Bad Mm Or Just Statistics
#1
Posted 29 September 2016 - 10:48 AM
Odds for that are 1:65536, that sound like a lot player playing a lot of games or did PGI rig the system to try to turn me into a whale? What is your take on that?
#2
Posted 29 September 2016 - 10:50 AM
#3
Posted 29 September 2016 - 10:51 AM
#4
Posted 29 September 2016 - 10:54 AM
Edited by The Nerf Bat, 29 September 2016 - 10:55 AM.
#6
Posted 29 September 2016 - 12:13 PM
Jehofi, on 29 September 2016 - 10:48 AM, said:
Odds for that are 1:65536, that sound like a lot player playing a lot of games or did PGI rig the system to try to turn me into a whale? What is your take on that?
On group-q? Then 16 win streak is normal when your group is good. On solo-q 16 win streak is very lucky, though personal performance also affects it. Either way, enjoy it.
#7
Posted 29 September 2016 - 12:22 PM
that is if the MM is working.
What Tier are you?
Edited by Dread Render, 29 September 2016 - 12:22 PM.
#8
Posted 29 September 2016 - 01:38 PM
FWIW, when i first got my STK-5M I lost my first 12 matches playing that mech; granted, i would alternate between it and my other mechs, but still... am I happy to not have to play the Oceanic servers anymore
#9
Posted 29 September 2016 - 02:36 PM
#10
Posted 29 September 2016 - 10:53 PM
El Bandito, on 29 September 2016 - 12:13 PM, said:
I am an okay player for my level, those 16 games contained 2 where i was just running behind my team to keep up finishing with 250 to 300 dmg.
But thank is was a strange experience, not bad ofc but somehow strange.
Dread Render, on 29 September 2016 - 12:22 PM, said:
that is if the MM is working.
Dread Render, on 29 September 2016 - 12:22 PM, said:
#11
Posted 30 September 2016 - 09:09 AM
Jehofi, on 29 September 2016 - 10:48 AM, said:
Odds for that are 1:65536, that sound like a lot player playing a lot of games or did PGI rig the system to try to turn me into a whale? What is your take on that?
That's not how you calculate probabilities. You engaged in cherry picking of your data. What are the odds that you got a 16 game win streak? 1 It already happened.
It's the basketball "hot hand" fallacy all over again.
Edit: To properly appreciate your "luck," you need to know the results for every 16 game sequence for every single other player in the game over your chosen time period. That's going to be hundreds of thousands of sequences. Within all those possible sequence results, 1 in 65,536 will be 16 wins. There will be several of those within the complete data set. You just happened to notice one of them. It really isn't an unusual event given the massive number of 16 game sequences generated by the game.
Every single 16 game sequence has the same probability! Assuming you used a 50% win chance it looks like this:
WWWWWWWWWWWWWWWW - 1 in 65,536
WWWWLWWWWLWWLLWW - 1 in 65,536
LLWWWLWWLLWLWLLW - 1 in 65,536
LLLLLWWWLLWWLWWL - 1 in 65,536
And so on. I don't see players creating threads to bemoan the matchmaker because they experienced any of those other sequences. In short, most people are statistically ignorant.
Second edit: If you really want to use occurrences of 16-game win streaks as some metric for the "goodness" of the matchmaker, here's what you need to do.
1 - Record the results for every single 16-game sequence for every player in the game. For example if you played 20 games, then there are five 16-game sequences to record (1-16, 2-17, 3-18, 4-19, and 5-20). You are going to end up with millions of sequences to analyze in just a few days.
2 - Turn each of those sequences into a win/loss total. For example: WWWLLWLWWLLLWWLW would be a 9-7.
3 - Create a frequency distribution graph for all of the win/loss totals.
4 - Run a statistical test to determine if this frequency distribution is significantly different from a normal distribution of the win/loss totals.
All of this only works as a test of the matchmakers effectiveness if you intended for the matchmaker to create situations that would force every single player to have a win/loss ration of 1. If that isn't what you designed the matchmaker to do then talking about having a 16-game win streak is meaningless in regards to the effectiveness of said matchmaker.
A more appropriate goal of the matchmaker is to have a normal distribution of win/loss ratios for all players. That means the vast majority of players would be right around a ratio of 1 with a small percentage at each extreme. (This assumes that you don't lose or gain new players during your evaluation period and that all players play roughly the same number of games.)
In short, programmers don't try to make matchmakers that produce statistically "fair" results because it's nearly impossible to do or impossible to verify if you've done it. Matchmakers are designed to make as many players as possible feel like the games are fair. That's not the same thing as being statistically normal.
Edited by vandalhooch, 30 September 2016 - 10:09 AM.
1 user(s) are reading this topic
0 members, 1 guests, 0 anonymous users
























