Page 4 of 18
Re: NAF Data for Number-Crunchers
Posted: Fri Feb 13, 2015 2:41 pm
by lunchmoney
lunchmoney wrote:There are 3 matches that have a score line of -1/-1. How did that happen? (Two of them from Thrudball 2013 and one from the idea thieving Speedbowl Cup in Australia. I'm slightly concerned over the Thrudball entries as I was the one who recorded results on the day......I pretty sure I didnt put in a -1 score line.... )
Did anyone take a look at this? Surely -1/-1 shouldnt be in there?
Re: NAF Data for Number-Crunchers
Posted: Fri Feb 13, 2015 2:45 pm
by lunchmoney
mubo wrote:lunchmoney wrote:mubo wrote:I still think it's bizzare though how often lunchmoney was the 'home' player.
I like the left....
If you entered your own results that might explain it...
I've always entered results by uploading the Score file, even when I do someone elses tourney (Thrud for example) I set it all up for Frank to just upload the Score file.
It really is just I like the left and tend to gravitate there
Re: NAF Data for Number-Crunchers
Posted: Fri Feb 13, 2015 11:54 pm
by Fassbinder75
Chris wrote:
The only one that makes me really sad is Humans. There performance at any TV doesn't seem to match their fluff position at all. Orcs to a lesser extent.
You'd figure that teams that take a big hit in percentage are those that are likely to be struggling. Humans losing 3% is a big black eye - the 'Orcs are hurt by n00bs' argument is more widespread in addition to being a pretty meh roster.
Having just about everyone between 45 and 55% does indicate very good game balance generally, certainly tournaments that hand out little inducement to Woodies and Undead and plenty to Chaos and Humans would be the way forward in my estimation - whether it screws up the numbers or not.
Re: NAF Data for Number-Crunchers
Posted: Sat Feb 14, 2015 11:23 am
by besters
Has anyone else looked at this at a coach level? For me that gave some surprises!
Re: NAF Data for Number-Crunchers
Posted: Sat Feb 14, 2015 2:52 pm
by sann0638
lunchmoney wrote:lunchmoney wrote:There are 3 matches that have a score line of -1/-1. How did that happen? (Two of them from Thrudball 2013 and one from the idea thieving Speedbowl Cup in Australia. I'm slightly concerned over the Thrudball entries as I was the one who recorded results on the day......I pretty sure I didnt put in a -1 score line.... )
Did anyone take a look at this? Surely -1/-1 shouldnt be in there?
Will pass it on.
Re: NAF Data for Number-Crunchers
Posted: Sat Feb 14, 2015 6:54 pm
by Oventa
Well not directly to the winning statistics, but still a nice view.
based on
a) games per nation summary:
http://www.thenaf.net/2015/01/naf-data-2013-14/
b) active membership per nation
http://www.thenaf.net/2015/02/active-na ... -snapshot/
only listing the nations mentioned in a)
I removed the europen/eurobowl event from belgium stat, to be somewhat fair.
sorted accodring to games per member.
So I guess belgium/netherland profit rom their neighbouring countries. active player award seems to go to France then
membergames.jpg
Code: Select all
Nation games played 2014(right axis) active member(right axis) games per active naf member(left axis)
Austria 0 6 0
Switzerland 78 17 4,59
United States 1304 271 4,81
Denmark 227 43 5,28
Canada 409 75 5,45
Sweden 520 77 6,75
Germany 1126 160 7,04
United Kingdom 2410 304 7,93
Spain 2631 294 8,95
Netherlands 212 23 9,22
New Zealand 192 20 9,6
Australia 895 92 9,73
Italy 2270 179 12,68
Belgium (w/o EuroEvent) 594 39 15,23
France 2660 140 19
Belgium 1386 39 35,54
Re: NAF Data for Number-Crunchers
Posted: Mon Feb 16, 2015 6:07 pm
by mubo
Updated graph, with win percentages in the cells.
Home/away now ignored, and mirror matches removed.
Re: NAF Data for Number-Crunchers
Posted: Mon Feb 16, 2015 6:13 pm
by mubo
Also, worked out the most successful players during this period. Removed games with and vs the weakest 6 races.
The performance metric is based on first quartile of the beta distribution, so basically you get credit for having a small variance as well as high win ratio, ie the more games played the more confident in a coach's ability abilities we can be. I wouldn't pay too much attention as tweaking the parameters makes a sizable difference to the order, depending on the balance between win % and games played.
Code: Select all
| ID|NAME | raw_performance| played_games| mean_losses| mean_wins| performance|
|-----:|:-------------|---------------:|------------:|-----------:|---------:|-----------:|
| 19538|cyberedelf | 0.9444444| 15| 0.8333333| 14.16667| 0.8526479|
| 13696|Gaixo | 0.8145833| 50| 9.2708333| 40.72917| 0.7675509|
| 20024|Turbo_Ralle | 0.9285714| 7| 0.5000000| 6.50000| 0.7668428|
| 16130|Noerciux | 0.8750000| 12| 1.5000000| 10.50000| 0.7624592|
| 9721|twak | 1.0000000| 4| 0.0000000| 4.00000| 0.7578583|
| 5290|Pippy | 0.7843112| 142| 30.6278061| 111.37219| 0.7578304|
| 13583|Harti | 0.7637845| 165| 38.9755639| 126.02444| 0.7389384|
| 20025|kapi | 0.8409091| 14| 2.2272727| 11.77273| 0.7384475|
| 5654|Pulpo_Fiction | 0.7954545| 36| 7.3636364| 28.63636| 0.7374653|
| 586|RentTreznor | 0.9166667| 6| 0.5000000| 5.50000| 0.7371558|
| 20455|envy89 | 0.9166667| 6| 0.5000000| 5.50000| 0.7371558|
| 3312|Joemanji | 0.7572917| 152| 36.8916667| 115.10833| 0.7311563|
| 1103|Tarra | 0.8333333| 14| 2.3333333| 11.66667| 0.7307547|
| 13561|Matt_le_Fou | 0.7731481| 61| 13.8379630| 47.16204| 0.7300217|
| 11777|Purplegoo | 0.7556452| 157| 38.3637097| 118.63629| 0.7299352|
| 8199|disk | 0.7791667| 46| 10.1583333| 35.84167| 0.7285604|
| 13417|Spartako | 0.7497830| 224| 56.0486111| 167.95139| 0.7284885|
| 7037|GardenGnome | 0.7500391| 113| 28.2455790| 84.75442| 0.7190982|
| 11491|BiBi | 0.7589286| 70| 16.8750000| 53.12500| 0.7187152|
| 13409|mubo | 0.8000000| 20| 4.0000000| 16.00000| 0.7173995|
Re: NAF Data for Number-Crunchers
Posted: Mon Feb 16, 2015 6:34 pm
by sann0638
sann0638 wrote:lunchmoney wrote:lunchmoney wrote:There are 3 matches that have a score line of -1/-1. How did that happen? (Two of them from Thrudball 2013 and one from the idea thieving Speedbowl Cup in Australia. I'm slightly concerned over the Thrudball entries as I was the one who recorded results on the day......I pretty sure I didnt put in a -1 score line.... )
Did anyone take a look at this? Surely -1/-1 shouldnt be in there?
Will pass it on.
Frank says Dockball for the Thrud ones. Don't know about Speedbowl.
Re: NAF Data for Number-Crunchers
Posted: Mon Feb 16, 2015 8:07 pm
by Shteve0
What happened in Austria 2013 and Belgium 2014? In 2014 Austria fell 508 games (100% drop!) while Belgium increased 780 games (129% increase) year on year. Moving event?
In the meantime, the best places to live for BB (based on highest # games played per million population):
2013: Austria (59.26), Spain (56.56), Belgium (54.11), Sweden (45.37), Australia (45.35)
2014: Belgium (123.77), Spain (56.62), Sweden (53.37), New Zealand (42.57), France (39.93)
Full list (all pops are latest estimates as taken from Wikipedia, which is never wrong):
Code: Select all
Nation 2013 2014 Population 2013 gpm 2014 gpm
Belgium 606 1386 11198638 54.11 123.77
Spain 2628 2631 46464053 56.56 56.62
Sweden 442 520 9743087 45.37 53.37
New Zealand 132 192 4509900 29.27 42.57
France 2914 2660 66616416 43.74 39.93
Denmark 240 227 5700000 42.11 39.82
Australia 1077 895 23746600 45.35 37.69
United Kingdom 2485 2410 64100000 38.77 37.60
Italy 2010 2270 60782668 33.07 37.35
Germany 1193 1126 80716000 14.78 13.95
Netherlands 234 212 16912640 13.84 12.54
Canada 584 409 35675834 16.37 11.46
Switzerland 66 78 8183800 8.06 9.53
United States 1502 1304 320206000 4.69 4.07
Austria 508 0 8572895 59.26 0.00
NAF Data for Number-Crunchers
Posted: Mon Feb 16, 2015 8:30 pm
by Lard
Eurobowl happend to belgium and Austria
Re: NAF Data for Number-Crunchers
Posted: Tue Feb 17, 2015 12:59 am
by Shteve0
Sounds reasonable. Thanks Lard
Re: NAF Data for Number-Crunchers
Posted: Thu Feb 26, 2015 1:08 am
by Wulfyn
Fassbinder75 wrote:Chris wrote:
The only one that makes me really sad is Humans. There performance at any TV doesn't seem to match their fluff position at all. Orcs to a lesser extent.
You'd figure that teams that take a big hit in percentage are those that are likely to be struggling. Humans losing 3% is a big black eye - the 'Orcs are hurt by n00bs' argument is more widespread in addition to being a pretty meh roster.
Having just about everyone between 45 and 55% does indicate very good game balance generally, certainly tournaments that hand out little inducement to Woodies and Undead and plenty to Chaos and Humans would be the way forward in my estimation - whether it screws up the numbers or not.
I had a look at these arguments and the results were surprising. Overall I recorded Orcs as having the 14th best win record. I then looked at the average number of tournament games played by each player, and then created a weighted average for each race. Orcs did indeed come out as very low, 23rd in the list with an average of 45 games played (only Chaos were lower at 43.4). This seemed to tie up the argument; a lot of 'noobs' play Orcs, so they do worse.
But that only answered half the question. I then decided to filter the win rate to remove coaches that had played a low number of games. I selected 13+, 31+, and 61+ as being roughly equal to 3+, 6+, and 11+ weekend tournaments. Orcs were 14/14/13/13, meaning that even amongst experienced players they don't do that well. I haven't looked at anything temporal yet, so it still might be that people start with Orcs, do badly, and then play other teams and improve.
Incidentally on that same basis humans were 15/12/10/8 - a significant improvement by experienced coaches.
Re: NAF Data for Number-Crunchers
Posted: Thu Feb 26, 2015 7:17 am
by sann0638
Superb statting! Interesting disbelief at the weekend at the numbers to do with nationality, so might look into this a bit more.
Re: NAF Data for Number-Crunchers
Posted: Thu Feb 26, 2015 8:16 pm
by Wulfyn
I've popped the results into a google spreadsheet if anyone wants more detail.
https://docs.google.com/spreadsheets/d/ ... edit#gid=0
Slann are the most experienced team faced, with players having had an average of 62.7 tournament games.
This is followed by Woodies (59.6), Pro Elves (58.7), Undead (57.2), and Underworld (55.7).
That's probably an unexpected mix of teams, and the reason is also mixed.
When dividing the average experience by the number of games played we get a loyalty score. Basically if a team has a high experience but not many games played it is because experienced coaches will be playing these teams infrequently - perhaps as part of making their way to playing all 24 teams. This will give a low loyalty score. On the other hand if they are played a lot then they should get a high loyalty score.
This matches an expected pattern. Slann (18.0), Pro Elves (27.
, Underworld (18.9), and Vampires (19.2) have low loyalty scores. Woodies (72.9), Undead (92.3), and Norse (81.5) have good loyalty scores. But interestingly the highest loyalty score is from Orcs (129.6). This means that although there are a lot of Orc players that come for a few games and then never come back, even the ones that do come back seem to stick with Orcs, or more likely that other experienced players tend to not pick them.