My idea (using raw point scores) works SA2B to be about 60 times that of SPA. So that's out of the equation (no pun intended).

My idea for a tweak is so simple I'm surprised SWedge didn't come up with it. A simple square root will reduce the weighting from 16:1 to 4:1.

From there, we are discussing the possibility of the equation being thus:

(percentage) * (sqrt)[(maxpoints) / (divisions)]

This leaves percentage alone (as grouping percentages closer was not the intention), but brings all the games closer together, making a fairer system.

My only concern is that it would be rather difficult to manually calculate, if a player was aiming for a specific position on the overall chart, for example. However, it can be summed up thus:

> percentage = good
> max points = good (ie. if crappy people submit, your score goes up)

Opinions?

EDIT: Something I forgot about last night.

www.soniccenter.org/mike89/overallrankingstest.xls - I plugged all the figures we had into the formula, and per Aitamen's request put a few players over to the side as demonstrations. Note, however, that when you change the criteria to M/D, instead of (sqrt) M/D, the percentage barely changes...

yse · « **Reply #6 on:** June 07, 2006, 02:20:04 am »

Okay, so for all my tweaking in the past couple of weeks, I've come up with nothing that betters this scenario.

Anyone willing to put forth an improvement, or god forbid, a better system? X)

yse · « **Reply #7 on:** March 08, 2007, 04:58:42 am »

Updated the above link. For the record, the new equation stands currently as follows:

As always each game's contribution factor is determined by performance x popularity:
* Performance - 8p^2 +2p, where p = your percentage expressed between 0 and 1. This formula yields values between 0 and 10. An additional 5 point bonus is given to the champion of any given game (shared between multiple).
* Popularity - currently expressed as ln(maxpoints/10), which makes the ratio of the highest to lowest ranked games approximately 4.1:1. Is this a fair difference between Sonic Adventure 2, which takes months to get to a high level at, and Sonic Jam which could probably be wiped out in an afternoon?

Note also that the figures on the left still date to last May when I first did the calculations so comparisons with today probably aren't fair, but feel free to try anyway.

EDIT: Rankings are up!

Current formula is p^2 * M/D, which seems to give pretty fair results.

KnucklesSonic8 · « **Reply #8 on:** March 09, 2007, 03:10:54 pm »

SWEET! CONGRATS ON GETTING YOUR IDEA FINALIZED :) (PC ACTING UP SO THIS MESSAGE WILL APPEAR IN ALL CAPS)

ONE IDEA< THOUGH< WOULD BE MAYBE WE CAN SEE HOW EACH MEMBER"S TOTALS ARE CALCULATED FOR THE SITEWIDE THING

OTHERWISE ITS A NICE ADDITION TO TSC :)

magnum12 · « **Reply #9 on:** March 09, 2007, 03:26:36 pm »

If what I think is true, then an overall top 10 might be based off of this. If things go right for me, I might actually make the list.

stanski · « **Reply #10 on:** March 09, 2007, 05:54:42 pm »

Mike creates a formula then puts himself in first... I CALL BS!

yse · « **Reply #11 on:** March 09, 2007, 06:54:00 pm »

Stanski, there is no reasonable formula that -wouldn't- put myself in first. Have you seen my member page lately?

douglas · « **Reply #12 on:** March 10, 2007, 01:31:22 pm »

I have one:

if(member.name=="mike89") member.position = last;
else { normal formula }

. . . . oh wait, reasonable?

Bilan · « **Reply #13 on:** March 10, 2007, 05:44:51 pm »

That is completely reasonable >__>

stanski · « **Reply #14 on:** March 11, 2007, 01:08:32 am »

mike, i see, you have sweet sonic 3d stats and your knuckles chaotix stats kick ass :-P

Possible Overall Rankings ideas

The Sonic Center > Forum > Sonic Central > Competition Central > Possible Overall Rankings ideas