21 replies to this topic

[Call_Me_Ishmael]

What a (good) new player can expect until he joins our TeamSpeak and gets into matches above 2000 MMR.

Use the TeamSpeak! Fight higher-tier players (this plot was made using general 3-star pub matches)

 

The x-axis is days - that's a 10-day span
 

[crockrocket]

I'm confused, how did you make this chart?

[Call_Me_Ishmael]

Confusingly enough, I played about 200 matches with a new account (Lawyer_Daggett), and recorded the MMR I had after each match.

 

 

(edit:  65 matches)

Edited by Call_Me_Ishmael, 21 March 2015 - 12:06 PM.

[crockrocket]

Ohh I see. This wouldn't be what most new players have to look forward to then.

[Shadeness]

Didn't notice periods on the Y Axis and was wondering how the hell you we're calculating 2 million MMR.

[ZombieBiscuit]

I know a lot of people complain about getting beat and stomped by better players but that is how you improve. 

 

You don't improve by playing people lower than you, you improve a little by playing people the same as you, and you improve a lot when playing people better than you. 

[FlamingBeaker]

Hopefully Ashfire will chime in about what the new average MMR is. It used to be 1492, but I suspect it to be higher now. I'd love to see a chart showing MMR distribution in the game (no names of course).

 

Definition of 'good player' varies widely. 2000 feels right to me, but more skilled players will say 2200, someone looking at the majority of games in the server browser will dispute that and say lower, etc. Price of having a high skill ceiling game.

 

 

I know a lot of people complain about getting beat and stomped by better players but that is how you improve. 

 

You don't improve by playing people lower than you, you improve a little by playing people the same as you, and you improve a lot when playing people better than you. 

 

I wish sometimes we had two sets of servers: MMR-balanced and free-for-all unbalanced, just for this reason. Maybe someday, with a larger player base and integrated VOIP or TS, and a better tutorial for new players.

[Call_Me_Ishmael]

I think they would not expect to break 1980MMR for a while, especially playing 3-star pubs.

[Shadeness]

 

I wish sometimes we had two sets of servers: MMR-balanced and free-for-all unbalanced, just for this reason. 

 

What do you think the Mayhem servers are? They ignore MMR restrictions.

[DieselCat]

Didn't notice periods on the Y Axis and was wondering how the hell you we're calculating 2 million MMR.

 

I believe he is showing an mmr calculation of 2000. If one pulls up their mmr using scrimbot they will get a number for example.... 1594.76, which would be an average of +/- 100 points.

 

I was a little confused at that chart number at first, also thought it was 2 million but it's showing a range of 1400 to 2000 mmr.

 

*+ 

[Call_Me_Ishmael]

Default Google Doc format, sorry. 

[DieselCat]

Default Google Doc format, sorry. 

 

No worries, you still did a wonderful job... 

 

*+

[EM1O]

not sure right now, but i answered a reddit inquiry about it by making another smurf, default entry level was 1250, 2 weeks ago.

[Call_Me_Ishmael]

I think your image needs a disclaimer added to it in order to be taken seriously.
 

Spoiler

* Actual user experience may vary.

Here's what I think is happening...

We have, likely, a bi-modal distribution (two humps in the curve below) in the MMR population. I need some data from Ash when he has a chance, to show this is real, not hypothesis. I think we have a dead-zone region where people who either are playing nice and picking three-star servers or are letting the matchmaker pick matches for them (just hitting the find-match button). My hypothesis is that the penalty for knocking a player backwards in MMR for failing to make the top 50% of the server score (or score/min if in match long-enough) is TOO LARGE, and should be smaller; alternatively the bonus adder (if any) for playing above your head should be larger.

Bi-modal - this is *not* real data, but it illustrates my point:


I think we have a deadzone like this:

 
I think the only way around this experience for the casual player (being capped at/not easily breaking 2100) is to play the 1-star matches that are OVER your MMR, and the way to do that reliably is to use clan or the Unofficial Hawken TeamSpeak servers (edit: and to actually be good enough to score in the top 50% of these servers...).  Since there are essentially no clans at the moment, everyone who thinks they're actually better than their current MMR should use the TS server: ts5.gameservers.com:9152

Well, there is another way: farm MMR by stomping, but you get smaller and smaller increases in MMR the more imbalanced your MMR is compared to the server average (this is what makes me believe there is a scalar/adder to the MMR you earn for finishing in the top half of the server score/min). But this is uncool and worsens the casual experience.
 

Anyway, you use the server to find matches (i.e. pick a server where a bunch of good players will start a match together, or fight-club or other activity), and join the named server.  Many players will friend you for a party if that's the only way to get you into a match with them and you're in-channel.

 

Good Luck, Pilots.  Go earn your MMR!

Edited by Call_Me_Ishmael, 21 March 2015 - 02:59 PM.

[crockrocket]

Hmm Hyginos' histogram would be relevant here

[Hyginos]

Hmm Hyginos' histogram would be relevant here

 

you called?

 

*warning: its kinda big*

 

Spoiler

 

 

This is based on data that is a few months old, mind you. I think early December. I deleted an outlier at the top that was like 3400 something, as well as I think some single digit mmrs at the bottom. IIRC there was a huge falloff after 50 mmr down to <1 range.

 

condensed (I only have this as a picture for some reason.):

Spoiler

Edited by Hyginos, 21 March 2015 - 03:36 PM.

[Call_Me_Ishmael]

I'd really like the data from the last month or so, and I'd really like to be able to track individual player's MMR movement (not by name), and do some regression on the data...

 

Shadeness, Ash, if you have a few moments, grab me on the TS sometime?

Edited by Call_Me_Ishmael, 21 March 2015 - 03:39 PM.

[Ashfire908]

I'd really like the data from the last month or so, and I'd really like to be able to track individual player's MMR movement (not by name), and do some regression on the data...
 
Shadeness, Ash, if you have a few moments, grab me on the TS sometime?

I've been trying to plan a thing like this, but never got it up and running. I'll try to be in TS today though (I'm there right now) and catch you.

[Call_Me_Ishmael]

SS - I quoted you because I think the experience really doesn't differ  that much for many/most players ('your mileage may not vary much').

 

I am encouraging four things:

1. use of the TS server, because public
2. better analysis and collection of available data to support the hypotheses out there (mine's only one)
3. revisiting our MMR parameters/coefficients
4. players to rise above and test themselves.  Because competition.

Edited by Call_Me_Ishmael, 21 March 2015 - 04:21 PM.

[deidarall]

Man, I never really tracked mmr, Also there is a TS? :0

[Merl61]

Man, I never really tracked mmr, Also there is a TS? :0

Yes. The post isn't out yet but the server is ts5.gameservers.com:9152

[SamEureka]

you called?

 

*warning: its kinda big*

 

Spoiler

 

 

This is based on data that is a few months old, mind you. I think early December. I deleted an outlier at the top that was like 3400 something, as well as I think some single digit mmrs at the bottom. IIRC there was a huge falloff after 50 mmr down to <1 range.

 

condensed (I only have this as a picture for some reason.):

Spoiler

I f-ing love numbers...