Dolphin Power Rankings after Week 3: predictions

[Dolfanalyst]

LOL ... It's hard to fathom some people actually believe that results don't matter. Kim Bokamper from those dreaded Killer Bs of the Dolphins said that was a philosophy they followed. Make them go the length of the field and keep the big plays from happening. The more plays they ran the more opportunites our defense has to capitalize on their mistakes.

Sure seems like that's is how defense plays too. We'e won 3 big games in 3 different ways. I'd think that would count for something.

Results certainly matter, because the playoffs are determined almost exclusively by record, but results aren't stronger predictors of future performance than other variables.

Last year Alabama beat Georgia convincingly in the SEC Championship game, yet Georgia was favored over Alabama in the National Championship game just a month later.  Obviously there were better predictors of both teams' future performance in that game than merely the result of the game they played previously.

[Spider-Dan]

That logic works both ways.  Did those same predictors say that Georgia would beat Alabama in the first game? You can't just ignore half the results and point to the remaining half as "proof" of validity.

Did these predictors say that KC would beat CIN in the regular season last year?  How about in their AFCCG rematch?

[Dolfanalyst]

That logic works both ways.  Did those same predictors say that Georgia would beat Alabama in the first game? You can't just ignore half the results and point to the remaining half as "proof" of validity.

Did these predictors say that KC would beat CIN in the regular season last year?  How about in their AFCCG rematch?

I'm not saying anything about the degree to which such predictor variables are accurate -- I'm merely saying there are variables in that regard more accurate than team win percentage over a relatively small sample of games or the result of a single previous game between two teams.

Consider that right now Buffalo and the Giants are both 2-1, yet the expectation would be for Buffalo to all but trounce the Giants.  That expectation has to be determined by something other than their records.  And if they actually played, the Giants might win the game, but if they played again the following day Buffalo would be favored again nonetheless.

It helps to pan back and ask yourself which team would win the majority of games if they played 100 times.  If the Giants played Buffalo and by chance beat them, and that represented one of the perhaps 10 times out of 100 they would beat them, well then Buffalo remains clearly the better team.

Again none of that matters in terms of who reaches the playoffs, because there again we're using team records almost exclusively -- this pertains again only to which variables are best used to determine the better team and the one expected to win.

[Spider-Dan]

When Eli's Giants played Brady's Patriots in SB XLII, the predictors said NE was favored.
When they played again in the 2011 regular season, the predictors said NE was favored again.
When they played in SB XLIV, again NE was favored.
No matter how many times Eli's Giants beat Brady's Patriots, Brady's Patriots were always the favorite.

The idea that the prediction industry is "accurate" is far from substantiated.

[Dolfanalyst]

When Eli's Giants played Brady's Patriots in SB XLII, the predictors said NE was favored.
When they played again in the 2011 regular season, the predictors said NE was favored again.
When they played in SB XLIV, again NE was favored.
No matter how many times Eli's Giants beat Brady's Patriots, Brady's Patriots were always the favorite.

The idea that the prediction industry is "accurate" is far from substantiated.

Prediction of single game outcomes is difficult using any variables, but there are variables that do it far better than win percentage over a relatively small sample of games and the outcomes of previous meetings between teams.  This is again why Buffalo would be favored against the Dolphins right now, even though the Dolphins are 3-0 and Buffalo 2-1 and the Dolphins beat them four days ago.

Again the larger point is that if you're generating power rankings as a measure of the "best" teams and not the "most deserving" teams, you do it with more than just win percentage and/or H2H outcomes.  Otherwise you would simply put money on the Giants "tying" the Bills in a game between them, since they're both 2-1, whereas such a bet would obviously be ridiculous.  There's obviously a far better way of determining what's most likely to happen in a game between Buffalo and the Giants.

[Spider-Dan]

Prediction of single game outcomes is difficult using any variables, but there are variables that do it far better than win percentage over a relatively small sample of games and the outcomes of previous meetings between teams.

Can you cite someone who has proven over time to be better at predicting the outcome of a game than our boy George Washington on a standard-issue United States quarter dollar?

The entire industry is a scam.

[Dolfanalyst]

Can you cite someone who has proven over time to be better at predicting the outcome of a game than our boy George Washington on a standard-issue United States quarter dollar?

The entire industry is a scam.

If we order 100 2021-2022 regular season and playoff games randomly and look at the percentage of favored teams that won, that percentage is 63%.

If you would've bet let's say $100 on each game and used just a coin flip, instead of the Vegas favorite, to determine the teams bet on (win or lose), you likely would've lost about $1300.

[Spider-Dan]

That's cool and all, but when talking about actual outlets that exist in the real world, I doubt any of them simply predict chalk down the line... especially since (in the early weeks) either you have very little data to work from, or you're using last season's data with outdated rosters.  So in practice, you would need to offset that 63% with the number of games that a given outlet predicts an upset that doesn't happen.

This is why I asked for a specific "someone who has proven over time to be better" rather than stats on amorphous favorites.

As a side note: please don't derail this conversation into yet another discussion about how Vegas works.  I'm not talking about Vegas; I'm talking about the analysis industry who claims to be able to predict the outcome of games, many of whom charge delusional marks a fee for their insight.