Would make "Newb stack" something to be feared of /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />Ksero wrote:QUOTE (Ksero @ Apr 13 2007, 07:14 PM) Would be even better... Imagine, commanders might start to yell for (0)'s to come and stack their teams /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
Autobalance bugs?
I have heard lots of arguments FOR the system Pook proposes and not a single arugment against it. So why is it not implemented? If these lobbyists are out there perhaps they should restate their case with sound reasoning as to why it shouldn't be implemented. If autobablance = max team imbalance = N/A is the only reason that isn't a very good argument. Just don't stack and it won't be a problem. Commander modifiers would be the next step. After that ranks could actually mean something.
What Allegiance needs is a little more cowbell. 
This should be implemented so that there is no restirction on the imbalance. It is not a problem, except if you are seriously trying to stack a team. There was a complaint about a 7vs 12 stack -- shame on the folks who joined the team with only 7 folks. Imagine the stack in a game like that.
Now, for why it has probably NOT been implemented (and Pook, please correct, subtract, or add to these reasons):
1) I belive the Senate has to vote on something like this. I don't know and I can't find what the vote was, but I am guessing that the mods were a part of the approved vote, no?
2) Core host(s) have objected to hosting an auto-balanced core, correct?
3) Vehement objections of some of the long-time vets (don't want to list all of the objections or all of the objectors here, but it is at least a significant majority of players).
Now, for why it has probably NOT been implemented (and Pook, please correct, subtract, or add to these reasons):
1) I belive the Senate has to vote on something like this. I don't know and I can't find what the vote was, but I am guessing that the mods were a part of the approved vote, no?
2) Core host(s) have objected to hosting an auto-balanced core, correct?
3) Vehement objections of some of the long-time vets (don't want to list all of the objections or all of the objectors here, but it is at least a significant majority of players).
Robin: "Gosh, Batman, this camel grass juice is great."
Batman: "Beware of strong stimulants, Robin."
I'll let Tiger respond to why some things were coded the way they were.
I do think the Newb Stack prevention code is problematic... and that we need a fix for. It can lock out players of up to rank 8 from both sides.
As for the threshold I can see why Tiger put it in as it allows a low ranked player to join the team with the higher total as long as the teams are not too unbalanced. This allows for some flexibility that a more rigid system (like always sticking the new player on the side with the lowest total) would not.
I won't list individual ranks, but say I am rank 5, there are ~15 per side AND the teams are fairly balanced, but I prefer the players/faction/commander on the side that has the slightly higher total. The threshold method allows me to join EITHER side. This I think is a good thing.
I also put in my two cents AGAINST having 0's count as a 5. My reasoning?
Team A: 4 4 4 5 5 10 10
Team B: 0 0 1 1 3 10 10
These teams just aren't equal. A 4 or 5 just isn't the same as a 0 or 1, warm body or not. I'd argue a 4/5 is almost 4 times as useful than a 0/1 that is still trying to figure out when people yell 'Defend the EXP', what the hell an EXP is, where is it, and how to defend it. I'd be okay with a floor of 2 however.
Personally I like the idea of auto balance allowing things like
Team A: 0 0 1 1 2 3 3 4 4 5 5 5
Team B: 9 10 10 10
I do think the Newb Stack prevention code is problematic... and that we need a fix for. It can lock out players of up to rank 8 from both sides.
As for the threshold I can see why Tiger put it in as it allows a low ranked player to join the team with the higher total as long as the teams are not too unbalanced. This allows for some flexibility that a more rigid system (like always sticking the new player on the side with the lowest total) would not.
I won't list individual ranks, but say I am rank 5, there are ~15 per side AND the teams are fairly balanced, but I prefer the players/faction/commander on the side that has the slightly higher total. The threshold method allows me to join EITHER side. This I think is a good thing.
I also put in my two cents AGAINST having 0's count as a 5. My reasoning?
Team A: 4 4 4 5 5 10 10
Team B: 0 0 1 1 3 10 10
These teams just aren't equal. A 4 or 5 just isn't the same as a 0 or 1, warm body or not. I'd argue a 4/5 is almost 4 times as useful than a 0/1 that is still trying to figure out when people yell 'Defend the EXP', what the hell an EXP is, where is it, and how to defend it. I'd be okay with a floor of 2 however.
Personally I like the idea of auto balance allowing things like
Team A: 0 0 1 1 2 3 3 4 4 5 5 5
Team B: 9 10 10 10
DOG PROPERTY LAWS:
2. If it's in my mouth, it's mine.
[unless it tastes bad, then it is yours.]
I disagree here.
I don't think the threshold is needed - I think that it's one way of trying to appease the folks that say things like "I won't fly for a $#@!tard". One benefit of getting rid of it - it makes the system more simple, more transparent, and less "WTF" prone.
As for the 0's counting as 5... the average player takes only around 20 games to reach a rank of 5. I think it's a good compromise, and I doubt that a 4/5 is that much more useful. At best, they've figured out the keymappings and perhaps know that they prefer one type of ship to the other... but a novice is a novice.
Bottom line I think is that we tried it the other way and it's gotten to the point that nobody will use it, and I'd like to try something else. Hell, if I can scrap ELO in search of a simpler and more straightforward solution I think we can do the same with autobalance.
I don't think the threshold is needed - I think that it's one way of trying to appease the folks that say things like "I won't fly for a $#@!tard". One benefit of getting rid of it - it makes the system more simple, more transparent, and less "WTF" prone.
As for the 0's counting as 5... the average player takes only around 20 games to reach a rank of 5. I think it's a good compromise, and I doubt that a 4/5 is that much more useful. At best, they've figured out the keymappings and perhaps know that they prefer one type of ship to the other... but a novice is a novice.
Bottom line I think is that we tried it the other way and it's gotten to the point that nobody will use it, and I'd like to try something else. Hell, if I can scrap ELO in search of a simpler and more straightforward solution I think we can do the same with autobalance.
A threshhold is needed, I personally don't want to sit on noat until I can join yellow, I rarely play blu and would rather not be foreced to. /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />
I would also say that players of rank <5 shouldn't equal 5. I definately see that issue, I'd be more in favour of something like <2 = 2... and even that's a stretch, I'm not sure what's perfect though I do see issues. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
I would also say that players of rank <5 shouldn't equal 5. I definately see that issue, I'd be more in favour of something like <2 = 2... and even that's a stretch, I'm not sure what's perfect though I do see issues. /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
n.b. I may not see a forum post replied to me or a pm sent to me for weeks and weeks...
Disagreement is healthy /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />
But I do agree that people don't use it because they 'heard' it didn't work a few times.
I also agree with Pook about the WTF factor, if they see people joining sides when they would think they shouldn't they will think its broken.
That said, I am still in favor of something a little more complex than the new player always goes to the side with the lower rank total. Although that may be okay for players joining after a game starts. We need something a little more complex for the initial 'balancing' of everyone in NOAT that is not AFK, but I think that part of the code is working just fine.
But I do agree that people don't use it because they 'heard' it didn't work a few times.
I also agree with Pook about the WTF factor, if they see people joining sides when they would think they shouldn't they will think its broken.
That said, I am still in favor of something a little more complex than the new player always goes to the side with the lower rank total. Although that may be okay for players joining after a game starts. We need something a little more complex for the initial 'balancing' of everyone in NOAT that is not AFK, but I think that part of the code is working just fine.
DOG PROPERTY LAWS:
2. If it's in my mouth, it's mine.
[unless it tastes bad, then it is yours.]
Ok, more complex in what way? Is this about the difference between three (0) vs three (5) - Should it try and even those out somehow?
What factors do you think make a good autobalancer, i.e. what does it need to do besides what I've suggested which is a simple (yet imo effective /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />) method?
Once we've laid out the "what", we can work on the "how" /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
What factors do you think make a good autobalancer, i.e. what does it need to do besides what I've suggested which is a simple (yet imo effective /wink.gif" style="vertical-align:middle" emoid=";)" border="0" alt="wink.gif" />) method?
Once we've laid out the "what", we can work on the "how" /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
Last edited by Pook on Sat Apr 14, 2007 4:12 am, edited 1 time in total.
Unless I missed previous posts, I dont understand one thing concerning the team rank sums :
it's pretty much impossible to tell that a high ranked player "will be better" than another high ranked one, no?
So basically, it should not be linear (even with your min/max 'cuts') when adding ranks for balance but rather asymptotic (BalanceValue = a + b/Rank instead of BalanceValue = Rank like now).
With well tuned constants (a & b), a 20 and 25 will have almost same balance values (and not a +5 diff) and a 5 and a 10 will have almost a linear diff (close de +5 or may be more). The key is how to 'shape' the asymptotic curve so that the balance is balanced /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
So unless the ranking system is complety redesigned (I guess that's another debate /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" /> ), using a linear rank->value for balance system would never be accurate because the ranking system itself is not linear. no ?
it's pretty much impossible to tell that a high ranked player "will be better" than another high ranked one, no?
So basically, it should not be linear (even with your min/max 'cuts') when adding ranks for balance but rather asymptotic (BalanceValue = a + b/Rank instead of BalanceValue = Rank like now).
With well tuned constants (a & b), a 20 and 25 will have almost same balance values (and not a +5 diff) and a 5 and a 10 will have almost a linear diff (close de +5 or may be more). The key is how to 'shape' the asymptotic curve so that the balance is balanced /mrgreen.gif" style="vertical-align:middle" emoid=":D" border="0" alt="mrgreen.gif" />
So unless the ranking system is complety redesigned (I guess that's another debate /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" /> ), using a linear rank->value for balance system would never be accurate because the ranking system itself is not linear. no ?
