Under what circumstances should a beatloop be broken?
When beatloops are removed, it means that data about team relationships are being taken out of the graph.
So far, beatloops have only been broken if smaller beatloops with common teams were removed first. This means the system is fairly liberal about removing beatloops.
But I’ve started experimenting with another reason to break beatloops: when a beatpath contradicts a portion of a beatloop.
Here’s an example. Denver has been in a beatloop with NE and MIA: DEN=>NE=>MIA=>DEN . As a result, Denver hasn’t been given credit for its win over New England.
New England has an alternate beatpath to Miami, but that doesn’t mean much. It just confirms something that was already in the beatloop. Reinforcing a portion of the beatloop doesn’t help us.
But, what’s more interesting is that Denver has an alternate beatpath to Miami. That contradicts part of the beatloop. Since Denver has managed to develop an alternate beatpath to a team that it had earlier lost to, it essentially means that it has exposed that earlier loss as a fluke. If you break the beatloop at that portion, you have a beatpath segment to add back into the graph: DEN=>NE=>MIA .
By breaking contradicted beatloops to account for fluke losses, we end up with slightly more beatpath data, and slightly taller graphs. We’ll also end up with slightly more dynamic power rankings, as losses that seemed flukey a week earlier end up seeming not so flukey anymore.
Here’s the post-Thanksgiving beatpath graph with “beatflukes” removed:
The main differences between this graph and the other post-thanksgiving graph?
- Denver sheds its fluke loss to the Giants, regaining its victories over San Diego and Dallas
- Denver sheds its fluke loss to Miami, regaining its victories over New England, the Jets, and KC
- Tampa Bay sheds its fluke loss to SF, regaining its victory over Washington
- Tampa Bay sheds its fluke loss to the Jets, regaining its victories over Buffalo and Atlanta
- Carolina sheds its fluke loss to New Orleans, regaining its victories over MIN and GB
- Jacksonville sheds its fluke loss to St. Louis, regaining its victory over Seattle.
I’m still deciding whether to switch to this beatpath graph style, but it looks pretty good so far. Tampa Bay really climbs up the rankings with this graph, and Seattle gets hurt.