Purplegoo
Joined: Mar 23, 2006
|
 Posted:
Sep 19, 2011 - 20:29 |
 
|
I too can't really say I play them often at all.
I don't think they're that prevalent, you just remember them when they smack you around a bit.  I really fancy a crack at Chuck Vs. BB for instance, we've just never crossed paths, I suppose. No Sure Hands currently, I note. Yummy!
Good to see this is going in a constructive way. |
|
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Sep 19, 2011 - 21:00 |
|
| Hitonagashi wrote: | | Without quoting the lot of that, as there's rather a lot there; I think Bloodbowl is balanced. This makes me a singularly bad person to come up with a definition for OP. |
Well, as the one doing the research (or at least proposing it) you have to come up with a working definition - its just the definition that you'll be using in your research. People can argue that it was a bad definition later, but at least you have something concrete to start with.
| Hitonagashi wrote: | | Onto your suggestion, I really like it, but I'm a touch concerned about whether we need to race limit it. As Bardazur said earlier, the outcome of a game depends on the race matchup as well as the casualty matchup. We can't analyse over all teams, because there are around 10x as many chaos/chaos games as there are chaos/elf games, which would skew the data heavily. |
I think we specifically do not want to examine the differences between race matchups beyond limiting ourselves to games in which at least one team contained that skill combination. Why? Because we're not really interested in whether that skill combination makes it more likely you'll win against certain races, but whether it makes it more likely you'll win in general.
To say certain skills give you an advantage over certain races is... almost not worth saying. Most skills are a benefit over some races more than others, but to say a skill (or combination of skills) is "too good" is really to say that it gives you an unfair advantage.
If there is a strong relationship between casualties caused and winning a match, then it should exist even if both teams have that combination, right?
| Hitonagashi wrote: | | I'm not a statistician, but it seems to me that that output over all teams and games isn't as useful if the racial balance of those teams is skewed. We could take a representative sample of the teams to apply it? (so say, a random sample of 30 with greater than 15 games teams for each race) |
Again, this is why you want a clear hypothesis. If you're interested in looking at the relationship between the presence of ClawPOMB and victory over dwarves, then sure, we want to split out that race and look at those results. If you want to examine the relationship between its presence and wins in general, then you don't want to analyze things on a per-race basis at all.
Personally, I don't think it'd be useful to say "ClawPOMB helps chaos beat dwarves at high TVs" - it wouldn't really affect the ClawPOMB argument in any relevant way. |
|
|
polardragon
Joined: May 07, 2011
|
| Posted:
Sep 19, 2011 - 23:12 |
|
Interesting thread.
If I am not mistaken, the one thing the game stats won't tell you is how many KO's were caused. Casualties will get you spp, but an entire team KO'd till then end of a drive is equally effective for scoring points.
I am sure there is a clever way to extrapolate that information in a well built model. I just don't know how it would be done. |
_________________
  |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Sep 21, 2011 - 01:53 |
|
Well, I decided to take a quick look at the relationship between casualties and touchdowns. To do so I took the first 5000 matches and then checked for correlation between relative casualties and relative touchdowns - that is to say, subtracting team 1's casualties from team 2's, and doing the same with their touchdowns.
There is a weak-to-moderate relationship between them, based on those 5000 matches (r = 0.32, p < 0.01) which translates into roughly 10% of the variance in relative number of touchdowns being attributed to the relative number of casualties.
Now, that doesn't mean bashy teams are necessarily king - it may simply mean that aggressive play tends to be rewarded. That said, the better your ability to cause casualties, the more you're likely to cause, and there certainly is a demonstrated relationship between casualties caused and winning games. Given the game, that should be true, just not in any overwhelming sense.. and I don't see it being overwhelming based on those 5000 matches.
| polardragon wrote: | | If I am not mistaken, the one thing the game stats won't tell you is how many KO's were caused. Casualties will get you spp, but an entire team KO'd till then end of a drive is equally effective for scoring points. |
This seems to be true - there is no information in the match records about KOs. That said, the relationship between casualties and KOs should be ridiculously strong (if you're getting casualties, you're getting in blocks, and you're likely getting more KOs than casualties... but the amount should be linked) so if we look at the casualties/td relationship we can assume that relationship extends to KOs as well.
Now, again, this is a pretty off-the-cuff analysis - I *believe* it includes all matches, which means it includes weird alternative rule groups where casualties might be less likely (like BB7). I'm certainly no pro at using the fumbbl API, so... if someone can suggest ways to get more precise info out of the api, I'm all ears! |
|
|
Hitonagashi
Joined: Apr 09, 2006
|
| Posted:
Sep 21, 2011 - 02:17 |
|
Nice work! 
Yep, it does include all matches though. If you need to get a division, you need to follow the team id for each team, and the team info has a "division" tag, which is 10 for the Box, and 1 for Ranked. That's why my dataset grabbed all games and all the teams that played in them, so I could track down division.
For reference, from my records, 80.9% of the games played since January have been in "competitive" divisions (R and B), so your conclusions are probably fairly strong.
I'm going to get back to my data sometime soon, but I'm fairly busy for the rest of the week (and Aftermath this weekend), so I'll get onto it when I have a spare minute. |
_________________
http://www.calculateyour.tv - an easy way to work out specific team builds.
 |
|
VoodooMike
Joined: Nov 07, 2010
|
| Posted:
Sep 21, 2011 - 02:42 |
|
Yeah, in order to turn the XML records available into the data needed to do more precise analyses the data would have to be turned into multiple database tables. I know you've done that with mongoDB, but I'm not familiar with that (I use mysql). We'll see how it goes.
It should be noted that when the number of matches is expanded from 5,000 to 25,000, the relationship is entirely unchanged. Until the box/ranked matches can be separated out, we can just run with the above numbers. |
|
|
|
|
| |
Borg Invasion
Search
