SinceThe70s Posted November 13, 2019 Share Posted November 13, 2019 6 hours ago, The Frankish Reich said: row33, I know math is hard... ... that’s ok. I’ll explain: teams with W-L records better than what their point differentials suggest tend to fall back, and vice versa. The Bills W-L record is better than what their meager positive point differential would suggest. Therefore, we may infer that they won’t keep winning games at a .670 rate. Got it? IMO math is easy, but stats, probability and trend analysis isn't. A .670 winning percentage implies an 11-5 season. I think most fans could agree that 11-5 is unlikely without relying on a point differential analysis based on a ridiculously small sample set. Stats are only useful to predict trends over large sample sizes IMO. Put another way, if the Bills blow out the Dolphins 45-0 this Sunday it would benefit the point differential but wouldn't lead me to believe they'll be an 11 win team. Link to comment Share on other sites More sharing options...
Hapless Bills Fan Posted November 13, 2019 Author Share Posted November 13, 2019 1 hour ago, NoSaint said: Is there a reason you are blending a per game with an overall net? Wouldn’t win percentage Jake more sense than total wins? It's not blending, it's looking at a correlation. It would probably make more sense to look at win percentage, yes, to account for the teams that haven't had byes yet. 1 hour ago, NoSaint said: that’s not a small amount though. Go back through many of our 6-10 type seasons and toss a touchdown on it score every other week and that record would shoot up often times. That could be true. Last year an extra TD would have made a difference in 3 games, so overall difference between 6-10 vs 9-7. In many others we either blew the team out or got blown. I think that's the reason why some pundits feel a team changing its record from, say, 7-9 to 9-7 doesn't necessarily indicate much real difference in the quality of a team - could be a couple lucky breaks. Link to comment Share on other sites More sharing options...
Hapless Bills Fan Posted November 13, 2019 Author Share Posted November 13, 2019 31 minutes ago, SinceThe70s said: IMO math is easy, but stats, probability and trend analysis isn't. A .670 winning percentage implies an 11-5 season. I think most fans could agree that 11-5 is unlikely without relying on a point differential analysis based on a ridiculously small sample set. Stats are only useful to predict trends over large sample sizes IMO. Put another way, if the Bills blow out the Dolphins 45-0 this Sunday it would benefit the point differential but wouldn't lead me to believe they'll be an 11 win team. That's actually implicit in looking at the spread of the data. It's clearly not predictive even for events that have already taken place - if you tell me the point differential, I can't tell you how many wins that team has. I thought it was interesting context to what's happening around the league and how small the point differentials really are for most teams halfway through the season Link to comment Share on other sites More sharing options...
NoSaint Posted November 13, 2019 Share Posted November 13, 2019 30 minutes ago, Hapless Bills Fan said: It's not blending, it's looking at a correlation. It would probably make more sense to look at win percentage, yes, to account for the teams that haven't had byes yet. That could be true. Last year an extra TD would have made a difference in 3 games, so overall difference between 6-10 vs 9-7. In many others we either blew the team out or got blown. I think that's the reason why some pundits feel a team changing its record from, say, 7-9 to 9-7 doesn't necessarily indicate much real difference in the quality of a team - could be a couple lucky breaks. I suppose shoot up is a bit overstated in my wording... but even an extra win or two is big. and agree a little health, good bounce or lucky call go a long way in sorting out the crowded middle of the pack. but yea, it’s definitely blending to compare total wins with average differential. Doing the average differential with the average outcome (win percentage) should be a better metric. You wouldn’t do a graph of average yards per game to total touchdowns, knowing there are different game counts, for instance. Not a big deal but it’s not the right build out Link to comment Share on other sites More sharing options...
SinceThe70s Posted November 13, 2019 Share Posted November 13, 2019 24 minutes ago, Hapless Bills Fan said: That's actually implicit in looking at the spread of the data. It's clearly not predictive even for events that have already taken place - if you tell me the point differential, I can't tell you how many wins that team has. I thought it was interesting context to what's happening around the league and how small the point differentials really are for most teams halfway through the season It's especially interesting given my perception that there are more really bad teams and less really good teams this year than usual. Could be just water not finding it's level just yet. Link to comment Share on other sites More sharing options...
Reed83HOF Posted November 13, 2019 Share Posted November 13, 2019 9 hours ago, Hapless Bills Fan said: But I could be wrong. Been wrong about lots of stuff here over the last 10 years. We all have good sir Link to comment Share on other sites More sharing options...
The Frankish Reich Posted November 13, 2019 Share Posted November 13, 2019 10 hours ago, SinceThe70s said: IMO math is easy, but stats, probability and trend analysis isn't. A .670 winning percentage implies an 11-5 season. I think most fans could agree that 11-5 is unlikely without relying on a point differential analysis based on a ridiculously small sample set. Stats are only useful to predict trends over large sample sizes IMO. Put another way, if the Bills blow out the Dolphins 45-0 this Sunday it would benefit the point differential but wouldn't lead me to believe they'll be an 11 win team. Correct. Football is difficult because there's no such thing as an adequate sample size with a 16 game season. That's why MLB has provided the laboratory for sports metrics; the season is literally 10X longer. But still ... point differential and various other things (turnover differential, W-L records in games decided by 3 points or less, etc.) are indicative of teams experiencing better or worse than typical luck as opposed to being where they are in the standings by virtue of true talent level. So these things are limited in value, but not worthless. And yes, if the Bills actually do blow out an inferior team and raise that point differential, that tells us something useful. 1 Link to comment Share on other sites More sharing options...
Recommended Posts