Tie breaker system

Everything dealing with the video game developed by Cyanide!
logain
Posts: 303
Joined: 11 March 2012, 08:34

Re: Tie breaker system

Postby logain » 24 December 2015, 08:50

I think Logain(?) was the other one that said their league had experienced the same thing and was also trying the 1/2 difference.
Indeed that's what was decided, still better than full CAS which was the case before, but i did not agree with that decision.
On my BB2 league i rely on TD difference only.

Looking at dode's stats CAS should count as 0.16, so better not counting it.

User avatar
dode74
Posts: 7040
Joined: 11 December 2008, 11:18
Location: Nr. Reading, UK
Contact:

Re: Tie breaker system

Postby dode74 » 24 December 2015, 10:42

No, the stats suggest using cas difference at all will likely give an advantage to bash teams, whereas TD difference does not advantage bash over agi or vice versa. For that reason I think it is better to not use cas difference at all.
Image

User avatar
- Karma -
Posts: 473
Joined: 23 September 2015, 15:16

Re: Tie breaker system

Postby - Karma - » 26 December 2015, 19:13

Interesting conversation all.

Spydyr, obviously GOBBLN is your (and the other Overlords') league so you want to keep CAS difference in some form would it be worth considering a value of .25 and decide how to go about rounding those off?

I mean honestly, from our standings after one season it looks like there will be hardly any (if any) playoff spots determined by TD/CAS difference tiebreakers. BUT then I can see where it is needed in determining seeding tiebreakers. I'd be willing to help with the math for GOBBLN PG as a test bed - we could add several different values and just see how much impact, if any it would have made. PM me on the league site if you want help with that.

I guess what I'm really wondering is, other than missing the playoffs entirely due to a tiebreaker on TD/CAS difference, is going up or down a seed spot one or two places that huge? It's not like "home field" advantage is that noticeable, do you think? For our upcoming playoffs for example I'd have no issue giving the team I'd play home field even though that'd be 16 spots worth of seeding I'd give up. Or does it become more so in your all's experience once more teams start to purchase stadium upgrades?
Last edited by - Karma - on 26 December 2015, 19:15, edited 1 time in total.

User avatar
- Karma -
Posts: 473
Joined: 23 September 2015, 15:16

Re: Tie breaker system

Postby - Karma - » 26 December 2015, 19:14

double post

User avatar
DarkTechnomancer
Posts: 5
Joined: 11 March 2013, 08:23

Re: Tie breaker system

Postby DarkTechnomancer » 18 January 2016, 18:40

No offense to Dode, I have seen your stats and they are very comprehensive, but I can't agree with the idea that TD differential doesn't bias agility teams. The other commissioners in the Legion League and I have discussed it at length and gone over a lot of statistical data to try to decide on the best tie-breakers to use while keeping things as fair as possible.

Consider this graph using stats taken from the NAF, a sample size of over 140,000 games played worldwide.
Image
Even with just a simple scatter plot, it is clear that there is a distinct correlation between match-point rating and TD differential. However there is also a clear pattern of teams with access to AG4 players having TD differentials much higher than other teams at similar ratings. Some of the most profound gaps at similar ratings are between High Elves and Khemri, and between Wood Elves and Undead.

For what it is worth, in Legion we use the average match-point rating of defeated opponents as a tie-breaker in place of TD differential, and it seems to have served pretty adequately, even helping us to resolve a potential 3-person tie at the end of our current season.

User avatar
dode74
Posts: 7040
Joined: 11 December 2008, 11:18
Location: Nr. Reading, UK
Contact:

Re: Tie breaker system

Postby dode74 » 18 January 2016, 22:36

You're using NAF data (140,000 games under which ruleset?), I'm using MM data (110k games under one ruleset). You're limited to very low TV only in a resurrection format and using means only to eyeball a scatter plot, while I am using a wider range of TVs, full injury rules, and conducting a t-test. Make of that what you will.

Edit: Here's a like-for-like chart for you, removing the non-T1 teams. Again, it's eyeballing rather than a proper statistical test, but you can see that while there is a correlation between win percentage (x-axis) and TD difference (y-axis)*, there is no big advantage for AG4 teams.
Image

* duh! ;)
Image

User avatar
DarkTechnomancer
Posts: 5
Joined: 11 March 2013, 08:23

Re: Tie breaker system

Postby DarkTechnomancer » 18 January 2016, 23:24

That is fair. I should have separated the LRB6 results. That was an oversight on my part. That said, the results are not significantly different when filtered to include only LRB6 games, as they make up the vast majority of the results anyway. Admittedly, my analysis is far less thorough but the patterns that I see always point to Bash teams having more consistent games while Agility teams have greater scoring potential. At the very least, most players I have spoken to have not been satisfied with the idea of using TD differential as a tiebreaker.

User avatar
dode74
Posts: 7040
Joined: 11 December 2008, 11:18
Location: Nr. Reading, UK
Contact:

Re: Tie breaker system

Postby dode74 » 18 January 2016, 23:53

I don't find it massively surprising that the agi teams, which tend to take more cas, do better (both in terms of wins and, given the correlation, TD difference) in a resurrection environment. We're not talking about resurrection environments, though, and I find things like the reversal of the WE/Undead dynamic to be fairly telling of there being a difference in those environments.
the patterns that I see always point to Bash teams having more consistent games while Agility teams have greater scoring potential
I think the agi teams are streakier, but the data from MM doesn't point to that scoring potential leading to TD difference advantages. I'll try to do the same with OCC data soon (TM).
Image

User avatar
DarkTechnomancer
Posts: 5
Joined: 11 March 2013, 08:23

Re: Tie breaker system

Postby DarkTechnomancer » 19 January 2016, 00:10

I didn't think that all NAF tournaments were resurrection format, but I have little experience with them except for a few emails from a league I used to participate in. But if they are really that common, it certainly would skew the results. I hesitate to assume, but if by MM you mean matchmaking then I would not want to use those statistics, myself. Ladder formats are significantly different from scheduled leagues, and obviously we would only care about tiebreakers in a scheduled league format.

User avatar
dode74
Posts: 7040
Joined: 11 December 2008, 11:18
Location: Nr. Reading, UK
Contact:

Re: Tie breaker system

Postby dode74 » 19 January 2016, 00:12

The vast majority of NAF tournaments are resurrection. And while I agree that MM is different to leagues, I'd say it's far closer in terms of consequence than a res tournament is. As I say, I'll have a play with the OCC (large scheduled league, 24k+ matches) data I have when I get the chance.
Image


Return to “General”

Who is online

Users browsing this forum: No registered users and 2 guests

cron