Warrior General Games

Originally posted by InRomoWeTrust
Yes it does. I don't think it's low (teams are moving, hence Virgins +10 and Queen City +8 in the earlier posts), it's just needing a tweak upwards.

There are dangers with increasing it too much, as the system becomes too volatile, so it's important to work in small adjustments to make sure we hit a sweet spot without a huge guess and check mentality.

How does it become too volatile? It is nowhere near the sweet spot right now when a team gets smashed in every single game on the season as a CPU and is still in the top 20 after 20 games. My own team should have fallen out of the top 20 by now let alone the Generals. Bigger movement is more fun all around and makes every game much more important. It also works for people who move up too fast because they can be pushed right back down quicker as a result of finding a team they can't contend with.

Originally posted by TxSteve
Sounds like we are all kind of agreeing: the k value is too low.




Yes, to a point. Some have advocated we ditch ELO completely or add weird strings to it that don't flow with ELO's autocorrecting nature. That's just an overreaction.

So what is this, the 3rd season The Nightmare has been at the Vet level. I had to rebuild more than half my team this season, I had 2 guys hit Legend about a quarter through and had their chem completely drop off. Yes we cant compete with a 1-3 agent team who has 100 chem and no legend players hanging around. We built a bond with many of the agents on our team and i respect them enough to not cut their players at random to make room for someone else. We still can compete with many of the top 10 teams who have been around for a while. The new age of GLB2 seems to be the 1 or 2 agent teams who will come in with their power run or pass games and take over and we have always expected that. But its not gonna happen in one season. We deserve to be where we are for competing in the top 10 for as long as we have. Once we lose enough I expect to drop out. But until then, let the elo do what it does and move along.

I see what you're saying - but my personal preference would be a little more volatility. We hardly even get any changes in the top 10. Maybe someone drops one spot (it seems).

I'd be all for more volatility - would make it a heck of a lot more fun than the way it is now. Top 10 is just too stable - which makes it boring and uninteresting.

Originally posted by bhall43
How does it become too volatile? It is nowhere near the sweet spot right now when a team gets smashed in every single game on the season as a CPU and is still in the top 20 after 20 games. My own team should have fallen out of the top 20 by now let alone the Generals. Bigger movement is more fun all around and makes every game much more important. It also works for people who move up too fast because they can be pushed right back down quicker as a result of finding a team they can't contend with.

Ladder scheduling*

If we had more teams in the pool, it could work really nicely. Not with a system where we basically play everyone once. Scheduling would have a ton to do with your ladder rank.

Originally posted by InRomoWeTrust
Common sense says it's bad. Biggest danger is the system overcorrects itself too fast. A team on a streak, regardless of elite competition, would be getting more ELO to their name for it. With the way ladder scheduling works, the rankings could get REALLY funky, not resembling something 'accurate'.

In theory, if you really jacked the k-value, you could create a system that reflected a ladder based on like the last 3 games, lol. That's the polar extreme. Now that wouldn't happen, but that gives you an idea of how that could flow.

The realistic side of having a k-value that is too high is that it rewards the streaky team over the team that has consistent success (consistent in the form of 'that season').

Consistent success rewards itself. I don't think the Stunners should be given a pass if they suddenly drop 3 games in a row. The same goes for my team being given its own pass near the top 10 when we should realistically be down around 20 or lower for the season.

Originally posted by TxSteve
I see what you're saying - but my personal preference would be a little more volatility. We hardly even get any changes in the top 10. Maybe someone drops one spot (it seems).

I'd be all for more volatility - would make it a heck of a lot more fun than the way it is now. Top 10 is just too stable - which makes it boring and uninteresting.

Agree with you. Not saying we don't want volatile, just putting caution tape on it. I'd love to have the 'perfect' ELO system in 2 or 3 seasons than 5+ guessed systems that are all imperfect.

Originally posted by InRomoWeTrust
Yes, to a point. Some have advocated we ditch ELO completely or add weird strings to it that don't flow with ELO's autocorrecting nature. That's just an overreaction.

I think we are all saying LET'S DO SOMETHING - BECAUSE THIS SYSTEM IS BROKE.

whether that is tweaking K values (as corndog mentioned previously) - or whether that is taking blowouts and streaks into account like ninja says -- I don't really care -- but we need to try something.

Originally posted by InRomoWeTrust
Ladder scheduling*

If we had more teams in the pool, it could work really nicely. Not with a system where we basically play everyone once. Scheduling would have a ton to do with your ladder rank.

It actually works better in that regard as you wouldn't be seeing CPU teams within your ladder rank for over the extent of an entire season.

Originally posted by jjet5552007
So what is this, the 3rd season The Nightmare has been at the Vet level. I had to rebuild more than half my team this season, I had 2 guys hit Legend about a quarter through and had their chem completely drop off. Yes we cant compete with a 1-3 agent team who has 100 chem and no legend players hanging around. We built a bond with many of the agents on our team and i respect them enough to not cut their players at random to make room for someone else. We still can compete with many of the top 10 teams who have been around for a while. The new age of GLB2 seems to be the 1 or 2 agent teams who will come in with their power run or pass games and take over and we have always expected that. But its not gonna happen in one season. We deserve to be where we are for competing in the top 10 for as long as we have. Once we lose enough I expect to drop out. But until then, let the elo do what it does and move along.

it is great that you want to keep going. no one is saying that you can't or shouldn't. But the ladder is inaccurate. You've lost 38% of your last 116 games. That is not top 10 worthy.

fwiw, they did win Tiger last season (over Stark and Ronin). Just saying they aren't pitiful to at least the start of the season.

Originally posted by bhall43
Consistent success rewards itself. I don't think the Stunners should be given a pass if they suddenly drop 3 games in a row. The same goes for my team being given its own pass near the top 10 when we should realistically be down around 20 or lower for the season.

I agree.

If we go on a losing streak -- or abandon the team - or start coasting - or go full CPU (any or all of those) - then we should be in danger of losing #1.


I'd like to see just a little more (notice LITTLE) volatility a'la the BCS (which sucks) system. In this system - the Virgins could go 32-0 and not even hit top 5. That isn't a good system in my book. Too much favor to the teams who got here first.

They need to be careful...looking at this seasons blocking change they seem to exxeggerate changes a bit...

I mean yes...the current season should count the most but should really count the season! If they mess things up you end up ladder decided too heavily by the last few games. A team going 27-3 deserves to be a head of a team beeing 23-7. yet, if latter team lost most games in the beginning and had a good streak at the end doesnt mean they should be ahead of the team who maybe screwed 3 of their last 5 ladder games.


I kinda agree that teams who get blown out need to fall faster...doesnt matter how hard it is to code. The K value needs some work but not as much as you guys want if you want to see an inactive Top10 team fall out the Top25 in half a season...if that is only done by K value then you see really really unfair standings at the end of the season.

Just my 2 cents on the topic

The ELO system we currently have was supposed to reward keeping your team together and continuing on to keep up rivalries. Though with the way chemistry currently is you can't really expect teams to continually produce high quality squads over and over. If we increased the k value I would also expect a quicker chem pick up in the offseason as I really don't find it fair that recruiting and continuing your team would actually be a huge disadvantage to a team.

http://fivethirtyeight.com/datalab/introducing-nfl-elo-ratings/

Nice article. I know I am getting away from the current formula used but he talks about how his takes into consideration last seasons rankings, margin of victory, consecutive losses/wins, etc.