?

Reply
#1
Old 8-05-2016, 01:31
Perth
27 Years Old
71 Posts
Steam Profile
muppet's Avatar
muppet90 muppet is offline aka bnwmuppet, el mupito
Member
I've had a pretty quiet week at work so I put together stats for the OZF 15 premier regular season. I'll just start throwing some graphs out and discuss them/methodology as they come up.

SCOUTS


Not surprising to many that the Jasmine scouts top the K/D and KA/D, probably helped by not losing a game this season. Elmo finishes the regular season with the highest kills per minute of all players in Prem.


I've gone through and weighted the DPM values for every game every player has played, based on the length of that map. The logic here is that if someone does 400 DPM in a 10 minute map, and 300 DPM in a 30 minute map, it wouldn't make sense to say that that player "averages" 350 DPM - so weighting values (i.e. a discrete weighting function) is used to take this into account.

Players that show a decrease to their DPM as a result of weighting generally hit higher DPM than normal in shorter maps. This makes sense (especially for scouts) if these longer games are a bit stale-matey and there's less direct fighting to boost damage stats.

The blue column to the left shows a players DPM per % heals received. Players with a greater value are more efficient in dealing damage with their heals. Elmo was the most efficient scout this season, averaging 324 DPM (305 weighted) with only 7% heals. This is in contrast to the likes of Yui (316 DPM, 311 weighted, 15%) and cooki (312 DPM, 298 weighted, 15%) who received over double the % of their teams' heals compared to elmo.



Shoutouts to my teammate teejay, who despite playing for our 7th place team, consistently managed to perform well in tough matches.

You'll notice Abender has a lower DPM than most scouts. Well,


To be honest, I'm surprised he even played scout 65% of the time. Moreover, how did he only spend 15% of the time playing pyro???

POCKETS



Unsurprisingly, Geoh and Yuki have the best season K/D's on pocket, while the bottom 3 pockets for K/D are from the bottom 3 teams.


The damage stats show that while Geoh, Yuki and Rocky averaged the highest DPM and weighted DPM; Aporia (when playing pocket) and lock were the most efficient in dealing damage with their heals. Rocky received a massive average of 42% of the heals over 6 maps, giving him the lowest DPM/%Heals. He did have the highest average airshots of 2.8 per map (back to this later).


We can see that when Aporia plays pocket, he on average receives the lowest amount of heals of any pockets this season, while dealing good damage.

ROAMERS



Jasmine's zaka was the only roamer to achieve a season K/D of 1, and also achieved the highest KA/D of all roamers. Mercy and Aporia tied for the highest kills per minute while playing on the flank.


Zaka tops the damage stats as well, the only roamer to average over 300 DPM. Ransibling (over 4 maps) is the most efficient roamer for damage, dealing 257 DPM off 6% heals, followed by Aporia and lambda.


We can see the difference in playstyles between some of the teams here, with the likes of Cooking with Coolio, FIFO, 11corners and Jasmine giving their roamers 10% or more heals.

DEMOMEN



Somewhat surprisingly, enrith finished the season with the highest K/D, ahead termo, bulk and paulsen. Termo achieved a significantly higher rate of kills per minute than the other demos this season.


termo finished the season with the highest DPM of all players (389 DPM, 378 weighted DPM) with 27% heals. Paulsen was the most efficient demoman over the season, averaging 350 DPM (337 weighted) with 21% heals.


I'd also like to highlight the performances of both bulk and tekk, who don't seem to have lost any ability after some time away from TF2.

MEDICS



Firstly sorry for the cut-off of Ampersand's name, it was a lot harder to fix than it seems and I've used enough of my time on this. We see that the medics from the top 2 teams, brutalix and Jersh, have the highest KA/D's. However, a fine season by Dave sees him come in 3rd. We also see Dave and melons (4 maps) dominate the kills per 100 minutes. Again, we see the bottom 4 medics in KA/D come from the bottom 3 teams.
Ampersander's high drop rate is attributed to the amount of times he was killed with at least 25% vaccinator charge.


We See that Dave and rHk embarassed the most people this season. While the death rate generally reflects a medics' teams' finishing position, we again see Dave with a low death rate playing for the team that finished 5th.
I tried to normalise charge rates for vaccs and kritz, by assuming that kritzkrieg builds 20% faster than uber, and vaccinator charges build 40% faster, where 1 vaccinator charges = 4 charges as presented on logs (i.e. 100% vaccs). The fact a vaccinator can be charged before you get 100%, and the amount of time Ampersand spent using it (46% of all his charges were not ubercharge), we see Amper with the highest rate of charges over the season.
Brutalix, Dave, Dramz and myself have either only used uber or used a non-uber charge such a small amount of times that it rounded to 0.

INDIVIDUAL AND DICK-WAVING STATS



*PLEASE NOTE* Abender recorded 32 kills as sniper - with the machina. Sorry to anyone else who used the machina, I didn't go through and look at this.
Geoh narrowly wins the dick-waving competition ahead of Render for the most airshots this season. Yui has the most headshots for the season, beating out Sam. While shooter has the most points captured for the season by a fair margin.






You can view and sort through all of the raw (totalled and analysed) data for all players who played at least 4 maps here!

Notes
  • Players with * next to their name on main class stats charts played at least 4 but less than 6 maps over the season
  • Where a player played a non main class (i.e. Yui playing demo for a map in R3), these stats are not included in their analysis

I'll probably update this as I find errors and add/remove things.

~ Thx for reading ~
Last edited by obla; 8-05-2016 at 19:40. Reason: .
Reply With Quote
#2
Old 8-05-2016, 02:12
Australia
1,753 Posts
QUACK!Steam Profile
Smithsonian's Avatar
racSmithsonian Smithsonian is offline aka Smithsonian, Smashsonian
Member
Discredited all of your stats when you said abender played scout 65% of the time.

For reals though sick writeup
________________________
[v3] Kaki : and shitstainian worst demo ever
Reply With Quote
#3
Old 8-05-2016, 03:40
United States
61 Posts
racmustardoverlord mustardoverlord is offline
Member
finally, the statistical proof I need to make fun of tom for being the worst roamer in tf2!!
Reply With Quote
#4
Old 8-05-2016, 04:11
Victoria
11 Posts
Steam Profile
Vanquish Vanquish is offline
Member
This is some awesome stuff, great read.
Reply With Quote
#5
Old 8-05-2016, 10:06
Zambia
2,556 Posts
QUACK!
Steam Profile
hobbes's Avatar
rachobbes hobbes is offline aka the tommunist, Mohawk, The Tommunist, Tom, Tommunist, Alpha G, tommu, hobbes, tomm
Member
Quote:
Originally Posted by mustardoverlord View Post
finally, the statistical proof I need to make fun of tom for being the worst roamer in tf2!!
________________________
enWRATH: dash is a fuckin minx
Reply With Quote
#6
Old 8-05-2016, 14:07
Victoria
480 Posts
Steam Profile
lock.'s Avatar
:B1:lock. lock. is offline aka 'Lockrice.
Member
awesome writeup, should get it put on tftv
________________________
many could beat anyone using red only
Reply With Quote
#7
Old 8-05-2016, 14:37
Perth
27 Years Old
71 Posts
Steam Profile
muppet's Avatar
muppet90 muppet is offline aka bnwmuppet, el mupito
Member
Quote:
Originally Posted by lock. View Post
awesome writeup, should get it put on tftv
guess i'm handballing that one to brodogs?
Reply With Quote
#8
Old 8-05-2016, 14:51
Victoria
318 Posts
Steam Profile
2y.Alex. Alex. is offline aka Santa
Member
10/10 write up
________________________
johnny : im having fun shut up u tryhard faggot shitter milf fuckwit nigger
Reply With Quote
#9
Old 8-05-2016, 14:51
New South Wales
25 Years Old
1,455 Posts
Steam Profile
Brodogs's Avatar
ozfortressBrodogs Brodogs is offline
Retired Media Coordinator
This probably isn't the type of news content TFTV are looking for but you should definitely post it as a thread!

Thespire ~may~ have interest in posting this though. I'll ask them so hold off on TFTV for now.
Last edited by Brodogs; 8-05-2016 at 14:53.
Reply With Quote
#10
Old 8-05-2016, 15:41
Melbourne
201 Posts
Steam Profile
ckmm's Avatar
ckmm.tbz ckmm is offline aka .ckm
Member
Awesome stuff
Reply With Quote
#11
Old 8-05-2016, 15:50
Brisbane
379 Posts
Steam Profile
Snazzy's Avatar
Nobel_Snazzy| scrap.tf Snazzy is offline aka snazzy, splinter, coach
Member
really good stuff, this writeup
________________________
http://ozfortress.com/showpost.php?p=999715&postcount=32
Reply With Quote
#12
Old 8-05-2016, 19:25
New South Wales
107 Posts
Steam Profile
solace''s Avatar
solace'mate solace' is offline aka aka, leos
Member
really interesting stats, cheers for this muppet!
Reply With Quote
#13
Old 8-05-2016, 21:24
Brisbane
22 Years Old
13 Posts
Steam Profile
Thyme's Avatar
Thyme90 Thyme is offline
Member
great read
Reply With Quote
#14
Old 10-05-2016, 02:02
Melbourne
49 Posts
Steam Profile
SBS |capesmate capesmate is offline
Member
awesome work man, good read
Reply With Quote
#15
Old 11-05-2016, 21:40
Australia
512 Posts
Steam Profile
LF's Avatar
RUFFIAN'LF#45&M LF is offline
Member
Awesome analysis piece and equally as awesome write up.

I'm quite interested in knowing how the weighting was calculated.
Reply With Quote
#16
Old 12-05-2016, 16:25
Australia
370 Posts
Steam Profile
abender777's Avatar
mG:sabender777has ze mindset abender777 is offline
EPIC Donator
Thanks for doing these stats!

I had a play around with the stats that were linked, have a look here:
https://public.tableau.com/views/OWL...es&:showTabs=y

You can use the filters on the right to show/hide teams or classes. Also hover over the dots for extra info.

Size represents heal % for their team, and colour is class. There are multiple points for Yuki/Aporia as the data has them under multiple classes.
Last edited by abender777; 12-05-2016 at 16:28.
Reply With Quote
#17
Old 12-05-2016, 16:32
Sydney
108 Posts
Steam Profile
quid_'s Avatar
quid_in 3D quid_ is offline aka Munch, munchtism, v0id
Member
melons # 1
________________________
pause: unbind g
Reply With Quote
#18
Old 12-05-2016, 17:27
Perth
34 Posts
Steam Profile
mlns's Avatar
hc.mlns mlns is offline aka MoistMelons
Member
:|
:O
:{D
________________________
ah!
Reply With Quote
#19
Old 12-05-2016, 18:43
Adelaide
538 Posts
Steam Profile
riotbz's Avatar
riotbz riotbz is offline aka Running.Riot, ri0t, rio
Member
was there not enough space on the graph for smithsonian 8)
________________________
dude nice
Reply With Quote
#20
Old 13-05-2016, 13:24
Perth
27 Years Old
71 Posts
Steam Profile
muppet's Avatar
muppet90 muppet is offline aka bnwmuppet, el mupito
Member
Quote:
Originally Posted by abender777 View Post
Thanks for doing these stats!

I had a play around with the stats that were linked, have a look here:
https://public.tableau.com/views/OWL...es&:showTabs=y

You can use the filters on the right to show/hide teams or classes. Also hover over the dots for extra info.

Size represents heal % for their team, and colour is class. There are multiple points for Yuki/Aporia as the data has them under multiple classes.
fuCK. data presentation gives me hardons. this is sick
Reply With Quote
#21
Old 13-05-2016, 13:49
Perth
27 Years Old
71 Posts
Steam Profile
muppet's Avatar
muppet90 muppet is offline aka bnwmuppet, el mupito
Member
Quote:
Originally Posted by LF View Post
Awesome analysis piece and equally as awesome write up.

I'm quite interested in knowing how the weighting was calculated.
I'm not sure how / if ozfortress supports latex math typefacing so this might look gross. But in general a weighted average of a discrete data series would take the form

weighted ave = (sum over all elements of (weight of value x value)) / (sum over all elements (weights))

The major difference map to map that would influence the dpm values (in my opinion) is time, as i discussed in the first post. So I took the weights as

weight of map = map time / total time played (player)

This also allowed for sum(all weights) for any given player to =1 and reduce the amount of shitty cell formulas.

So a 2 map example would be something like, a player plays 2 maps:
Map 1: 12 mins - 372 DPM
Map 2: 30 mins - 290 DPM
W1 = 12/42 = 0.286; W2 = 30/42 = 0.714

Weighted DPM = 0.286*372 + 0.714*290 = 106.39+207.06 = 313 DPM
Versus your standard mean where all maps are equally weighted => AVG (nonweighted) DPM = 331 DPM.

So the differences in general weren't huge, but are just more *statistically* accurate.
In an ideal world I could've taken all the DPM's achieved and plotted them versus map time and seen if there was a correlation/weibull/normal dist etc. Developing pdfs and cdfs to accurately 'weight' data.
You'd then probably have to start looking at things like an effective 'ELO' for opponents, whether the map played favours certain classes over others, and then comes to the fact that the data sample size is actually really too small to be doing any of these things.

A quick plot of all of Jasmines DMG over map time

Shows that in general longer maps product less damage per minute than fast maps. As you'd intuitively expect.
Reply With Quote
#22
Old 13-05-2016, 15:07
Australia
242 Posts
Steam Profile
eggnormous's Avatar
Sneaky Snakeseggnormous eggnormous is offline aka Eggnormous
Member
crazy good! thx bud
________________________
╰⋃╯ヽ༼ • •༽ノ✂
Reply With Quote
#23
Old 13-05-2016, 16:24
Melbourne
24 Years Old
194 Posts
Steam Profile
nope's Avatar
nope.mE nope is offline aka Damien, d4m0
Member
How did you take pauses into account?
Reply With Quote
#24
Old 13-05-2016, 17:19
Perth
27 Years Old
71 Posts
Steam Profile
muppet's Avatar
muppet90 muppet is offline aka bnwmuppet, el mupito
Member
Quote:
Originally Posted by nope View Post
How did you take pauses into account?
From zoob's September 2014 logs.tf changelog
Quote:
Pauses no longer affect match time and per minute stats (with plugin support)
Reply With Quote
#25
Old 13-05-2016, 21:07
Melbourne
24 Years Old
194 Posts
Steam Profile
nope's Avatar
nope.mE nope is offline aka Damien, d4m0
Member
Well I know that some of our (recent) logs have counted pauses in the time and dpm, eg http://logs.tf/1383976?highlight=76561198078214483
Reply With Quote
#26
Old 13-05-2016, 23:25
Melbourne
527 Posts
Steam Profile
v4na's Avatar
Thev4naGoose v4na is offline aka ☆Seadra☆, Kgd, Vanadium, Vana, Vana.
Member
epic stuff muppet

modus#1
Last edited by v4na; 13-05-2016 at 23:27.
Reply With Quote
#27
Old 23-05-2016, 13:41
Australia
1,061 Posts
QUACK!
Steam Profile
spitty's Avatar
«spitty spitty is offline aka Spitty
Member
Nice stats
________________________
<@valk_> !kb bowt go follow a fucking team sport u dumb cunt
Reply With Quote
#28
Old 24-05-2016, 16:31
Australia
116 Posts
Steam Profile
Jiho's Avatar
mxJiho Jiho is offline
Member
Elmo is my hero
Reply With Quote
Reply
Thread Tools