Free Allegiance Forums

Ok as requested here by Pook I'm starting a thread to discuss improvements to the current Newbie Rank Modifier.

The current situation is that each player starts with an ELO of 1500 (to help keep the system close to 0 sum). Newbies then have a modifier on their rank that starts at -15 and reduces by 1 for every full week since their account was created. The flaw with this system is that it affects a newbie who plays one game a month in the exact same way as it affects people who play daily.

There have been some reported cases of people who've played a few games returning to find themselves without newbie protection and rated as Vet 1. This causes problems with the current game balance situation as a 'Vet 1' newbie will skew the team rank total. This will also be a problem once autobalance is introduced.

The old newbie number was based on the unique number of days that that person had logged in for. My suggestion is to combine these two methods and create a newbie modifier that is based on days logged in rather than on time since account creation.

Any opinion on this suggestion Pook/Tiger? Would it be too much trouble? How much work would it be to change the current formula to an equivalent based on unique login days rather than time since account creation?

I'm not trying to piss anyone off about this but it seems to me that my proposed idea would be better than the current arrangement.

Please Note: This is not a thread about autobalance, the merits and accuracy of ELO or the relative worth of different players. Please post about those topics elsewhere.

Awhile ago, unique login days was the best we could do as we didn't have specific game data.

However, we now have specific game information including the number of minutes someone has actually played. Because the "number of minutes played" is not pre-aggregated and I'd rather not do SUM queries across the TeamMember table at every login...

My proposal would be to use the total number of valid (game counted) games they've played where they played for at least half of the game. This information is already aggregated and the performance hit for the query would be almost nothing.

The query itself would look like this:
The question then becomes, what's the rank to gamecount ratio? Do you lose 1 rank-modifier point for every 10 games you play? 1:5? 1:25?
(i.e. at 1:10, you would have no modifier after you'd played 150 games, since the modifier starts at -15)

Cool, I figured on days because I knew it could be done and it would be better than the current situation. Basing it on games would be even better so it's great to hear that it's an option.

Ok, if we want to simulate the effects of the current system then we need to decide how many games a newbie plays in a week. I'd say that 15 games seems about right (2 per day * 7 days + 1 to make a nice round number).

I'd agree with that ratio... Assume 2 games per day.

--TE

Sounds perfect, and it'll help quite a bit once it's implemented in some fashion. Being able to roughly gauge skill for newer players is an important part to getting elo more on track, it'll also make comms more able to be fair and newbs less frightened of over responisiblity. More or less. /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />

75 games to only be considered a (5) seems way too excessive for me. People that play 2 games a day every day of the week end up being much more valuable than their number would indicate.

Ok then, 10:1 it is.

See if it's any better, if not it can be adjusted again.

In the meantime feel free to continue the discussion if you think it should be done differently.

Has this been implemented already?

Just curious as I was a (15) yesterday--got 60 elo from the last 3 games I played--and when I joined today I was a (14)...

I've only started playing again regularly in the last few weeks and to be fair (14) is still an overestimation of my skill level, assuming that elo is a true representation of skill.

I hear 'everything' is going to be reset when R3 comes out... is this true?

Why is the rank shown when it is clear to see a game is stacked or not by looking at the callsigns? Is it just for ennoblement?

Sorry, I've been out of the loop for a while.

So xyz player that joined and played for a week a year ago was a (15), just as someone who's played the last year. Callsigns are definately the best way to balance games, if everyone knows everyone intimately. This is a good change, especially for the newer players, as well as commanders. /smile.gif" style="vertical-align:middle" emoid=":)" border="0" alt="smile.gif" />

Was a Vet 1 on Monday and Tuesday morning I was a Novice(4). By the math used and the logic behind it, I don't see how you can determine a player's value by the frequency by which he plays.

I left this game for 6 months and came back with just a little rust. A couple games later I was in full form. How can you measure experience/capability by the number or frequency of games played? That doesn't make sense. I know guys who play constantly and never get any better. On the other hand, I also know players who rarely play, and they are 10 times better than the average vet.

If a rank can suddenly go from Vet to Novice overnight, then it's obvious that the ranking system bares no reflection of a player's skill. Skill is not "time based". By that logic, fighter pilots in the air force would be senior citizens. Skill is also not a matter of "games played". It's true that a noob will get better the more he plays, however, there will come a time where he will reach his peak performance and plateau as all pilots in the real world do. Skill can ONLY be measured against another's capability. It is relative to each other's skills. If you're going to attempt to rank a player, do it by his true performance. This means evaluating each and every player individually and comparing his capability to the top and bottom player. Outside of that, rank cannot be calculated accurately.

Wasp
AKA GluteusMaximus