Making a better power ranking

Hey so thought I would start this thread as the power rankings on CFL/TSN/3Down/etc. tend to just be blah and wanted to see if it would be possible to create a more number driven/statistical model for the weekly power rankings.

I'm going to reach out to the CFL.ca and see if they have any master lists of stats for as far back as possible but in the meantime anyone have any suggestions for any type of information that should be included feel free to comment and we can go from there.
Also if there is anyone out there with a background in statistics/sports statistics I would love the help as my experience is of a more geographical background so there might be times when I'm not using the data perfectly.

1 Like

I'm with as long as we still get to argue, discuss, & bicker over the results. (otherwise boring!)

2 Likes

Well of course if I didn't get to argue I'm pretty sure my head would explode

1 Like

Ok so then first question (maybe hardest to answer) is the power ranking

  1. a weekly ranking based on the most likely place for each team to finish in the standings/win the grey cup
  2. a weekly ranking based on which team has the most likely chance of winning the following week with those teams that are ranked higher having a more dominant win

As you know from the thread on where this was hatched, I have some thoughts on this and am willing to help but don’t have s background in statistics like you do.

I won’t even try to get into details, other than pointing out that #1 you stated above needs some refinement. For example, where a team might finish in the standings and their chance of winning a Grey Cup might be completely different. We can sort that out later but I think we want to apply statistics to the teams based on various criteria and taking into account past performance and future predicted performance with weighted values assigned to various stats such as wins, points/points allowed, strength of opposition and I am sure some more.

Great that you are reaching out to CFL.ca. I suggested early we might want to examine how golf does their rankings. Their system is probably the most accomplished in the sporting world as it drives everything. It also just occurred to me that power rankings are done weekly by TSN for NHL, NFL and MLB. For hockey at least it is stats driven with an emphasis on predicting the immediate future. Maybe finding out how they do their rankings would be useful as well.

Might be interesting if we had a poll for who should be in first place. Have a points system for the results. 9 points for 1st, 8 points for 2nd etc. If Calgary, BC and Winnipeg all got votes and no one else did then you could do another poll for who should be in 4th place. Then see what the result are until you have the last pick.

Literally the opposite of trying to do a statistical method of coming up with a power ranking as what you are talking about is how the current power rankings are done via the writers/contributors to cfl.ca/tsn.ca/3downnation.com and they are meh

The reason I think of using the team's end of year standings is that it is something that, once I create some sort of model, I can then apply it to random sampling of years to see how close it is. I won't bother you with the way its done but its basically predictive modelling very similar to what insurance companies do.
I've heard back from the CFL.ca but am waiting for the guys in statistics to get back to me as so far all I have is the 2022 stats book which doesn't have a lot of useful information.

Sounds good. You’re the statistical guy. Someone just posted about the lack of meaning of the current power rankings and I directed him to this thread. Good luck