Talk Fantasy Football

GalakStarscraper wrote:Sandwich ... thanks for that summary by race.

That chart makes me pretty happy. I will readily admit that there are still things in the rules that could be tweaked. But in terms of our work to hit the Tiers ... it is really good to see that 20 of the 24 teams land in the Tiers that we wanted them to and the 4 that do not JUST miss them but not by much (1.26%, 0.58%, 0.87% and 1.04%)

Wood Elf is over 55%
Vampire is over 45%
Chaos is under 45%
Halfling is over 35%

Otherwise the other 20 are in their Tiers. Just having a happy developer moment seeing that.

European tournaments that make up the bulk of the numbers tend to be low TV + skill packs which tend to favour rosters like Undead, Wood Elves and Lizards at the expense of raw outfits like Chaos, High Elves and Nurgle. Down here in 'riches aplenty' Australia they run at 51% (although much of that is probably Smeborg), so I don't think it's necessarily a fault of the CRP itself, rather Tournament Organisers picking winners.

Ran up a quick heatmap.
It's pretty simple- just took the results from the home team vs the away team, so it is missing some data, although matrix is not symmetrical because of this. I can give code if anyone is interested. Read the row for the home team, ie: Amazons have their best results vs Ogres.

One thing really sticks out at me, is that the home team is FAR more likely to win that the away team:

HomeWin: 13783
AwayWin: 11903
Tie: 7294
This is WAY over anything you would expect by chance. The only thing I can think is that it's something to do with Swiss? I.e. higher ranked teams are more likely to be home. Any other ideas? Reporting errors possibly? Mike, if you happen to have round number, I could check this.

Sandwich/Galak:
I think you may be oversimplifying the data slightly with halflings > 35%. Remember most games are in a swiss format. They will lose more, and hence play weaker races. Even considering the unexplained home team bias, they only have a > 35% record vs 11/23 teams. As the 'away' team 9 teams have a < 65% score against them.

Thinking about it... the home team bias is probably an artefact of tournaments that do not use score? When they manually upload the results, TOs may list the winning team first? Any other reasons?

Would be interesting to compare to the win loss for mid and high tv teams - of course plotted on a chart showing the rise and fall Is the data for FUMBBL and Cyanide out there?

Fumbbl has definitely been done, have a look on the forums there.

Mubo, I think the data is in original order, but I don't have a column for rd 1 etc.

Hi Sandwich,
very interesting win percentages.

Any chance you could list the number of games played for each race alongside the win percentage?
That would be very useful for calculating the confidence interval for those numbers.

Also, any chance you could remove mirror matches?
Mirror matches do nothing but cover up which teams are good and which ones are really bad.
Even more so in a Swiss format - as Mubo has already hinted.

I think that the high win percentage of some of the worst teams, and the (comparatively) low win percentage of some of the power teams is heavily influenced by mirror matches and the Swiss format.
[I did some math on undead and woodies with the previous batch of naf stats, and the impact of removing mirror matches was huge).
Not to mention the fact that tiered bonuses seems to be catching on - which much as a love it will further distort the stats.

Cheers
Martin

Number of games played by each race. Lazy formatting of course. Most of your assumptions about the data are correct, although its too much of a pain in the arse to post them. Continental Europe is an Undead-fest, so many mirror matches drag its win % towards 50 a lot. Without them they're 56.2, not far off Wood Elves 56.8.

Edit: sorry this is raw totals (includes mirror matches)

• Race Number of Records
  Orc 5831
  Undead 5277
  Dwarves 4504
  Norse 4436
  Skaven 4373
  Wood Elves 4345
  Dark Elves 4163
  Lizardmen 3827
  Chaos Dwarves 3822
  Humans 2789
  Necromantic 2667
  Amazons 2273
  Goblins 2029
  Chaos 1798
  Khemri 1682
  Elves 1630
  Halflings 1622
  Ogres 1591
  High Elves 1503
  Chaos Pact 1459
  Slann 1128
  Nurgle's Rotters 1097
  Vampires 1063
  Underworld 1051

Interesting fact... as far as I can see there were only 19 matches between Nurgle and Vampires worldwide in two years, making it the rarest match up. Vampires went 9-4-6 (WTL)

Fassbinder75 wrote:Interesting fact... as far as I can see there were only 19 matches between Nurgle and Vampires worldwide in two years, making it the rarest match up. Vampires went 9-4-6 (WTL)

Must be the Vampires being afraid of food poisoning

So you want to know how teams really stack up? This is results minus mirror & stunty matches. Stunties are removed because they're rubbish and are almost never at the pointy end of a competition. The stunty values here are vs non-stunty teams.

• Team Win%
  Wood Elves 55.54%
  Undead 55.48%
  Lizardmen 53.62%
  Dark Elves 53.14%
  Amazons 51.85%
  Elves 50.75%
  Chaos Dwarves 50.40%
  Necromantic 49.85%
  Norse 49.85%
  Dwarves 49.68%
  Skaven 48.80%
  Orc 47.33%
  High Elves 47.07%
  Khemri 46.64%
  Slann 46.40%
  Humans 45.90%
  Chaos Pact 44.86%
  Nurgle's Rotters 43.67%
  Vampires 43.31%
  Chaos 42.65%
  Underworld 42.13%
  Halflings 33.20%
  Goblins 30.26%
  Ogres 30.04%

Oh look, one of the non-timmy teams with the second best win%.

Well the below is an attempt to show the move between the unfiltered list and the filtered one...

• Team UnWin% Adjusted% change in postion/
  Wood Elves 56.26% (55.54% 1/1)
  Undead 55.64% (55.48% 2/2)
  Dark Elves 54.80% (53.14% 3/4)
  Lizardmen 54.32% (53.62% 4/3)
  Amazons 53.21% (51.85% 5/5)
  Elves 52.58% (50.75% 6/6)
  Norse 52.55% (49.85% 7/9)
  Dwarves 51.63% (49.68% 8/10)
  Chaos Dwarves 51.36% (50.40% 9/7)
  Necromantic 51.14% (49.85% 10/8)
  Skaven 50.61% (48.80% 11/11)
  High Elves 49.37% (47.07% 12/13)
  Slann 49.16% (46.40% 13/15)
  Orc 48.67% (47.33% 14/12)
  Humans 48.51% (45.90% 15/16)
  Khemri 46.88% (46.64% 16/14)
  Chaos Pact 46.16% (44.86% 17/17)
  Nurgle's Rotters 45.81% (43.67% 18/18)
  Vampires 45.58% (43.31% 19/19)
  Underworld 44.67% (42.13% 20/21
  Chaos 44.13% (42.65% 21/20)
  Halflings 36.04% (33.20% 22/22)
  Goblins 33.49% (30.26% 23/23)
  Ogres 32.21% (30.04% 24/24)

How do you get the list function to display columns?

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.

Does the chart match the % of overall tourny wins a team has?

mubo wrote:Thinking about it... the home team bias is probably an artefact of tournaments that do not use score? When they manually upload the results, TOs may list the winning team first? Any other reasons?

I think this might be likely. Even when using Score!, many historical events will have entered results manually (I did a whole NAFC once this way). In this situation natural human bias might lead to the winning team being listed first, I can certainly imagine doing this myself.

Joemanji wrote:

mubo wrote:Thinking about it... the home team bias is probably an artefact of tournaments that do not use score? When they manually upload the results, TOs may list the winning team first? Any other reasons?

I think this might be likely. Even when using Score!, many historical events will have entered results manually (I did a whole NAFC once this way). In this situation natural human bias might lead to the winning team being listed first, I can certainly imagine doing this myself.

Cool, I think it's the most likely explanation too. I can check this 100% if you could give me a few tourneys that were entered manually and some that definitely used score.

I still think it's bizzare though how often lunchmoney was the 'home' player.

mubo wrote:I still think it's bizzare though how often lunchmoney was the 'home' player.

I like the left....

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...