Warrior General Games

You fucked up my whole revised ranking system when you gakked it up against the Templars in week 6.

Just wanted to let you all know why I haven't published week 9 here or in A5 yet.

If I published what I have right now, the Templars would be in 3rd and the Guardians a DISTANT 5th (behind the Templars and the Nightmare), despite the fact that Italy just beat Salzburg...

BTW - Math nerds feel free to chime in here.

The problem I was trying to solve: How to give more weight to close wins versus higher quality opponents.

The old formula I was using:

MoV/(60-OppR)*(OppW+1)

where MoV = Margin of Victory, OppR = opponent's scouting bar, and OppW = opponent's win total.

I set it up like this to make sure that a win got some kind of points (beating a zero win team would give a multiplier of 1) but that it was tempered by the quality of the opponent (beating the Blizkrieg, for example, divides your margin of victory by 33)

Problem was, if you beat an opponent by 3 points, no matter how good they were, you got shit for points. You had to blow someone out to really affect your ranking.

So I tried this:

(Mov+OppR)/(60-OppR)*OppW

This gives a close win a boost by incorporating that rating a second time as an add-on to the base. It didn't have any significant impact on the blowouts vs bottom tier teams, because the MoV was already higher than the OppR anyway, and their low scouting bars tempered the win. By removing the +1 on the back, I further tempered it by basically giving a team zero points for a win against a winless opponent - you SHOULD win that game by a ridiculous margin, and it SHOULDN'T boost your ranking.

Problem was:

Week 6, the Templars blew apart the Syndicate by 62 points. The Syndicate has a 57 scouting bar. This gave the Templars a total of 238 points for that victory - the biggest point total of any game across the entire division. Boom, my formula just exploded....

The only other way I can see to fix this situation is to substitute (33-OppPR) where OppPR = whatever the opponent's power ranking was the week before as the boost instead of their scouting bar, because the Sooners are the fly in the scouting bar ointment - if we go by scouting bar, the Sooners should be 9-0, just like the Niners and Outlaws in Alpha, but they're not.

If I do that, I'll probably do the inverse on losses - add OppPR to the base margin of defeat, to balance out the point accumulation.

Of course, if I go this direction, you can basically forget about me getting Week 9 published, because I'm going to have to tear the whole thing down and rebuild it week by week, which won't be done until week 10.

Thoughts?

Could you give less weight to MOV and more weight to wins.

Honestly I'd remove the scout bar calculation from the mix and rely more on Strength of Record formulas. I would have to play with the math of course to find the right balance...

Static:
SoR(Strength of Record) = ((# of OPP wins) - (# of OPP losses)) + ((# of OWN wins) - (# of OWN losses))
The SoR would be static as of the time played so you would manually enter this data. The SoR should probably be the record BEFORE the game was played rather than the record after the game was played.

Non-Static:
MoV (Margin of Victory) = (Own Seasonal Average MoV)
OPPMoV (Opponents Margin of Victory) = (OPP Seasonal Average MoV)
The MoV and OPPMoV would continually adjust as the season went on so you would have this set to pull from seasonal data.

Then you take those formulas and:

SoR * (MoV - OPPMoV) = x

This will give more ranking to teams that beat other teams with a strong record and devalue teams that have blow outs against non competing teams. (Now you'll have to further adjust the formula as you use it to make more sense from it as you get real numbers. Something like:

SoR * ((MoV +/- y) - (OPPMoV +/- z)) = x

in order to make the numbers more real.) Doing it this way would take the seasons average points against and for instead of a single games numbers. Making the numbers more fluid instead of jumpy from game to game. Again, without actually playing with the numbers this is just off the top of my head.

EDIT: The only way to use this formula from Game 1 of the season would be to use: (SoR +1) * ...... So that no team has a 0 in the first match up. If you want to share the data collection you've done so far I can help tinker with formulas too.

Originally posted by azfredbird
Could you give less weight to MOV and more weight to wins.



That's what I'm trying to do here - MOV is highly mitigated by the quality of the opponent (scouting bar and total wins) - but the formula as it sits basically kills the team that only wins by 3, regardless of how good the opponent is.

I do feel strongly, though, that if you blow out a quality team, that should have weight over a narrow victory.

Tahl: PM me your e-mail addy and go download OpenOffice.org and I'll send you my spreadsheet, which has all the data I've culled over the course of the season.

nvm

Originally posted by griffin8r
Originally posted by azfredbird

Could you give less weight to MOV and more weight to wins.



That's what I'm trying to do here - MOV is highly mitigated by the quality of the opponent (scouting bar and total wins) - but the formula as it sits basically kills the team that only wins by 3, regardless of how good the opponent is.

I do feel strongly, though, that if you blow out a quality team, that should have weight over a narrow victory.

Tahl: PM me your e-mail addy and go download OpenOffice.org and I'll send you my spreadsheet, which has all the data I've culled over the course of the season.

What about taking in account win percentage. As the season goes on the win percentage will get better for the good teams and worse for the bad teams. Give a bonus to any win against a team with high win percentages.