Christer
Joined: Aug 02, 2003
|
Does the championship scoring system need to be changed?
If you have any suggested changes I am requiring the following criteria to be fulfilled:
- Cumulative scoring
The site shouldn't have to go through a list of games for each result reported in order to update the championship score. The current system stores 6 integer values for each coach and month (wins, ties, losses, two strength counters and a score for optimization reasons), allowing easy and quick update of scores.
- Should encourage playing against equal or stronger opponents.
Someone who wins all their games playing against opponents with half the team stength as their own should get less points than someone who wins against stronger opponents.
- Should not be favourable to play few or many games
The championship is supposed to be something that a newcomer to FUMBBL can strive for. Also, I don't want to force people to play 100 games per month in order to compete. As a guideline, coaches who play 10 games in a month should have an equal chance of winning as a coach who played 150. This is where the current scoring system is the weakest. |
|
|
Colin
Joined: Aug 02, 2003
|
 Posted:
Aug 07, 2003 - 04:01 |
 
|
There has to be some sort of penalty for a loss to satisfy the 3rd criteria - perhaps you could change the formula to something like:
Score = 10 * min(10, gamesPlayed) * (wins + ties/2 - LOSSES) / gamesPlayed * difficulty
The penalty should be greatest to those who lose to lower rated teams, and least to those who lose to higher rated teams, and vice versa for win bonuses.
Ties could be removed from the equation:
Score = 10 * min(10, gamesPlayed) * (wins - LOSSES) / gamesPlayed * difficulty
A tie to an equal ranked opponent would mean no change to the Championship score, but a tie with a higher rated opponent would still get a slight bonus, and a slight penalty for ties with lower rated opponents.
These formula doesn't account for difficulty of individual matches, it just takes an average. Would I be right in saying that trying to make it more accurate would break the 1st criteria?
My other idea would be to simply add up the total ranking gains or losses accumulated over the time period, with a minimum of 10 matches to satisfy entry. Again, this may break the first criteria. |
_________________
Join The Cult of Tzeentch, mutate randomly! | Hug a newb! Join the Faculty of Academy Instructors! |
|
cataphract
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 04:27 |
|
| Colin wrote: | There has to be some sort of penalty for a loss to satisfy the 3rd criteria - perhaps you could change the formula to something like:
|
there is though... it comes under the games played factor |
_________________
"the eunuch should not take pride in its chastity" |
|
Colin
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 15:36 |
|
| Quote: |
Colin wrote:
There has to be some sort of penalty for a loss to satisfy the 3rd criteria - perhaps you could change the formula to something like:
there is though... it comes under the games played factor
|
I knew I'd miss something. In that case, it's ok as it is. You could always magnify the effect of difficulty. My suggestion for adding up ranking changes still stands, though, as an alternative.
I have a query about the formula, though. To be more accurate with mathematical syntax, should it be written with some extra brackets:
Score = 10 * min(10, gamesPlayed) * {(wins + ties/2) / gamesPlayed} * difficulty ?
As it is currently written, wins and ties are divided by the difficulty, which would penalise coaches playing 'up' instead of reward them. |
_________________
Join The Cult of Tzeentch, mutate randomly! | Hug a newb! Join the Faculty of Academy Instructors! |
|
boobis
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 16:18 |
|
| Colin wrote: |
I have a query about the formula, though. To be more accurate with mathematical syntax, should it be written with some extra brackets:
Score = 10 * min(10, gamesPlayed) * {(wins + ties/2) / gamesPlayed} * difficulty ?
As it is currently written, wins and ties are divided by the difficulty, which would penalise coaches playing 'up' instead of reward them. |
Mathematically speaking, I believe the "slash" used for division here makes the extra parenthesis unnecessary, although I would include them for clarity.
/Joel |
|
|
Christer
Joined: Aug 02, 2003
|
| Quote: | | Score = 10 * min(10, gamesPlayed) * {(wins + ties/2) / gamesPlayed} * difficulty |
Yes, this is the way to interpret the formula. I was under the impression that people read formulas left to right normally (as most calculators and programming languages do). ie: a/b*c is interpreted as (a/b)*c and not a/(b*c).
The problem with the formula is currently that you are being penalised for playing games after getting a good record on your first 10 games as you have much more to lose than to gain from this point.
Come to think of it, there are a few ways to get around this:
1. The score is only updated when you win. Championship score would be a kind of average of the difficulty of the games you have played. This would most likely end up giving lots of coaches with the same score.
2. Update the score for the 10 first games only. Slightly boring as you'd have no chance of improving after the first 10 games.
3. After the 10th game, update the score only if it would increase. Effectively, this would mean that only wins were counted after the first 10 games. |
|
|
Chickenbrain
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 17:02 |
|
4. Only count the 10 best games... |
_________________
Join Themed Blood Bowl for the joy of Themed Teams. |
|
Christer
Joined: Aug 02, 2003
|
| Chickenbrain wrote: | | 4. Only count the 10 best games... |
Well, you see.. Here is where the optmization problems appear.. I'm not particularly keen on keeping a list of 10 games and check/update it after each game...  |
|
|
Chickenbrain
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 18:20 |
|
Pity, but would be the fairest 
Like droping worst and best result on those high diving stuff.
But i understand the problem. A list of 10 results per coach would be a new table entry in the SQL stuff .. not very neat at all. |
_________________
Join Themed Blood Bowl for the joy of Themed Teams. |
|
Grumbledook
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 18:22 |
|
i think its fine as it is |
|
|
Chickenbrain
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 21:27 |
|
Actually i have never looked at it. But well, in RL some systems use the "best x result" approach. That was the reason why i voted for that.
I can live very well with current or any other system.. I don't use that so much. Lack of coaching abilities to be competive in that approach anyway... *grin* |
_________________
Join Themed Blood Bowl for the joy of Themed Teams.
Last edited by Chickenbrain on %b %07, %2003 - %21:%Aug; edited 1 time in total |
|
Emmers
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 21:27 |
|
It seems what we are looking for is quality wins. A quality win would be when you defeat an opponent that is higher ranked than you. And of course losses against weaker opponents would hurt you. So each game should be worth exactly the ratio of the team strenghts of (winner str)/(loser str). The winner would have this score added to his champ rank, and the loser would have it subtracted from his. A draw is worth half that score, with the score added to the weaker team and subtracted from the stronger. Now, for the last point. Only change the score for coaches who have played less than 20 games. Once you reach 20 games, your champ score is set and will not change.
This plan keeps a running total, it does penalize a couch who has not played 20 games, but that number can be changed, however it does nothing at all for players who play more games. Only your first 20 games count. Active coaches can spend the first few days establishing a score, dormant coaches can spend the whole month. It clearly favors quality wins and penalized moronic losses. |
|
|
yavanaril
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 22:17 |
|
Just one from an old boring guy.
I like the idea of putting in some sort of limit on games. And althought I prefer the top 10 games idea, I understand Christers reservations. So it is fine to use the first X games. It's just that for someone like me, who is married, has a job and has other private activities, 20 games per month is still a lot. 15 seems more attainable.
But that's just an old fart complaining about how things were better when he was young.  |
|
|
Mully
Joined: Aug 02, 2003
|
| Posted:
Aug 07, 2003 - 22:34 |
|
I dont see how you can cap the number of games and still call it the "monthly championship". For the active players, what makes the first 15 days of the month anymore important than the last 15. Is an active coach who started the month at 10-10 and finished the month with a 20-0 run supposed to be rated lower than a coach who went 13-7 over the whole month ??? That doesn't make much sense.
The only issue I have with the scoring (July championships) is it seemed to promote beating up on weaker STR teams. A lot of the teams at the top had good records against bad opponents. It does not look that way this month (yet) so I don't know if that's because Christer's already re-jigged the formula or not. |
_________________
Owner of the REAL Larson
Come join the CCC League |
|
Grumbledook
Joined: Aug 02, 2003
|
| Posted:
Aug 08, 2003 - 00:12 |
|
um if you have someone who has won 10 matches against weaker str teams and someone who has 10 wins against equal str teams then the person who beat 10 equal str teams would rank higher
like i said i think its fine as it is |
|
|
|
|
Conceding v Goblins/...
Search
