VoodooMike
Joined: Nov 07, 2010
|
 Posted:
Dec 02, 2011 - 19:45 |
 
|
I recently got around to looking at the data scraped from the FUMBBL api by hitonagashi, which covers the games played since CRP was implemented until the end of October. Just using simple, descriptive stats the tiers in blackbox seem to be:
Tier 1
Amazon
Chaos Dwarf
Dwarf
Lizardman
Skaven
Tier 2
Chaos
Chaos Pact
Dark Elf
Elf
High Elf
Human
Khemri
Necromantic
Norse
Nurgle
Orc
Slann
Undead
Wood Elf
Tier 3
Goblin
Halfling
Ogre
Underworld
Vampire
Though, of course, it comes down to the question of "how do you make the divisions?". In this case, I broke the races down into "good against" and "poor against" when it comes to other races, defining good against as having a 50% or higher win percentage against that race, and poor as having less than that. Win percentage is based on Galak's win percentage method of wins counting as 1, losses counting as 0, and draws counting as 0.5.
The tiers are then defined by saying Tier 1 races are poor against 5 or fewer races, Tier 3 races are good against 5 or fewer races, and Tier 2 races fall between those two extremes.
Obviously this is specific to blackbox, though, since the rating difference between races is always relatively small.
I also mapped each race's success (in general) across the team values that had enough information to actually use for that purpose, in order to get some idea of where each race's "sweet spot" might be under these circumstances, and very few races have one - most seem to either do worse as TV increases, or do better as TV increases, or remaining relatively constant. I even made graphs for that information, since everyone seems convinced that graphs actually give useful information.. but its not practical to put 24 images into the post.
Anyway, the main reason for posting this is that, as usual, the numbers don't quite match the past declarations... in this case, regarding team tiers... at least as far as blackbox goes. |
|
|
blader4411
Joined: Oct 18, 2009
|
| Posted:
Dec 02, 2011 - 20:13 |
|
Dwarves stronger against more races than all the ClawPOMB teams save CDs?
Are you smoking something sir? 
-Blader |
|
|
Dolls
Joined: Jul 26, 2006
|
| Posted:
Dec 02, 2011 - 20:15 |
|
| blader4411 wrote: | Dwarves stronger against more races than all the ClawPOMB teams save CDs?
Are you smoking something sir?
-Blader |
Or maybe he is just using statistics? |
|
|
Azure
Joined: Jan 30, 2007
|
| Posted:
Dec 02, 2011 - 20:17 |
|
My first impression of the Tiers listed is that perhaps you would get different results if you only looked at TV 1200+ matches. If you are looking at total number of games - then you are including more games at lower TV than higher TV (under the assumption that all teams start at low TV, not all of them get to high TV).
Dwarves and Amazons are still very strong...but they tend to get crushed at higher TV...and then people retire the teams. So is the data you using really accurate?
You should give a bit more information about the data and methodology if you want to disprove current beliefs. |
|
|
Shraaaag
Joined: Feb 15, 2004
|
| Posted:
Dec 02, 2011 - 20:21 |
|
Indiscriminate use of statistics and data is questionable. |
_________________
 |
|
Purplegoo
Joined: Mar 23, 2006
|
| Posted:
Dec 02, 2011 - 20:22 |
|
I always welcome someone actually providing some numbers (or results derived from numbers) rather than just spouting things.
I suppose there are a lot of factors at work under the numbers that lead to these tiers. I would certainly jiggle things about a bit based upon experience, but the plural of anecdote isn't data (oh where are you, Torso?). I suppose it depends on who's coaching. Rats, for instance, are almost certainly not as good in the box as a whole as something like Woodies or Pro Elves, but what they do provide is a very good minmax opportunity to prey on the system. Likewise, Lizzies are excellent until every other game destroys a Saurus a turn. Etc. Etc.
Welcomed as these insights are, it's always worth a think as well. Just like any other set of numbers. |
|
|
koadah
Joined: Mar 30, 2005
|
| Posted:
Dec 02, 2011 - 20:43 |
|
|
pizzamogul
Joined: Jun 13, 2005
|
| Posted:
Dec 02, 2011 - 20:45 |
|
I'd love to see the results broken down into TV groupings, like 1,000-1,250, 1251-1500, 1501-1750 and so on. I believe that different races should jump into or drop out of Tier 1 according to their TV "sweet spot." |
_________________
"Don't expect mercy."
-Woodstock |
|
pythrr
Joined: Mar 07, 2006
|
| Posted:
Dec 02, 2011 - 20:54 |
|
| Dolls wrote: | | blader4411 wrote: | Dwarves stronger against more races than all the ClawPOMB teams save CDs?
Are you smoking something sir?
-Blader |
Or maybe he is just using statistics? |
indeed.
most clawpomb teams (bar cds) suck at low TV (unless they minmax hard) |
|
|
Beerox
Joined: Feb 14, 2008
|
| Posted:
Dec 02, 2011 - 20:57 |
|
Interesting, but TV grouping would be interesting-er! Many ways to look at it, I guess everyone has their opinion of how to skin a cat. <--odd phrase |
|
|
Sutherlands
Joined: Aug 01, 2009
|
| Posted:
Dec 02, 2011 - 20:57 |
|
The statistics are bad. A team may be "good against" all but 5 teams, but if those 5 teams are the 5 most populous races in the box, then it is not a "Blackbox top tier team" |
|
|
Wreckage
Joined: Aug 15, 2004
|
| Posted:
Dec 02, 2011 - 21:31 |
|
To me the list looks very interesting but I don't really understand the calculation or what system garak used.
Calculating 0 for loss and 0.5 for tie seems odd but i guess it always depends on compared to what. Divided by number of games?
Another concern the distribution has to me is: What means 'doing well' against some race? Having an average win % of 45% or more like 75%? When am I doing bad? |
|
|
Wreckage
Joined: Aug 15, 2004
|
| Posted:
Dec 02, 2011 - 21:36 |
|
| Sutherlands wrote: | | The statistics are bad. A team may be "good against" all but 5 teams, but if those 5 teams are the 5 most populous races in the box, then it is not a "Blackbox top tier team" |
And regarding that a bit harsh criticism:
I have been wondering about that too a bit. What made me think of an interesting idea.
If one was actually to divide the teams in tiers... why not look for maybe 3 teams that do well against everyone and then look for 3 teams that do well against the three previously chosen teams rather than random teams and put them all in a group  |
|
|
Sutherlands
Joined: Aug 01, 2009
|
| Posted:
Dec 02, 2011 - 22:50 |
|
| Wreckage wrote: | To me the list looks very interesting but I don't really understand the calculation or what system garak used.
Calculating 0 for loss and 0.5 for tie seems odd but i guess it always depends on compared to what. Divided by number of games?
Another concern the distribution has to me is: What means 'doing well' against some race? Having an average win % of 45% or more like 75%? When am I doing bad? |
40 wins, 20 ties, 40 losses would be 50%. 30 wins, 70 ties would be 65% (good). 30 wins, 10 ties, 60 losses would be 35% (bad). |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Dec 03, 2011 - 08:57 |
|
| blader4411 wrote: | | Dwarves stronger against more races than all the ClawPOMB teams save CDs? |
That is correct, yes, based on FUMBBL's own match results from nearly a year under CRP. Dwarves have have a 50% or higher win average against all but four races: Dark Elf (47%), Chaos (46%), Nurgle (45%), and Chaos Dwarf (44%).
| Azure wrote: | | My first impression of the Tiers listed is that perhaps you would get different results if you only looked at TV 1200+ matches. If you are looking at total number of games - then you are including more games at lower TV than higher TV (under the assumption that all teams start at low TV, not all of them get to high TV). |
I'm sure you'll get different results if you decide to limit yourself to special cases like only at specific team values - but what exactly is the point of doing so? Blackbox does not limit itself to specific TV ranges beyond being limited to specific TV *differences* between teams. Any results you got would have pretty limited external validity.
| Azure wrote: | | Dwarves and Amazons are still very strong...but they tend to get crushed at higher TV...and then people retire the teams. So is the data you using really accurate? |
The data really is accurate because we know the data to be true. What you really want to ask is "Will these teams all perform in this exact order at all team values" and that's not really a useful question to begin with - we know that no one team will always win in all cases, but we can look for trends across ALL games, and that's what you're getting.. the win percentages of ALL games of one race against another at ALL team values. Because of that, we can say those are the tiers (given that definition of tiers) rather than having to say "well this is the 1200 TV tiers..." which you have to say if you decide to exclude other cases.
| Azure wrote: | | You should give a bit more information about the data and methodology if you want to disprove current beliefs. |
See, you take all the wins and add them to half the draws, and then divide them by the total number of games... or do you think, maybe, I don't actually need to explain the methodology behind percentages? I did explain my definition for good or poor versus, and what criteria I used for the tiers.. so what is it you're asking beyond that?
| Shraaaag wrote: | | Indiscriminate use of statistics and data is questionable. |
Not half as questionable as using intuition to determine things, which is what the blood bowl community has always done in the past - including the BBRC.
| Purplegoo wrote: | | Rats, for instance, are almost certainly not as good in the box as a whole as something like Woodies or Pro Elves, but what they do provide is a very good minmax opportunity to prey on the system. |
Skaven win percentage: 56.04% +/- 1.3%
Wood Elf win percentage: 54.98% +/- 1.39%
High Elf win percentage: 50.34% +/- 1.72%
I suppose I'd need to ask you for your definition of "good" in this case to know whether skaven are or are not as "good" as the elf teams. If we're talking about winning a higher percentage of games, then certainly they're more "good"... they have a 50% or higher win percentage against more races than either, across all TVs... though wood elves are right on the dividing line between Tier 1 and Tier 2 (skaven have 5 races they get lower than 50% on, wood elves have 6.. high elves have 11)
| pythrr wrote: | | most clawpomb teams (bar cds) suck at low TV (unless they minmax hard) |
As it happens, most clawpomb races, with the exception of chaos, don't get a lot better as their TVs increase. Chaos does, in a huge way. Chaos Dwarves actually lose more often as TV increases.. though "more often" doesn't mean they drop below 50% overall. Chaos Dwarves are a monster team.
| Sutherland wrote: | | The statistics are bad. A team may be "good against" all but 5 teams, but if those 5 teams are the 5 most populous races in the box, then it is not a "Blackbox top tier team" |
That would show up in total win percentage, then. No race in what I listed as Tier 1 has less than a 50% win rating across all games played. Some Tier 2 races have higher than 50% as well, but I didn't base the tiers on overall win % specifically because that could be skewed by the distribution of races.
| Wreckage wrote: | | Another concern the distribution has to me is: What means 'doing well' against some race? Having an average win % of 45% or more like 75%? When am I doing bad? |
I gave the definition in the original posting. A race is considered "good against" another race if it has won 50% or more of its games against that race, and "poor against" if it has won fewer than 50% of its games against that race... with win % calculated based on galak's 0.5 for draws formula. |
|
|
|
|
| |
Conceding v Goblins/...
Search
