New MVP rule and race win percentage

[garion]

So I was wondering what is considered a win percentage that is too high for a race?
By this I mean a win percentage considered higher than the brackets are meant to go.

Here are the stats from CRP - http://forum.bloodbowl-game.com/viewtop ... 988#p58988 scroll down to Dodes post.

So far the two races that have the biggest win percentage appear to be Lizardmen and Necromantic
Lizardmen 3314 games, Win pecentage = 59.44%. Percentage increase from CRP: 4.65%
Necromantic 4986 games, Win pecentage = 57.23%. Percentage increase from CRP: 3.65%

The two changes effecting this are Piling On removal, and picking 3xMVP.

I'm not saying any team is broken and its still only 2 years we have had these rules for. But is this a cause for concern? These are big win percentage increases on what were already among the best teams in the game. The previous samples from CRP were from approximately 20,000 games per team. We dont have that data for 2016 edition yet.

Obviously removal of PO from the kill stack is effecting things, however I would argue the two races mentioned above is more to do with picking 3 MVP. Quick skilling up of Gollems and Saurus is absolutely huge on what were already top end tier 1 teams.

Thoughts?

[lunchmoney]

Cant view the other side of that link whilst at work, but you state it's CRP stats/win rates.
Where did you get the Lizard and Necro game stats, that you quote, from?

[garion]

its from a fumbbl script -

Blackbox CRP
Race Record
RACE Games Played Win Percentage
Amazon 15803 58.29
Lizardmen 22821 54.79
Undead 25819 54.15
Wood Elf 21288 53.85
Necromantic 32264 53.58
Skaven 24268 53.37
Chaos Dwarf 47759 53.36
Dark Elf 23853 51.89
Elf 10216 51.34
Dwarf 37988 50.99
Norse 24140 50.84
Chaos 61314 49.73
Human 31450 49.3
Nurgle 39347 48.89
Chaos Pact 29343 48.89
Khemri 19632 48.77
High Elf 11966 48.72
Orc 39449 47.19
Slann 10782 46.08
Vampire 10413 43.96
Underworld 11646 42.92
Halfling 5419 29.58
Goblin 8130 28.52
Ogre 7320 27.43

Blackbox 2016 edition
Lizardmen 3314 59.44
Necromantic 4986 57.23
Amazon 1940 56.68
Undead 3707 55.56
Wood Elf 2494 54.53
Skaven 3172 53.66
Dwarf 4754 53.38
Chaos Dwarf 5431 52.68
Elf 2235 52.3
High Elf 2865 51.73
Khemri 3361 51.19
Dark Elf 3177 50.83
Human 3550 50.75
Norse 2708 50.24
Orc 5195 50.07
Chaos 48297 47.49
Slann 1537 45.87
Chaos Pact 2286 45.36
Nurgle 4190 45.33
Vampire 2345 45.16
Underworld 2454 43.03
Halfling 1125 34.22
Goblin 2180 29.61
Ogre 1559 27.2

[plasmoid]

Hi Garion,
I'll save you the wait and give you the 3 most likely brush-offs straight way:

1) You need to calculate the CI95 intervals. The numbers you are listing are not statistically reliable.

2) Black Box is an environment which is very different from proper League play. If there is a problem in Box, don't change the rules, change the Box. Box stats tell us nothing about balance in League play.

3) The listed win percentages are very likely not to refelect actual power, but instead to reflect the composition of the Box meta. So what you ought to do is correct for composition. However, to correct for composition, you'd have to look at performance against each opposing race - giving you much smaller data sets, and thereby much larger confidence intervals - thereby making it impossible to correct for composition. I.e. "you know nothing, Jon Snow... or at least you can't prove it".

Cheers
Martin

[Mystic Force]

What seems weird to me after a cursory glance through both tables is that so many teams went up, by significant amount (more than 1%) but only 4 teams, including all flavours of chaos went down significantly, making me wonder how the math balances. I will have to actually check the numbers to see if both sets average out to 50% but i think they might not.

[garion]

@ plasmoid -
hehe

you're right, someone will no doubt say some or all those things.

But these stats are what they are, we have a comparison from CRP to 2016 edition, so take from it what you will.

I also never liked the 3 MVP rule change (though it does make sense when coupled with seasons, which I'm also not a big fan of), so I have confirmation bias to add in there.

I also never said there is an issue. I was just wondering what the designers would consider a problem, win percentage wise. I know CRP was designed with teams to work inside specified win percentage brackets and it did so looking at data from blackbox, Cyanides matchmaker and TT tournaments. Can you remember what was considered too high back then?

@ Mystic Force - the numbers are correct.

[sann0638]

What I take from that - CHAOS PLAYED 48000 GAMES!

Box people are crayzee.

[king_ghidra]

Obviously removal of PO from the kill stack is effecting things

Seems unquestionable.

Blackbox CRP
Race Record
RACE Games Played Win Percentage
Lizardmen 22821 54.79
Dwarf 37988 50.99
Chaos 61314 49.73
Nurgle 39347 48.89
Chaos Pact 29343 48.89
Orc 39449 47.19


Blackbox 2016 edition
Lizardmen 3314 59.44
Dwarf 4754 53.38
Orc 5195 50.07
Chaos 48297 47.49
Chaos Pact 2286 45.36
Nurgle 4190 45.33


The notorious clawpomb teams dropped off, by 2-3%, and the non-clawpomb AV9 teams rose by similar or greater numbers. I feel like Occam's razor applies at that point.

I don't know how much it's possible to break down the stats by TV, but that would be the really interesting bit in terms of gauging the value of the MVP argument. You would imagine that the win% in the c. 1050-1300 kind of range would be most significantly varied were that the case, as those teams that can really leverage it see the benefits. And did those clawpomb teams that used to dominate vanilla av9 at higher tv see their win% drop off particularly in that range?

Interesting stuff though, no doubt.