|
Post by mijapi300 on Jul 10, 2017 16:50:26 GMT -5
I've never liked the current point distribution, now or when they first used it. I felt like it artificially inflates people's ratings, forever pushing them higher in league. I decided to do a statistical analysis in Excel to "prove" my theory.
I was surprised by the results. First, I created a table of the possible league points: +13 +10 +7 +4 +3 +2 +7 +3 -5 -11 -17 -22 +3 (top beacons) +2 (top beacons) +1 (top beacons)
The net of all of these distributions is 0. Meaning every point that is earned is lost by someone else. I'm more interested in the distribution of these points. I believed that this distributions caused the majority of players to climb continuously, thus creating a vaccuum between the top leagues and bottom leagues.
The next step was creating a randomized sample of players. To do this, I created a player table, with 1000 rows. So our random sample is 1000 players.
I then created our randomized "outcomes". Each cell contained two RANDBETWEEN formulas, with values of 1-12. The first determined their damage rank, and the second determined their beacon rank. Used a VLOOKUP to pull the appropriate point distribution from the table. Dragged this down to all 1000 players, and then over to 3000 games. Three million data points.
I made a macro button that would just alter a blank cell whenever pushed, which re-randomizes the data. I created a table that counted how many players had a positive overall distribution and how many had a negative. I also ranked the overall points from high to low.
The first randomization produced 480 positive, 1 zero, and 519 negatives. The spread from top to bottom was very consistent and gradual. I was surprised. So I pressed the button. Over and over. It was evenly distributed every time. I plan on running some more analysis, using different win ratios and whatnot, and will update this thread when I do.
Just though the results were somewhat interesting, and have changed my mind so far.
|
|
|
Post by DarkVagabond on Jul 10, 2017 16:56:52 GMT -5
It's nice to see someone test their long held assumptions. I applaud you for this, especially your willingness to consider the possibility that your assumption may have been incorrect. I think we can all learn from this, myself included.
|
|
|
Post by Dejnov on Jul 10, 2017 16:57:22 GMT -5
Can you publish a graph showing the distribution in points per player.
Dejnov.
|
|
|
Post by mijapi300 on Jul 10, 2017 19:12:23 GMT -5
Can you publish a graph showing the distribution in points per player. Dejnov. I will this weekend when I am at my home computer.
|
|
Deleted
Deleted Member
Posts: 0
Karma:
|
Post by Deleted on Jul 10, 2017 21:35:33 GMT -5
No, the league point distribution will create stars and superstars, meaning players that do not advance and are stuck in lower tiers. Where do you think the 13K points of Legend players come from? They aren't created from thin air. It's like if Kasporav wasn't happy with his 3000 rating and went down into a junior high chess tourny and clubbed all the students there so he could go up to 10,000.
Now imagine not just one top tier playing engaging in this, but all the Legend players join in on the clubbing. That's a lot of points stolen, so this causes them (the clubee or seal) to be at lower tiers since you know, they have less points than they should. This creates the Wall that exists between Expert and Diamond since that's when clubbing begins (iOS).
|
|
|
Post by dreamslayer28 on Jul 10, 2017 22:08:20 GMT -5
48-52 and anything that does not digress by more than three points seem to be a sweet spot.
The real question is what type of hangar and what type of in-game skills should one have in theory in order to break through the Wall.
To give a comparison, in CS:GO, one is considered average when one reaches Gold Nova 2 (that's 8th out of 18 ranks). Anything above that is above average - and one is considered very good at the game when one reaches DMG (that's 14th step, highest is GE at 18th league...not counting pro players who play good regardless of whether they play in MM or not).
|
|
Deleted
Deleted Member
Posts: 0
Karma:
|
Post by Deleted on Jul 10, 2017 22:43:37 GMT -5
Leagues below silver have different values for win/loss, with no negative points on loss iirc. Hence why people that haven't even played a full week can run into plasma Griffins, stalkers, etc.
Your chart doesn't account for the league points under silver being created from thin air, which is the component that destroys the mathematical equilibrium. The 'vacuum' between high and low is no more than the beginner leagues forcing progress far faster than upgrade times and resource gathering deem fair. Destriers and Cossacks can see plasma Griffins, rhinos and Fujins before the Player can amass more than 2 levels for any particular bot and/or weapon. This also plays a part in Griffin/Leo mania in lower tiers, because these bots have no downsides compared to leveling Ag mediums to oh 4-5 to have them serviceable (aka purchase and 2-3 days vs spend 3x The Ag for an instant drop in with 4 weapons.
Without hard limits however, there is no fixing this MM.
90% of the broken scores in Legends league are thanks to the LQ and account sharing.
|
|
|
Post by drake1588 on Jul 10, 2017 22:45:20 GMT -5
The real world results have mostly been positive in the short term. What I'm concerned about is the point vacuum at the legend level and the point generation at private/recruit. ELO systems can get really screwed up when points are not static per player.
|
|
|
Post by Firebeard on Jul 11, 2017 7:19:14 GMT -5
No, the league point distribution will create stars and superstars, meaning players that do not advance and are stuck in lower tiers. Where do you think the 13K points of Legend players come from? They aren't created from thin air. It's like if Kasporav wasn't happy with his 3000 rating and went down into a junior high chess tourny and clubbed all the students there so he could go up to 10,000. Now imagine not just one top tier playing engaging in this, but all the Legend players join in on the clubbing. That's a lot of points stolen, so this causes them (the clubee or seal) to be at lower tiers since you know, they have less points than they should. This creates the Wall that exists between Expert and Diamond since that's when clubbing begins (iOS). Exactly. That is the difference between Theory (Paper) and Practical Application (Real World). Well said.
|
|
|
Post by seanh on Jul 11, 2017 7:38:12 GMT -5
Firstly well done, I like this, its great to see someone spending that much time and effort trying to help us all work out the system we are all participating in.
Is the issue of ‘unfairness’ not one of perception, relativities and peoples rationalisation of their abilities? I am not much good with the fancy numbers and clever thinking and stuff (not entirely true, but the numbers are only ever half a story), but I would any ranking system with a set population, zero sum points distribution and a somewhat fair allocation method would do this relatively well. As the population is not strictly set in reality large sways of new players may impact the distribution, but in general it has to be ‘fair’.
The bit that causes consternation is the its not ‘fair’ for me, element. I have a predetermined idea of where I am / should be, unless I am the one guy who had zero movement I am going to go either up or down. I am going to have an emotional reaction to that, rationalise it, reset my expectations and repeat that whole process with just about every game. It is inevitable that at some point every single one of us will feel hard done by. We get promoted to the level of our ability / hanger and then get knocked up and down in waves.
The bit I do think is not ideal is the partitions between leagues, it appears the MM selects you first by reference to your league then by points, with certain leagues more prone to grouping (I don’t know this for a fact just going from personal experience). Meaning there is a point where every player has to take a big step up. That jump can be quite harsh and I would think may the reason for much of the consternation about the system. I would think if the machine just pulled the 11 available players with the nearest league points to you (with no reference to leagues in that selection) then matches would be fairer.
Anyway, I could have capped 10 beacons in the time it took me to type this, I am going back to grinding for a sexy carnage.
|
|
|
Post by dangit on Jul 11, 2017 7:49:47 GMT -5
mijapi300 like what you did there. But honestly I need to take you fishing with me. You need to get out for awhile. ? ?
|
|
|
Post by llama4president on Jul 11, 2017 8:18:10 GMT -5
Good Job mijapi300You need to get out for awhile. ? ? What does mean the last part? he needs to go on a boat and fish using a Girl as bait?
|
|
|
Post by dangit on Jul 11, 2017 8:23:59 GMT -5
llama4president hmmm, had to put my specs on. No that is an emoji of a fishing rod with a fish on it. Or least Apples representation of it. Fish with a girl for bait. Too funny.? I think that would be against the law. sent pm, sorry for hijack mijapi300
|
|
|
Post by llama4president on Jul 11, 2017 8:28:59 GMT -5
llama4president hmmm, had to put my specs on. No that is an emoji of a fishing rod with a fish on it. Or least Apples representation of it. Fish with a girl for bait. Too funny.? I think that would be against the law. i see a fish too, it's the 2nd symbol i was referring to Boat - Female - Fishing ? ?
|
|
|
Post by Deadeye on Jul 11, 2017 8:59:01 GMT -5
Good work mijapi300. You definitely lived up to your signature. Thanks for the input.
|
|
|
Post by every1jockzjay on Jul 11, 2017 9:17:42 GMT -5
I consider myself very good at this game. With old distribution i was stuck in e2-e3 bc of 50% win ratio. No matter how hard i tried gaining ground was an uphill battle. New distibution im pushing champion and dont see signs of slowing down. I have a 9/10 hangar with meta bots and play for damage andddd wins. Imo this distrubution is 100% better. Still have 50% average wins but it doesnt matter like it used to and makes it much more fun to try to up my league. Happy to hear people like it lol
|
|
|
Post by mijapi300 on Jul 11, 2017 10:41:52 GMT -5
dreamslayer28 I noticed you had mentioned this thread in one of your own. I wanted to update you on the status of it: Just want to add that my post isn't 100% complete. As I knew, and people pointed out - it doesn't account for the upwardly skewed points distribution in Silver league and below, and it doesn't account for differing win ratios. I'm going to add those aspects into the formula, so it'll hopefully give a more accurate representation of true spreads within the game.
|
|