An interesting situation happened with Carolina this week. If you review the previous entry, you’ll see three of Carolina’s wins are currently considered beatflukes. This one is a little hard for me to defend, so I’m going to explore it a little more.
Look at the beatfluke graph again:
Normally there would have been the following beatloops:
CAR=>TB=>CIN=>CAR
CAR=>NO=>ATL=>CAR
CAR=>BAL=>CIN=>CAR
However, there is a key beatpath segment of TB=>PHI=>DAL=>CAR. BAL, NO, and ATL all have a beatpath to that segment. It means that all of those teams have an alternate beatpath to Carolina, aside from the beatloops above. Under the beatflukes rule, it means that you can’t argue that the relationship between those teams and Carolina is ambiguous anymore – these other teams are clearly better. So the above beatloops are broken at that part of the chain, restoring the following:
TB=>CIN=>CAR
NO=>ATL=>CAR
BAL=>CIN=>CAR
(TB=>CIN is then taken out due to an entirely different beatloop.)
The basic story is that Carolina is losing the games it needs to win. What’s the upside? Well, the longer a beatpath segment, the more opportunity there is for a beatloop to occur in that segment. BAL=>NO=>ATL=>TB=>PHI=>DAL=>CAR is very long, and all that needs to happen is for any downstream team to beat an upstream team, and all of a sudden these beatfluke victories won’t look so flukey anymore.
However, even if these fluke loops are restored to the graph, it doesn’t mean that Carolina’s placement in the graph improves. Here is how the graph looks if beatflukes aren’t taken out, and all the loops are introduced into the graph (this is the original beatpaths graph style, before we started using beatflukes):
Carolina looks slightly better, but the restored beatpath is just CLE=>OAK=>PIT=>MIA, not exactly impressive. And in the power rankings, it doesn’t make as much difference as you’d think. In the normal variant, Carolina would be ranked #25. In the beatflukes rankings (coming out soon), they’ll be ranked #29.
So essentially, Carolina is suffering more from their losses to “poor” teams than their defeats of “good” teams. They can do a lot to help themselves in the second half, especially if they can win their remaining divisional games vs TB, NO, and ATL. An upcoming game vs PHI is another place where they could help themselves.
The one that is stranger to me Tampa Bay’s high position on the graph, given they only have 2 wins so far this season. I know it’ll shake out some in the rankings though, so I look forward to seeing those. I’m not quite ready to go do the digging into what happened in week 9 to restore TB’s win vs PHI to the graph, which really seems to have played a big part in all of this.
Yep, that TB/PHI game is clearly the source of the problems. The obvious solution is for Carolina to beat Tampa this Monday. If that happens, then the previous TB=>CAR game is no longer a beatfluke, and the TB=>PHI game is likely to be beatlooped away. Carolina will still not get credit for their wins over NO or BAL, but they will lose their losses to PHI and DAL, and will regain their wins over TB and CLE.
All in all, order will be restored. And if, on the other hand, Carolina loses to TB this week, then maybe they aren’t that good after all.
This week’s power rankings explain the Tampa Bay situation a little bit.
Now with the results of week 10, how will this affect CAR (asking before the results of CAR-TB). PIT forms a beatloop with the NO-ATL (PIT>NO>ATL>PIT) But NO has an alternate beatpath to NO (NO>CLE>OAK>PIT). But PIT has an alternate beatpath to NO (PIT>MIA>KC>SF>MIN>CAR>NO). With the wild results in week 10, there could be more, and probably should have been even more ‘upsets’ than that.
It’s pretty crazy. TB is still favored over CAR, but CAR is now within three power rank slots of TB (before the NYG/CHI results), which brings up the whole thing about how sometimes the picks are more accurate if you give the home team a 2-3 slot boost…