sleepy
Platinum Hoya (over 5000 posts)
Posts: 6,079
|
Post by sleepy on Feb 7, 2019 12:02:54 GMT -5
Didn't we have the discussion already, Pomeroy doesn't take recency into account from reports. Is there a reason to think it does? Either way, I don't totally by that margin of victory is a good way to reflect relative strength. There are a lot of factors, especially early in the season, that influence that final margin of victory that don't adequately reflect the difference of two teams. It shouldn't be totally disregarded, but as it seems like there is too much emphasis on that going into these analytic models to give an accurate picture. At the end of the day, the point of the game is to score one more point than your opponent, no matter how you go about doing it. I also think these models used to severely overrate III teams because of how they value certain metrics over others. For example: 2017 (III's last year) 69 (14-18) 2016 62 (15-18) 2015 22 (22-11) 2014 57 (18-15) Our ranking was severely inflated in those years because III's teams played well to how these systems run numbers. They don't place enough value on the actual ability to win games. You are right. KenPom does not take recency into account. At the beginning of the season, he builds in predictive effects based on the previous year, but by conference season, that is no longer a factor. But I think you keyed in on the item in bold above. Models like KenPom are based on offensive and defensive efficiency, and that's how they determine who they predict will score at least one more point than the other team. Clearly, the models aren't perfect (and I don't think Ken Pomeroy or the author of any other model would argue with that), fluid, and subject to a variety of factors. The value of these models is analyzing a large set of data over time. For example, you point to the JT3 teams, and say you think KenPom overrated them. Maybe. But, what KenPom does provide is a basis for comparing the same sets of data over a long period of time. For example, JT3's poor teams of 2014, 2015, and 2017 were ranked better than last year's team. Those teams were more efficient. How that translates to wins and losses is another matter, of course, and depends on things like strength of schedule, luck, etc. I also agree with your statement that these models are best used at the end of the season with full amounts of data. Keep in mind predictive models are just that. If Georgetown has a 55% chance to win against Butler, that doesn't mean we "should" win. It just means that the model predicts that if the game was played 100 times, Georgetown would win 55 times, Butler 45. But, the games are just played once. That's why, when results happen that contradict the models, it doesn't necessarily mean the models are wrong. It's just a reflection that the models are just that - estimates and predictions - and not a prediction of one specific outcome. If you think a 15-18 team isn't severely overrated at 62 in the country... IDK what to tell you. The issue is that people take these models and rankings as who the best teams are, but as you say, they are ranking better teams. They are ranking the efficiency of teams, not quality. Sometime they are one in the same, and efficiency will certainly make you better. But better team is sometimes the less efficient team, because they know how to win basketball games and the other team can't. I'm not talking just intangibles but there are data points these systems don't or can't take into account that tell you could tell the quality of a basketball team more than just how efficient they are. The types of turnovers, what points in games you play more/less efficient, etc etc. But that isn't want KenPoms are designed to do. And that is ok, but people need to recognized what the models are actually measuring and use it as a datapoint to determine how good a team is, not as an actual ranking.
|
|
jwp91
Platinum Hoya (over 5000 posts)
Posts: 6,013
|
Post by jwp91 on Feb 7, 2019 12:12:28 GMT -5
Just so we are on the same page, this is a Butler team that has won fewer games than Georgetown, lost more games than Georgetown, lost head to head vs. Georgetown at home and sits at 4-6 in the conference. If your model shows Butler as 24 places higher than Georgetown, I start losing trust in your model. Wins and losses do not mean anything directly to the KenPom rankings, but rather, they are based on the underlying efficiency and stats. There is a site (Massey Ratings) that, each week, aggregates a variety of different rankings for all teams. Without listing the specific models for ease, here are Butler's ratings among the first 20 models listed (there are more), as of Sunday, February 3, 2019 (so doesn't include results since then): 57, 58, 58, 55, 67, 62, 56, 51, 55, 65, 56, 62, 57, 60, 55, 54, 59, 66, 44, 49. And here are Georgetown's first 20: 71, 65, 68, 77, 84, 69, 72, 68, 82, 67, 87, 86, 61, 67, 79, 78, 61, 79, 71, 93. So of these 20 models - all of which rank the teams in different ways and using different methods - Butler is ranked higher than Georgetown in every single one. (By the way, among those above are NET, KenPom, and Sagarin.) My point is that if 20 different models are ranking Butler ahead of Georgetown there has to be something to it. There are multiple factors explaining this: - As others said, Georgetown's underwhelming wins against some very bad teams hurt it in the rankings (which is different from strength of schedule). - Butler has some decent wins - using KenPom rankings, they have beat Florida (37), Mississippi (46), St. John's (47), Creighton (49). - Georgetown's best wins are: St. John's (47), Butler (55), Liberty (61), Illinois (71). - Butler has two sub-100 losses (St. Louis (118) and Xavier (102)), to Georgetown's three (Loyola Marymount (139), SMU (114), and Xavier (102). I admit that based purely on wins and losses, the teams probably look more similar to one another than in rankings. That said, after last night, Georgetown is ranked 79 on KenPom, and Butler 55. If Georgetown convincingly beats Butler, we very well could have a situation where they're much closer ranked after the game. I think people put too much stock in the ranking number, too. If you look at the underlying ranking stats, it is more telling, and shows how tightly packed the teams in the 25-75 range are. For example, these are the Adjusted Efficiencies of the teams that are ranked #1, #25, #50: #1: 36.07 #25: 18.42 #50: 12.72 #75: 9.35 In other words, in terms of performance, there is a way bigger difference between #1 and #25, than there is between #25 and #50. So Butler being ahead of Georgetown on KenPom by 24 spots really isn't as drastic a difference as it seems. I’ll give you Butler and a point. How much do you want to bet? Your models tell you that you should take my offer...
|
|
prhoya
Blue & Gray (over 10,000 posts)
Posts: 23,304
|
Post by prhoya on Feb 7, 2019 12:14:32 GMT -5
Donny Marshall was awful and unprepared last night, but I’ll give him credit for noticing Pickett’s set shot routine when wide open: brings the ball up from the hip and high arc for the swish. He’s rinsing and repeating at a high percentage. Beautiful to see! Kudos to Mom Wallace?!
|
|
|
Post by aleutianhoya on Feb 7, 2019 12:32:40 GMT -5
|
|
|
Post by johnnysnowplow on Feb 7, 2019 12:37:06 GMT -5
Is this game on a national channel? Only thing I see listed is MASN2.
|
|
LCPolo18
Diamond Hoya (over 2500 posts)
Posts: 4,406
|
Post by LCPolo18 on Feb 7, 2019 12:43:18 GMT -5
Is this game on a national channel? Only thing I see listed is MASN2. Fox Sports Network (FSN) regional coverage MASN2 YES AT&T SportsNet Pittsburgh AT&T SportsNet Rocky Mountain Cox YurView - New England FOX Sports Arizona FOX Sports Detroit PLUS FOX Sports Florida FOX Sports Midwest PLUS FOX Sports North PLUS FOX Sports Ohio FOX Sports Southeast FOX Sports Southwest PLUS ROOT Sports
|
|
|
Post by johnnysnowplow on Feb 7, 2019 12:44:33 GMT -5
Is this game on a national channel? Only thing I see listed is MASN2. Fox Sports Network (FSN) regional coverage MASN2 YES AT&T SportsNet Pittsburgh AT&T SportsNet Rocky Mountain Cox YurView - New England FOX Sports Arizona FOX Sports Detroit PLUS FOX Sports Florida FOX Sports Midwest PLUS FOX Sports North PLUS FOX Sports Ohio FOX Sports Southeast FOX Sports Southwest PLUS ROOT Sports Ah thank you
|
|
hoyainla
Diamond Hoya (over 2500 posts)
Suspended
Posts: 4,719
|
Post by hoyainla on Feb 7, 2019 12:44:43 GMT -5
My bigger problem with the model is that I think it puts too much weight on OOC as once you get into conference the strength of your wins and losses are limited to how good your conference is. It’s almost a self-fulfilling prophesy. Last year when the BE had a strong OOC season, all the teams came into conference with great RPIs and therefore all future wins were “good” wins while losses remained “good” losses. Most teams stayed in the Top 40 as a result. This year the conference was not nearly as strong early on and the rankings starting January showed that. As a result it’s really hard to move up when the teams you are playing are at a certain level and then everyone just beats up on each other, with rankings converging to the mean. Look at the Big 12 this year, last I looked 8/10 teams were considered “in” because the relative rankings were strong and therefore every game virtually becomes a quality game. I get the OOC sets the foundation for relative strength, but it seems as a result once a conference is deemed good in January it stays that way because there’s no more comparison point for the teams in the conference to get recalibrated. What is your fix to this? Split the conference schedule in half with games early then play OOC and than the rest late? Just because a team shows well against other teams that didn't show well OOC what do you want the model to do? The BE dug it's own grave this year. Also this is why the BE encouraged teams to play tougher OOC schedules because it helps everyone in the conference for the reason you mentioned. I don't think the conference can mandate who a school plays OOC but it should do it's best to. As for a barometer as to how good the BE is this year, the current third place team that just beat the 2nd place team on the road. That same team lost to Duke by 30 the game before. Yes Duke is very good but that doesn't exactly help the BE case this year.
|
|
|
Post by centercourt400s on Feb 7, 2019 12:45:47 GMT -5
Donny Marshall was awful and unprepared last night, but I’ll give him credit for Pickett’s set shot routine when wide open: brings the ball up from the hip and high arc for the swish. He’s rinsing and repeating at a high percentage. Beautiful to see! Kudos to Mom Wallace?! I liked Pickett's shooting last night as well but I believe Marshall actually described it as coming up from his shin
|
|
NCHoya
Diamond Hoya (over 2500 posts)
Posts: 2,924
|
Post by NCHoya on Feb 7, 2019 14:06:52 GMT -5
Well, I hope the Hoyas just keep winning and the numbers will take care of themselves. The problems being discussed have been debated for decades. November results need to matter because mid-majors do not get any more chances after December to play power teams. And yes, that means when a conference is initially rated strong, it simply continues to build on itself during conference play. Look at Indiana, that team lost 7 straight games, but never fell out of the top 50 in Kenpom or NET, it is very hard to fall once you achieve a top 20 rating.
The only practical solution is to get BE teams to play tough OOCs, and we certainly are not doing our share in that regard.
As for this game, I expect to win against Butler at home. Butler is 1-5 on the road, I surely do not see it as a pick'em or a 1 pt spread. Yes, the Hoyas should feel some pressure to win this one and expect to win. I hope the gray out promotion draws a larger crowd than normal.
|
|
|
Post by hoyasaxa2003 on Feb 7, 2019 14:26:45 GMT -5
I’ll give you Butler and a point. How much do you want to bet? Your models tell you that you should take my offer... Again, I think you're taking this the wrong way. Nobody said that KenPom will predict the actual result. It's intended to predict the most likely result. In the case of a nearly 50/50 game, the 1 point difference is essentially meaningless because it's so close and so marginal. A Georgetown 1 point win is nearly as likely as a Butler 1 point win. The actual result could go in either direction or be a blowout for either team. Who knows...That's why the games are played, and that is why "upsets" happen all the time.
|
|
hoyainla
Diamond Hoya (over 2500 posts)
Suspended
Posts: 4,719
|
Post by hoyainla on Feb 7, 2019 14:30:13 GMT -5
I’ll give you Butler and a point. How much do you want to bet? Your models tell you that you should take my offer... Ill think you will be able to bet as much as you want on something close to that line before the game. Most books release close to the KenPom line.
|
|
jwp91
Platinum Hoya (over 5000 posts)
Posts: 6,013
|
Post by jwp91 on Feb 7, 2019 14:46:42 GMT -5
I’ll give you Butler and a point. How much do you want to bet? Your models tell you that you should take my offer... Again, I think you're taking this the wrong way. Nobody said that KenPom will predict the actual result. It's intended to predict the most likely result. In the case of a nearly 50/50 game, the 1 point difference is essentially meaningless because it's so close and so marginal. A Georgetown 1 point win is nearly as likely as a Butler 1 point win. The actual result could go in either direction or be a blowout for either team. Who knows...That's why the games are played, and that is why "upsets" happen all the time. And if the models are not predictive, they are a bit of academic mastur*ation. My assessment is that in the real world this is not a 50/50 game. Based on the KenPom construct of the difference in offensive and defensive efficiencies and the marginal difference between the two teams, it is.....but if that doesn’t mirror the likely outcome in the real world, then KenPom has failed.
|
|
|
Post by hoyasaxa2003 on Feb 7, 2019 15:08:19 GMT -5
And if the models are not predictive, they are a bit of academic mastur*ation. My assessment is that in the real world this is not a 50/50 game. Based on the KenPom construct of the difference in offensive and defensive efficiencies and the marginal difference between the two teams, it is.....but if that doesn’t mirror the likely outcome in the real world, then KenPom has failed. To some extent, everything used to compare teams is academic. That's part of the point. You can't evaluate using wins and losses since there are 353 teams, and each team only plays about 30 games a year. So that's really what RPI, NET, KenPom, Sagarin, or whatever models do. Even your subjective "jwp91 model" (whatever criteria you're using for saying it's not a 50/50 game) is academic to the extent it's somewhat arbitrary based on your opinion (and if you're able to sonsistently handicap game better than KenPom/Las Vegas, you should probably get to a state where you can start betting). People like Ken Pomeroy take their models seriously and put a lot of work into them. I do think it would be inaccurate to say KenPom, as a whole "doesn't mirror the likely outcome in the real world" in the aggregate. In any single game? That can certainly be the case, especially when the model is in flux, especially early in the year. But as a whole, I think it's proven fairly accurate.* But, we can agree to disagree. I hope you are right, we are the better team, and we run away with it. I want to win the game!
|
|
|
Post by RockawayHoya on Feb 7, 2019 16:33:16 GMT -5
According to the HHC e-mail I just received, those planning to use Metro to get to the game on Saturday should know the following:
"Due to scheduled maintenance, the Metro's red line will not be operational at stops near Capital One Arena. The red line will be closed at Farragut North, Metro Center, and Gallery Place. Instead, shuttle buses will operate between Dupont Circle and Judiciary Square. However, the blue/orange/silver lines will be operational at Metro Center and the green/yellow lines will be operational at Gallery Place. For more information, please visit WMATA's website."
|
|
mdtd
Golden Hoya (over 1000 posts)
Posts: 1,567
|
Post by mdtd on Feb 7, 2019 16:34:04 GMT -5
KenPom has us 79 this AM NCAA NET rankings also 79 - down from 82. We are going in the right direction. NET is 73 now, was 79 before the update today.
|
|
guru
Platinum Hoya (over 5000 posts)
Posts: 6,600
|
Post by guru on Feb 7, 2019 19:10:01 GMT -5
Wins and losses do not mean anything directly to the KenPom rankings, but rather, they are based on the underlying efficiency and stats. There is a site (Massey Ratings) that, each week, aggregates a variety of different rankings for all teams. Without listing the specific models for ease, here are Butler's ratings among the first 20 models listed (there are more), as of Sunday, February 3, 2019 (so doesn't include results since then): 57, 58, 58, 55, 67, 62, 56, 51, 55, 65, 56, 62, 57, 60, 55, 54, 59, 66, 44, 49. And here are Georgetown's first 20: 71, 65, 68, 77, 84, 69, 72, 68, 82, 67, 87, 86, 61, 67, 79, 78, 61, 79, 71, 93. So of these 20 models - all of which rank the teams in different ways and using different methods - Butler is ranked higher than Georgetown in every single one. (By the way, among those above are NET, KenPom, and Sagarin.) My point is that if 20 different models are ranking Butler ahead of Georgetown there has to be something to it. There are multiple factors explaining this: - As others said, Georgetown's underwhelming wins against some very bad teams hurt it in the rankings (which is different from strength of schedule). - Butler has some decent wins - using KenPom rankings, they have beat Florida (37), Mississippi (46), St. John's (47), Creighton (49). - Georgetown's best wins are: St. John's (47), Butler (55), Liberty (61), Illinois (71). - Butler has two sub-100 losses (St. Louis (118) and Xavier (102)), to Georgetown's three (Loyola Marymount (139), SMU (114), and Xavier (102). I admit that based purely on wins and losses, the teams probably look more similar to one another than in rankings. That said, after last night, Georgetown is ranked 79 on KenPom, and Butler 55. If Georgetown convincingly beats Butler, we very well could have a situation where they're much closer ranked after the game. I think people put too much stock in the ranking number, too. If you look at the underlying ranking stats, it is more telling, and shows how tightly packed the teams in the 25-75 range are. For example, these are the Adjusted Efficiencies of the teams that are ranked #1, #25, #50: #1: 36.07 #25: 18.42 #50: 12.72 #75: 9.35 In other words, in terms of performance, there is a way bigger difference between #1 and #25, than there is between #25 and #50. So Butler being ahead of Georgetown on KenPom by 24 spots really isn't as drastic a difference as it seems. That is a a major flaw in a system when the main goal of a single ame is to win. Not win by 5/10/15/20. One point. If you don't take that into account, your analysis of teams is always going to be flawed. When we beat Butler on Saturday, we will still be behind them on KenPom. There are very very few reasons a team who beats another team twice in one season should be behind that team in any ranking system. KenPom’s consistent blindness to the latter year JT3 Teams obvious deficiencies forever ruined my faith in that rating system. JT3 teams were almost perfectly constructed to game the kenpom system. Those 3 terrible JT3 teams in his final 4 years were seen as “bubble” teams until the season was essentially over when anyone with eyes knew just how bad they were. It’s also why using it to call this year’s team “clearly the worst Hoyas team since 2004” is also ridiculous, as I think we are starting to see.
|
|
sleepy
Platinum Hoya (over 5000 posts)
Posts: 6,079
|
Post by sleepy on Feb 7, 2019 19:21:28 GMT -5
That is a a major flaw in a system when the main goal of a single ame is to win. Not win by 5/10/15/20. One point. If you don't take that into account, your analysis of teams is always going to be flawed. When we beat Butler on Saturday, we will still be behind them on KenPom. There are very very few reasons a team who beats another team twice in one season should be behind that team in any ranking system. KenPom’s consistent blindness to the latter year JT3 Teams obvious deficiencies forever ruined my faith in that rating system. JT3 teams were almost perfectly constructed to game the kenpom system. Those 3 terrible JT3 teams in his final 4 years were seen as “bubble” teams until the season was essentially over when anyone with eyes knew just how bad they were. It’s also why using it to call this year’s team “clearly the worst Hoyas team since 2004” is also ridiculous, as I think we are starting to see. Did someone say it is clearly the worst hoyas team since 2004?!
|
|
guru
Platinum Hoya (over 5000 posts)
Posts: 6,600
|
Post by guru on Feb 7, 2019 19:45:19 GMT -5
KenPom’s consistent blindness to the latter year JT3 Teams obvious deficiencies forever ruined my faith in that rating system. JT3 teams were almost perfectly constructed to game the kenpom system. Those 3 terrible JT3 teams in his final 4 years were seen as “bubble” teams until the season was essentially over when anyone with eyes knew just how bad they were. It’s also why using it to call this year’s team “clearly the worst Hoyas team since 2004” is also ridiculous, as I think we are starting to see. Did someone say it is clearly the worst hoyas team since 2004?! Yes
|
|
EasyEd
Platinum Hoya (over 5000 posts)
Posts: 7,272
|
Post by EasyEd on Feb 7, 2019 20:46:08 GMT -5
Models, schmodels. Let's talk Hoya/Butler basketball.
|
|