Interesting Take on Playoff Records for Coaches - PalmBeachPost

Rule #1: If you're going to jump head first into a debate ensure you actually know what the debate is about.

I agree. You have broken the rule.

 

Conglomerate stats I think aren't good predictors for individual games though.

Like the 57% HFA, which has held steady as you point out, but each team does not have a 57% chance of winning home games, some teams are higher, and some teams are lower, and then matchups also change the %.

For an extreme example, the odds of NE beating the Browns don't change much whether it's in Cle or NE.

It's kind of like shark attack, or plane crash stats, they don't mean a heck of a lot, as in there is a 1 in whatever chance you become a shark attack victim, not if I don't go in the water there isn't, lol, and for a guy who surfs in tiger shark territory, or the guy that surfs where there are seals and great whites, their chances are exponentially higher than that of someone who goes to the Jersey shore once a year.

When they do those type stats, they are basing it on population numbers, but there is a large % of people that have never even seen the ocean, much less been in it, and nearly all the shark attacks in the US all occur in Hawaii, Cali and Fla, and they are mostly surfers and divers.

Plane crash stats are misleading as well, the odds of a plane crashing going from Newark to O'hare are lower than the national avg, because most plane crashes in the US happen to Alaskan bush pilots, so obviously if you are an Alaskan bush pilot your chances of crashing are way higher, and the skill, experience and capabilities of the pilot factor into that as well.

Sports is the same, league wide predictors don't help much on individual games.

 

Got to remember what I was arguing against. Fame in post #21 said "The moral of the story here is that history can not be trusted and using past statistics to predict future outcome is 100% garbage. Statistics are not predictive."

lol.. 100% garbage? In math that would mean you can do no better than chance. Well 57% HFA is better than chance, meaning if you had truly even odds (no bookie) and you just bet on the home team all the time you'd make money over time.

All you're saying is that the 57% might not be good enough for your purposes. But that all depends on what your purpose is. If all you need to make money over many games is on average 56% accuracy per game, then 57% is good enough. If you need more, then it's not good enough and you use more stats (or your own opinion mixed in) or you just lose haha.

So I can agree with what you're saying once the word "much" in your quote is quantified so that you're asking for something "greater than the predictive power afforded by stats". Either way, history can be used to predict the future with better than chance probability, even in the NFL, so it's not "100% garbage".

 

You get points for using the Dude. lol

 

What you're doing is one of the best examples of cherry picking I've seen in awhile.

You can't just use 2 teams when the stats use 32. You can't just use one game when the stats are using 12+.

The argument was, "You can't use past playoffs to predict future playoff outcomes."

Cbrad showed that over the course of NFL history the #1 seeds win more. Now, have there been years when the #1 team doesn't get past the first round? Of course. However, as he and I have been saying, over a given course of time, the #1 seed has a better overall chance of making it to the Super Bowl.

 

I agree that they aren't garbage, and with some, if you stick to your guns, yeah, that'll pan out over time.

I know guys that love to bet on the home dog, and according to them, that pays off in the long run, but it's also picking your spots, based on personal knowledge, you shouldn't take every home dog(although I know a guy that does, and he claims it pays off), but I think stats back that up, that home dogs cover the spread at a higher rate, so as you point out, they can help in spots.

On the other hand, personal knowledge can be a trump card vs stats, take the 05 Steelers, they were a WC team, but the truth is, they were one of the best teams in the league, they had a 3 game stretch where Ben was injured, so 3 losses, but a 4 game win streak before, and another 4 game win streak to end the season after, so with Ben, they were dangerous, with Maddox, not so much.

So statistically treating them like another WC team would be a mistake, they were elite.

This years Steelers are eerily similar, 11-1 with a healthy Ben, 0-4 with a less than healthy Ben, but that 05 team was better overall, especially on D, but not as good on O as this team.

So yes, stats can help you make a more informed opinion, especially in general, but predicting a win vs Pitt based on past results isn't a very good idea, like many people are saying, "we beat them earlier in the year", but that has no bearing on this game, both teams are different now.

 

Yeah, whenever one tries to evaluate how good population stats are in predicting "individual games" it's important that one not just look at those cases where the prediction failed, but also include those cases where the prediction succeeded.

So measuring how well a given method predicts "individual games" requires looking at a sample of games, preferably with large sample size.

 

Never disagreed with any of this Finster. Just keep in mind the actual predictive power of stats is better than HFA only haha! Oh.. and the predictive power of humans + stats is better, but even that combo isn't that good. Really, no one does it well.