11 replies to this topic

[Duke Nedo]

Edit: Wrote this in a hurry... I did not post this to brag or anything, but to point out the weird distribution of wins and losses that the MM set me up for as usual during tournaments. Drop in the most populated class = loss loss loss, drop in the underpopulated class: win win win. This has been very consistent for me during tournaments and challenges lately. Interested if others see something similar?

Would be lovely if PGI would look into how the MM enforces 3-3-3-3 during tourneys with the mish-mash of Elos from the group queue mixed with solo queue Elo... I doubt I "carried" so much harder in the raven compared to the timberwolf.

------------------------------------------------------
Games,Wins,Qualified,Kills,Deaths,K/D,win%
Raven-2x: 26, 20, 14, 30, 8, 3.75, 76.92
Timberwolf-prime: 26, 9, 6, 23 ,14 ,1.64, 34.62
Stormcrow-prime: 1, 0, 0, 2, 1, 2, 0

I had a horrible time trying to get on winning teams in my Timby, so I switched to my rvn-2x and everything got completely different... from 35% win ratio up to 77% win ratio. Did any of you experience something similar? MM always goes nuts for me during tourneys... whined before.

Would be interesting to see corresponding stats whatever they are... what did you go for, how did it work out?

Edited by Duke Nedo, 23 November 2014 - 09:15 AM.

[Duke Nedo]

...and just to underline the effect (and lack of my own importance) I dropped in my unskilled Mist Lynx and got another 3 straight wins. My Elo is probably lower in lights than in heavy, but I doubt that is all there is to it.

[Angel of Annihilation]

Challenge Stats:

Mech used: Mist Lynx. (All three variants, none elited)

59 battles required to complete the Challenge.
25 wins, 34 losses
46 kills
38 deaths


Of the 25 wins:

4 wins, survived but did not manage a kill.
1 win, managed at least one kill but did not survive.

[Viges]

Duke Nedo, on 23 November 2014 - 09:15 AM, said:

My Elo is probably lower in lights than in heavy, but I doubt that is all there is to it.

Why? You just proved that elo exists and changes the outcome.

[Abivard]

Viges, on 23 November 2014 - 03:00 PM, said:

Why? You just proved that elo exists and changes the outcome.

About all that may have proved is that the way PGI rates players and the way MM uses that to build matches proves PGI really screwed the pooch on those two things.

[ApolloKaras]

Abivard, on 23 November 2014 - 04:06 PM, said:

About all that may have proved is that the way PGI rates players and the way MM uses that to build matches proves PGI really screwed the pooch on those two things.

I have noticed something with the Pirates Bane. My normal W/L is about 1.5. In that locust after 50 drops (all pug drops with the exception of 5), its @ 3.40 W/L.

[aniviron]

Viges, on 23 November 2014 - 03:00 PM, said:

Why? You just proved that elo exists and changes the outcome.

Got to say, I think this is it. I rarely drive heavies, and took a TDR-9S out for a few spins in its stock loadout (hint: the stock loadout with two flamers). Got a match with 900 damage and a few kills, and no matches under 400 damage.

This doesn't say that the TDR-9S is an amazing killing machine in its stock loadout, which is in fact a terrible build; it says that someone who has over 1000 drops in other weight classes and less than a dozen in the weight class they are playing is probably going to clean up pretty easily, thanks to Elo.

It also says that this is an incredibly small sample size, to the point that calling it a sample size instead of an anecdote is almost insulting.

Here's an anecdote: I play a lot of light mechs, and I'm pretty good with them. I didn't bother dropping more than a couple times in my lights this weekend, because my Elo is high enough that dying is a relative certainty.

[Ragnar Darkmane]

Well, I managed to get 10 qualifying wins out of the first 16 games I played with my Stalker-5S right away...
...Then it took me another 56 matches to get the other 10 qualifying wins... god, that survival condition sure made for bad team play, makes the whole getting 8 loss streaks over and over again thanks to ELO-hell even more painful.

Edited by Ragnar Darkmane, 23 November 2014 - 04:26 PM.

[RockmachinE]

Ragnar Darkmane, on 23 November 2014 - 04:24 PM, said:

Well, I managed to get 10 qualifying wins out of the first 16 games I played with my Stalker-5S right away...
...Then it took me another 56 matches to get the other 10 qualifying wins... god, that survival condition sure made for bad team play, makes the whole getting 8 loss streaks over and over again thanks to ELO-hell even more painful.

It seems the average was ~50.

[Ragnar Darkmane]

Louis Brofist, on 23 November 2014 - 04:29 PM, said:

It seems the average was ~50.

That means I have a ~0.72 qualifying rate compared to the average. Pretty disturbing that it coincides with my ELO-hell win rate... which is exactly the same (yes, despite having a 1.60 KDR and ending up with top damage over and over again...) <.<...
God, I wish PGI would finally adjust the way ELO influences match making and team composition...

[Darlith]

49 total matches
29 wins
6 non qualifiers I made no kills, 4 non qualifiers I died before the end anyway

I don't think I was playing much different from normal, but the event did make me a lot more annoyed than normal at dying to help secure the win. I did notice a lot of still dangerous enemy mechs wandering the battlefield with orange or red torsos because people didn't want to kill that I think contributed to losses on both sides. A great deal more friendly fire then even the past bunch of events it seems, I lost count of the number of times someone ran up behind me and tried to shoot through me turning my core orange before the enemy even fired a shot. And a whole lot of teams being unwilling to push and losing because often in the game the team that seizes the initative wins (mostly due to how fast enemy counterfire withers as mechs die off), this of course happened on both sides of the match.

[Duke Nedo]

aniviron, on 23 November 2014 - 04:18 PM, said:

Got to say, I think this is it. I rarely drive heavies, and took a TDR-9S out for a few spins in its stock loadout (hint: the stock loadout with two flamers). Got a match with 900 damage and a few kills, and no matches under 400 damage.

This doesn't say that the TDR-9S is an amazing killing machine in its stock loadout, which is in fact a terrible build; it says that someone who has over 1000 drops in other weight classes and less than a dozen in the weight class they are playing is probably going to clean up pretty easily, thanks to Elo.

It also says that this is an incredibly small sample size, to the point that calling it a sample size instead of an anecdote is almost insulting.

Here's an anecdote: I play a lot of light mechs, and I'm pretty good with them. I didn't bother dropping more than a couple times in my lights this weekend, because my Elo is high enough that dying is a relative certainty.

I have no reason to believe that my Elo in lights is extremely different from my heavy Elo, but probably a bit lower for sure. Certainly my lights Elo should not be much worse than my mediums Elo. I did however get a very strong feeling that dropping in the significantly underpopulated lights repeatedly put me on the winning teams. Some games I contributed quite a bit, but really, there is limits to how much you can carry a win in a light ridge-humper. I certainly don't think me alone could drive a 77% win ratio in a raven. I have to assume that the MM put me up to it, and a few others have reported similar win-streaks in lights (one in this thread, one mist lynx and one huginn in other threads). Sure, sample is always small, but someone interested in statistics could probably calc the confidence by which 26 games giving a 77% win ratio is not 50:50 chance of winning each game. I am guessing it is significant, but small enough for it to be a real chance that it's random fluctuation.

Anyways, just curious if others see similar trends during tournaments. I have seen similar things every tournament and it usually drives me to drop in low populated classes and then I do much better (and waiting times go down so win-win). Could be self-confirmatory behavior on my behalf though, I see that one and that's why I am asking others what they experienced.

My main curiosity is if there is something in the MM that has a systematic or cyclic behavior to it when it comes to filling the teams, especially at the point when it is releasing the valves for 3-3-3-3. Does it do it fully symmetrically at that point? Perhaps a final re-scrambling of the built teams as a final action would help to remove any bias caused by the filling sequence?