Reconsidering the ruling on Khorne?

[dode74]

It's NAF data so it's tournaments.

I've not suggested buffing or nerfing anything, certainly not based on that data: it doesn't support the idea.

[GalakStarscraper]

Just a note ... I realize that it was just NAF data ... but it was good to see that 19 of the 24 teams landed where we tried to get them and the other 5 are slightly out of place (Wood Elf (+), Undead (+), Vampire (+), Chaos (-) and Halflling (+)).

[koadah]

It's NAF data so it's tournaments.

I've not suggested buffing or nerfing anything, certainly not based on that data: it doesn't support the idea.

Ah of course. I must have been seduced by the idea that the NAF was already the custodian of all Blood Bowl.

NTBB does nerf woods though.

In terms of proper statistics no it doesn't support anything. In terms of gut feeling, eye balling, Woods & undead look as though they could do with a nerf for short res tournaments.

Nonsense you say, they are only 56.25 & 55.64. But gut feel says if they do well in their first three games they are more likely to be playing vs woods or undead in the last three. That would make them underestimated. What does the data tell us about that?

The data does tell us that only orcs were taken more often than woods/undead and orcs are 'free' in the box.

One ruleset to rule them all? Hmm, I don't know.

[straume]

Does the data take into account tiered tournaments?

[dode74]

koadah - NTBB is not my thing

Nonsense you say, they are only 56.25 & 55.64. But gut feel says if they do well in their first three games they are more likely to be playing vs woods or undead in the last three. That would make them underestimated. What does the data tell us about that?

First, the numbers are wrong. They are not 56.25 and 55.64, they are 54.78-57.74 and 54.3-56.98. The mean is meaningless for the inferences you are making, only the range matters since we can't say where within the range (if it is in the range - 95CI remember) the actual mean is.
Second, the data says nothing, and will never say anything, about your gut feel.

straume - in all honesty I don't know. I don't even know if the NAF knows. It does say which tournaments the data comes from, so it should be possible to filter out the ones which are tiered ones if I can get info on which ones were. I'm not at my PC so can't do that right now but will look into it.

[Darkson]

Without looking at each individual rule pack you won't find that information as it's not something that you need to put in to get a tournament sanctioned ( same as when you asked about when LRB5 started).

[dode74]

I thought that might be the case, but I also thought there might be a good source for that stuff on this forum in the tournaments section. At least with the tournament names I have a start point, whereas without them I have nothing.

[babass]

Hi guys,
Babass - well, there's something I can respond to with the testing I'm doing. There are human teams in my current NTBB league. And I've played one myself. None of those had just 1 catcher. Nor did they only use them on short drives.

You are focusing too much on (your) league.
Humen catcher won't be more popular on tourney with such a change.

[Darkson]

Humen catcher won't be more popular on tourney with such a change.

Do they need more than 2 at tournaments?

[koadah]

Hi guys,
Babass - well, there's something I can respond to with the testing I'm doing. There are human teams in my current NTBB league. And I've played one myself. None of those had just 1 catcher. Nor did they only use them on short drives.

You are focusing too much on (your) league.
Humen catcher won't be more popular on tourney with such a change.

It is not about non-prog tourneys.

It is about better longevity so a better team in a longer league.

Yes, I definitely want at least two catchers.

[Chris]

Can't see it affecting tourney performance, too marginal and won't affect team make up much.

In leagues I assume it would keep them alive longer so they can develop, but they are still a gimped player...

[straume]

I thought that might be the case, but I also thought there might be a good source for that stuff on this forum in the tournaments section. At least with the tournament names I have a start point, whereas without them I have nothing.

If the goal is to measure racial strength, clearly the tiered tournaments must be removed. Also: the Swiss draw which I assume is used all over will make the assumed better teams face each other. Perhaps better to only add the first match of every tournament? Which is not tiered or have anything silly in the rules pack.

[plasmoid]

Hi Dode,
very nice chart. I wish I could do stuff like that.

On the NTBB site there is some similar math.
I went to Ian Williams original nafstats site, and pulled all games marked as LRB6 up to january 2014.
That's not quite double the games you have. But it is definately more.
I then removed mirror matches - because even though the BBRC didn't mention them in their tier balance definition - they are devoid of content and will only mask a teams true performance.
Even more so in Swiss tournaments. For example more than 10% of the Undead's games are mirror matches!

Doing that puts both Undead and Wood Elfs completely over the top of tier 1.

And that's not even considering 2 actual sources of error:
1. The Swiss format - matching strong against strong and weak against weak will pull the stats for both towards the middle (50%).
Maybe VoodooMike can figure out how strong the effect is in a standard 6 game tournament.
2. More and more tournaments seem to be giving tier-dependant bonuses to teams, which again pulling team performance towards the middle.

Cheers
Martin

[dode74]

Martin

If you have match-level data from DoubleSkulls then could you share it? Because your site suggests you are using this data (also from DoubleSkulls), which is summaries rather than match level.

Pulling out mirrors is contrary to what the BBRC did. You rightly say that including mirrors will pull the win percentages towards 50 (since each will add a win and a loss or two draws to the stats), but if the definition includes mirrors then we should include mirrors regardless of our opinions on the matter. I can understand why they would include mirrors: teams are allowed to play against teams of the same race, so it is part of their lifetime performance.

Regarding your two sources of error:
1. I have no idea if the tiers were developed with the idea that tournaments would be swiss in mind. Do you know that they weren't? If so then it's not a source of error at all.
2. Yep, and that's true of your own data too. Without knowing which tournaments use tiering all we can say is that our ranges are probably too small to account for that error.

[Vanguard]

1. I have no idea if the tiers were developed with the idea that tournaments would be swiss in mind. Do you know that they weren't? If so then it's not a source of error at all.

I was under the impression that CRP was developed with reference to Open Leagues, no other format was considered relevant? Or was that just JJ's view?