jwp91
Platinum Hoya (over 5000 posts)
Posts: 5,892
Member is Online
|
Post by jwp91 on Nov 13, 2019 8:40:51 GMT -5
At what point does PPP take into account who is on the floor for the other team? To the extent that a B line player has better production against B line guys from the other team, there is reduced value as a predictor of what that player would do against the opponents starters, amiright? I think that is a reasonable criticism of PPP, but we already have small sample sizes. Utilizing the approach mentioned would make the sample size problem even greater.
|
|
|
Post by tribeninerhoya on Nov 13, 2019 9:18:42 GMT -5
I like to use barttorvik a lot. I think it's a lot of great (free) information, and has some fun projection tools for you to play with (teamcast, etc.). Thought I'd share: barttorvik.com/team.php?team=Georgetown
|
|
hoyainla
Diamond Hoya (over 2500 posts)
Suspended
Posts: 4,719
|
Post by hoyainla on Nov 13, 2019 9:19:55 GMT -5
It’s definitely not the only stat to look at but it is widely considered one of the better ones as it shows the effects of things that can’t be easily measured. It’s definitely better than looking at the traditional box score stats. But too often people try to use it incorrectly. Were Blair and Malinowski the best players on the team last year? PPP differential says that they were. It’s an interesting stat, but like we have both said, it’s only one of many stats someone should look at when evaluating a player. No they weren’t but I think it’s fair to compare guys who play the same position when compare PPP. If Player A same his name is James plays 90% of possessions then his PPP diff is likely to be pretty similar to the team. That is where seeing how the team does with them off the court comes into play. If we see guys that are outliers in PPP diff we can the get a better picture of what is going on. For example if player A and B play together and have a PPP diff of 0 but Player A without B has a diff of +.1 and Player B without A has a diff of -.1 then we can see where the problem lies. Or if Player A and C have a diff of +.1 then A and C should play together more than A and B. Single player PPP over the course of a season where there are enough possessions to make judgement gives us a good idea. I don’t think we are there after 2 games but I also don’t think it takes an entire of year of games to get that data. Last year Leblanc had a better PPP diff than Mourning who he eventually replaced. Mosley had a better PPP diff than Pickett who he eventually replaced. I don’t think anyone here would argue those moves didn’t make the team better.
|
|
|
Post by tribeninerhoya on Nov 13, 2019 9:27:56 GMT -5
But too often people try to use it incorrectly. Were Blair and Malinowski the best players on the team last year? PPP differential says that they were. It’s an interesting stat, but like we have both said, it’s only one of many stats someone should look at when evaluating a player. No they weren’t but I think it’s fair to compare guys who play the same position when compare PPP. If Player A same his name is James plays 90% of possessions then his PPP diff is likely to be pretty similar to the team. That is where seeing how the team does with them off the court comes into play. If we see guys that are outliers in PPP diff we can the get a better picture of what is going on. For example if player A and B play together and have a PPP diff of 0 but Player A without B has a diff of +.1 and Player B without A has a diff of -.1 then we can see where the problem lies. Or if Player A and C have a diff of +.1 then A and C should play together more than A and B. Single player PPP over the course of a season where there are enough possessions to make judgement gives us a good idea. I don’t think we are there after 2 games but I also don’t think it takes an entire of year of games to get that data. Last year Leblanc had a better PPP diff than Mourning who he eventually replaced. Mosley had a better PPP diff than Pickett who he eventually replaced. I don’t think anyone here would argue those moves didn’t make the team better. See barttorvik’s PORPAGATU! (PRPG!) ratings, which I think are interesting and take into account usage, etc., for player value measurement. It’s an interesting approach. FYI- PORPAGATU! is Point Over Replacement Per Adjusted Game At That Usage
|
|
hoyainla
Diamond Hoya (over 2500 posts)
Suspended
Posts: 4,719
|
Post by hoyainla on Nov 13, 2019 9:28:38 GMT -5
I like to use barttorvik a lot. I think it's a lot of great (free) information, and has some fun projection tools for you to play with (teamcast, etc.). Thought I'd share: barttorvik.com/team.php?team=GeorgetownBarttovik is good but the problem assuming you are using his PRPGI which is his main stat is that it calculates minutes and possessions so how much a coach plays makes them look better. He also uses a baseline for ORTG that is not dependent of the team they are playing on. As long as you are higher than that baseline and your coach plays you a lot you will have a better PRPGI than other players on your team that may in fact be better. PPP is independent of total playing time.
|
|
hoyainla
Diamond Hoya (over 2500 posts)
Suspended
Posts: 4,719
|
Post by hoyainla on Nov 13, 2019 9:29:25 GMT -5
No they weren’t but I think it’s fair to compare guys who play the same position when compare PPP. If Player A same his name is James plays 90% of possessions then his PPP diff is likely to be pretty similar to the team. That is where seeing how the team does with them off the court comes into play. If we see guys that are outliers in PPP diff we can the get a better picture of what is going on. For example if player A and B play together and have a PPP diff of 0 but Player A without B has a diff of +.1 and Player B without A has a diff of -.1 then we can see where the problem lies. Or if Player A and C have a diff of +.1 then A and C should play together more than A and B. Single player PPP over the course of a season where there are enough possessions to make judgement gives us a good idea. I don’t think we are there after 2 games but I also don’t think it takes an entire of year of games to get that data. Last year Leblanc had a better PPP diff than Mourning who he eventually replaced. Mosley had a better PPP diff than Pickett who he eventually replaced. I don’t think anyone here would argue those moves didn’t make the team better. See barttorvik’s PORPAGATU! (PRPG!) ratings, which I think are interesting and take into account usage, etc., for player value measurement. It’s an interesting approach. FYI- PORPAGATU! is Point Over Replacement Per Adjusted Game At That Usage I was typing my answer to you as you were to me and I explained why I think it’s flawed.
|
|
rhw485
Silver Hoya (over 500 posts)
Posts: 740
|
Post by rhw485 on Nov 13, 2019 9:49:06 GMT -5
Ignoring the small sample size issues, as that will sort themselves out over time, there are a few issues that will never be solved by the analytics piece alone.
1. the PPP issue discussed. Yes, when comparing Blair to McClung last year as an example, the metric will never be able to take into account that McClung was going against starters and Blair usually was not. Teams stagger their rotations differently, Georgetown changes its rotation over the season as well, it will never be a true apples to apples comparison
2. Efficiency vs. Usage. Just because a player is highly efficient, it doesn't mean they're our best offensive player if their usage is very low. Josh is a perfect example of this. There were some arguments last year that Josh was our best offensive player last year based on his ORating was the highest. But his usage rate was 16%, the 18th percentile in college hoops. We can't simply have Josh shoot more, his baskets were the result of dump offs or offensive rebounds. I love Josh as a player and he's critical to our team. It's just an example where the metrics are missing a critical piece of how the basket was really created.
That being said, the only way to actually prove whether the metric is insightful or not is to actually experiment w different lineups to try and prove out. Does Blair PPP change if given more minutes w starters? Does a player maintain efficiency levels when given more opportunities? Does it match the eye test of what we're seeing on court? The answer isn't to ignore them, it's to try and see whether there's something real that can help the team w different lineup construction / minutes distribution. Ewing going deep in these first two games is great, but he didn't really try different lineups. He had primarily starter lineups and bench lineups and I think that was a missed opportunity to see how the pieces could fit together.
|
|
seaweed
Diamond Hoya (over 2500 posts)
Posts: 4,629
Member is Online
|
Post by seaweed on Nov 13, 2019 11:27:46 GMT -5
I like to use barttorvik a lot. I think it's a lot of great (free) information, and has some fun projection tools for you to play with (teamcast, etc.). Thought I'd share: barttorvik.com/team.php?team=GeorgetownIf I am reading the chart at the bottom left correctly, opponents are shooting threes on 44.9% of possessions and assisting on 56.6% of their scores. The first number is an interesting question of styles and may help sometimes and hurt others. The second number tells me we are NOT guarding passing lanes AT ALL. Unless those #s mean something else - I am not that versed in advance stats.
|
|
hoyasaxa2003
Platinum Hoya (over 5000 posts)
Posts: 9,685
Member is Online
|
Post by hoyasaxa2003 on Nov 13, 2019 12:51:18 GMT -5
Ignoring the small sample size issues, as that will sort themselves out over time, there are a few issues that will never be solved by the analytics piece alone. 1. the PPP issue discussed. Yes, when comparing Blair to McClung last year as an example, the metric will never be able to take into account that McClung was going against starters and Blair usually was not. Teams stagger their rotations differently, Georgetown changes its rotation over the season as well, it will never be a true apples to apples comparison 2. Efficiency vs. Usage. Just because a player is highly efficient, it doesn't mean they're our best offensive player if their usage is very low. Josh is a perfect example of this. There were some arguments last year that Josh was our best offensive player last year based on his ORating was the highest. But his usage rate was 16%, the 18th percentile in college hoops. We can't simply have Josh shoot more, his baskets were the result of dump offs or offensive rebounds. I love Josh as a player and he's critical to our team. It's just an example where the metrics are missing a critical piece of how the basket was really created. That being said, the only way to actually prove whether the metric is insightful or not is to actually experiment w different lineups to try and prove out. Does Blair PPP change if given more minutes w starters? Does a player maintain efficiency levels when given more opportunities? Does it match the eye test of what we're seeing on court? The answer isn't to ignore them, it's to try and see whether there's something real that can help the team w different lineup construction / minutes distribution. Ewing going deep in these first two games is great, but he didn't really try different lineups. He had primarily starter lineups and bench lineups and I think that was a missed opportunity to see how the pieces could fit together. A few things: 1. While PPP does not account for your opponents, I think the argument above is overstated. Sure, McClung started. But he also played with the bench at times, and Blair often played with a bunch of the starters. Basketball isn't hockey, most teams play relatively short rotations, so while there's some truth to the statement, I think it's really a minor factor overall, unless, for example, a guy like a true walk-on only plays during garbage time. 2. Again, your point on usage is not necessarily correct. Sure, LeBlanc's strength was that he cleaned up bad shots, took high percentage shots, etc. I disagree that you can't just have LeBlanc shoot more. You absolutely can, by running plays for him, and having him be a bigger part of the offense. Will he be as good? I don't know because he's never been used in that capacity, but if you look at the guy (again, stepping away from analytics for a moment), it's clear the guy is one of those players who gets it. He's not going to suddenly start taking awful shots (like, say, Trey Mourning did) because they allow him more opportunities. (Good example: In 2012-2013, Otto Porter's usage went way up AND so did his efficiency. Increasing usage does not always mean that efficiency goes down.) 3. I know McClung is popular (for good reason) and Blair less so, but I don't think the gulf in talent last year between McClung and Blair was as big as people think. Yes, they are very different players, and McClung is certainly more flashy, but there's at least an argument that Blair was the better player (not saying I think so, but it's a legitimate position). I am not saying I agree with it, but Blair had higher efficiency, shot threes better, turned it over less, and rebounded better. Personally, I think McClung is more valuable because he's a more versatile player and can do more, plus I think his ceiling is way higher. I would also note that PPP differential takes into account defense, which I think most fans either largely ignore or are poor at examining. I think most of HoyaTalk would agree that McClung was a pretty bad defender last season. Blair certainly wasn't amazing, but that's also part of the reason why the PPP differential favored Blair in some ways. While it was en vogue to rip Govan for everything relating to defense, guys like McClung were significant problems on that end, and that's where something like PPP can expose a player where it might not be immediately evident just by watching.
|
|
LCPolo18
Diamond Hoya (over 2500 posts)
Posts: 4,406
|
Post by LCPolo18 on Nov 13, 2019 13:48:04 GMT -5
I would also note that PPP differential takes into account defense, which I think most fans either largely ignore or are poor at examining. I think most of HoyaTalk would agree that McClung was a pretty bad defender last season. Blair certainly wasn't amazing, but that's also part of the reason why the PPP differential favored Blair in some ways. While it was en vogue to rip Govan for everything relating to defense, guys like McClung were significant problems on that end, and that's where something like PPP can expose a player where it might not be immediately evident just by watching. Yes PPP differential takes into account defense. But if someone is going to look at PPP differential, they should also look at the PPP offense and PPP defense numbers. Blair's defense in the Big East was incredibly bad by this measure, even compared to Mac who was generally in line with the rest of the team. Blair All Games Differential: 0.08 All Games Offense: 1.1 All Games Defense: 1.02 Big East Differential: 0.0 Big East Offense: 1.11 Big East Defense: 1.11 McClung All Games Differential: 0.0 All Games Offense: 1.01 All Games Defense: 1.01 Big East Differential: -0.03 Big East Offense: 0.99 Big East Defense: 1.02
|
|
hoyasaxa2003
Platinum Hoya (over 5000 posts)
Posts: 9,685
Member is Online
|
Post by hoyasaxa2003 on Nov 13, 2019 14:25:46 GMT -5
I completely agree that Big East numbers are probably more relevant than overall numbers, particular since the OOC was poor last year. This year, with the better schedule, I think the OOC games will be a bit more telling.
|
|
LCPolo18
Diamond Hoya (over 2500 posts)
Posts: 4,406
|
Post by LCPolo18 on Nov 13, 2019 14:30:54 GMT -5
But too often people try to use it incorrectly. Were Blair and Malinowski the best players on the team last year? PPP differential says that they were. It’s an interesting stat, but like we have both said, it’s only one of many stats someone should look at when evaluating a player. No they weren’t but I think it’s fair to compare guys who play the same position when compare PPP. If Player A same his name is James plays 90% of possessions then his PPP diff is likely to be pretty similar to the team. That is where seeing how the team does with them off the court comes into play. If we see guys that are outliers in PPP diff we can the get a better picture of what is going on. For example if player A and B play together and have a PPP diff of 0 but Player A without B has a diff of +.1 and Player B without A has a diff of -.1 then we can see where the problem lies. Or if Player A and C have a diff of +.1 then A and C should play together more than A and B. Single player PPP over the course of a season where there are enough possessions to make judgement gives us a good idea. I don’t think we are there after 2 games but I also don’t think it takes an entire of year of games to get that data. Last year Leblanc had a better PPP diff than Mourning who he eventually replaced. Mosley had a better PPP diff than Pickett who he eventually replaced. I don’t think anyone here would argue those moves didn’t make the team better. My only point was that individual PPP differential is not the end all be all statistic. I'm not against using it, and just like any stat, it can lead to further investigation and it should be considered with other factors. Just as you mentioned, looking at PPP for player combinations is an example of adding other factors to individual PPP differential to better understand the situation. I'm not here to discredit statistics or dismiss certain statistics. As a career data scientist, I just want to make sure my friends here on this board don't come to premature conclusions based on a narrow view of the data.
|
|
rhw485
Silver Hoya (over 500 posts)
Posts: 740
|
Post by rhw485 on Nov 13, 2019 14:46:16 GMT -5
Ignoring the small sample size issues, as that will sort themselves out over time, there are a few issues that will never be solved by the analytics piece alone. 1. the PPP issue discussed. Yes, when comparing Blair to McClung last year as an example, the metric will never be able to take into account that McClung was going against starters and Blair usually was not. Teams stagger their rotations differently, Georgetown changes its rotation over the season as well, it will never be a true apples to apples comparison 2. Efficiency vs. Usage. Just because a player is highly efficient, it doesn't mean they're our best offensive player if their usage is very low. Josh is a perfect example of this. There were some arguments last year that Josh was our best offensive player last year based on his ORating was the highest. But his usage rate was 16%, the 18th percentile in college hoops. We can't simply have Josh shoot more, his baskets were the result of dump offs or offensive rebounds. I love Josh as a player and he's critical to our team. It's just an example where the metrics are missing a critical piece of how the basket was really created. That being said, the only way to actually prove whether the metric is insightful or not is to actually experiment w different lineups to try and prove out. Does Blair PPP change if given more minutes w starters? Does a player maintain efficiency levels when given more opportunities? Does it match the eye test of what we're seeing on court? The answer isn't to ignore them, it's to try and see whether there's something real that can help the team w different lineup construction / minutes distribution. Ewing going deep in these first two games is great, but he didn't really try different lineups. He had primarily starter lineups and bench lineups and I think that was a missed opportunity to see how the pieces could fit together. A few things: 1. While PPP does not account for your opponents, I think the argument above is overstated. Sure, McClung started. But he also played with the bench at times, and Blair often played with a bunch of the starters. Basketball isn't hockey, most teams play relatively short rotations, so while there's some truth to the statement, I think it's really a minor factor overall, unless, for example, a guy like a true walk-on only plays during garbage time. 2. Again, your point on usage is not necessarily correct. Sure, LeBlanc's strength was that he cleaned up bad shots, took high percentage shots, etc. I disagree that you can't just have LeBlanc shoot more. You absolutely can, by running plays for him, and having him be a bigger part of the offense. Will he be as good? I don't know because he's never been used in that capacity, but if you look at the guy (again, stepping away from analytics for a moment), it's clear the guy is one of those players who gets it. He's not going to suddenly start taking awful shots (like, say, Trey Mourning did) because they allow him more opportunities. (Good example: In 2012-2013, Otto Porter's usage went way up AND so did his efficiency. Increasing usage does not always mean that efficiency goes down.) 3. I know McClung is popular (for good reason) and Blair less so, but I don't think the gulf in talent last year between McClung and Blair was as big as people think. Yes, they are very different players, and McClung is certainly more flashy, but there's at least an argument that Blair was the better player (not saying I think so, but it's a legitimate position). I am not saying I agree with it, but Blair had higher efficiency, shot threes better, turned it over less, and rebounded better. Personally, I think McClung is more valuable because he's a more versatile player and can do more, plus I think his ceiling is way higher. I would also note that PPP differential takes into account defense, which I think most fans either largely ignore or are poor at examining. I think most of HoyaTalk would agree that McClung was a pretty bad defender last season. Blair certainly wasn't amazing, but that's also part of the reason why the PPP differential favored Blair in some ways. While it was en vogue to rip Govan for everything relating to defense, guys like McClung were significant problems on that end, and that's where something like PPP can expose a player where it might not be immediately evident just by watching. On the PPP, I'm not trying to throw out the metric entirely, I think it does have value. Just pointing out that there's no possible way to account for everything. That's where lineup analysis can start to help when we have enough data. Again, use the metric to help investigate new things to experiment with to see if the benefits hold. On Josh, and again I love Josh, we agree that we don't know how he would handle a different role in the offense. My point is we simply can't assume it will translate. It's certainly possible he added new skills over the summer and would love to see us give him an opportunity to show it. Love Otto and yes he absolutely took a leap simultaneously in usage and efficiency as his skill set and game increased. Nowhere above did I say that the two are always inversely correlated. The only way to know whether a player is capable of handling that increase is to give them the opportunity. I'm all for trying new things on both offense an defense.
|
|
jwp91
Platinum Hoya (over 5000 posts)
Posts: 5,892
Member is Online
|
Post by jwp91 on Nov 13, 2019 16:15:59 GMT -5
I would also note that PPP differential takes into account defense, which I think most fans either largely ignore or are poor at examining. I think most of HoyaTalk would agree that McClung was a pretty bad defender last season. Blair certainly wasn't amazing, but that's also part of the reason why the PPP differential favored Blair in some ways. While it was en vogue to rip Govan for everything relating to defense, guys like McClung were significant problems on that end, and that's where something like PPP can expose a player where it might not be immediately evident just by watching. Yes PPP differential takes into account defense. But if someone is going to look at PPP differential, they should also look at the PPP offense and PPP defense numbers. Blair's defense in the Big East was incredibly bad by this measure, even compared to Mac who was generally in line with the rest of the team. Blair All Games Differential: 0.08 All Games Offense: 1.1 All Games Defense: 1.02 Big East Differential: 0.0 Big East Offense: 1.11 Big East Defense: 1.11 McClung All Games Differential: 0.0 All Games Offense: 1.01 All Games Defense: 1.01 Big East Differential: -0.03 Big East Offense: 0.99 Big East Defense: 1.02 Can we talk about how to interpret this? I think I understand PPP differential which I interpret as the difference between PPP when on offense vs. defense. If that is the case, I am not sure I understand PPP offense differential. Differential between ? and ? Thanks in advance.
|
|
LCPolo18
Diamond Hoya (over 2500 posts)
Posts: 4,406
|
Post by LCPolo18 on Nov 13, 2019 16:41:05 GMT -5
Yes PPP differential takes into account defense. But if someone is going to look at PPP differential, they should also look at the PPP offense and PPP defense numbers. Blair's defense in the Big East was incredibly bad by this measure, even compared to Mac who was generally in line with the rest of the team. Blair All Games Differential: 0.08 All Games Offense: 1.1 All Games Defense: 1.02 Big East Differential: 0.0 Big East Offense: 1.11 Big East Defense: 1.11 McClung All Games Differential: 0.0 All Games Offense: 1.01 All Games Defense: 1.01 Big East Differential: -0.03 Big East Offense: 0.99 Big East Defense: 1.02 Can we talk about how to interpret this? I think I understand PPP differential which I interpret as the difference between PPP when on offense vs. defense. If that is the case, I am not sure I understand PPP offense differential. Differential between ? and ? Thanks in advance. My bad, I was just showing the underlying PPP on offense vs PPP on defense numbers behind the PPP differential figures from the article on Casual Hoya. My point was that based on PPP differential, Blair appears to have better results than McClung. However if you look at the PPP on offense and PPP on defense figures that go into the PPP differential you'll see that Blair's PPP on defense is lower than McClung's, which is something that should be considered before stating that Blair should get more playing time.
|
|
jwp91
Platinum Hoya (over 5000 posts)
Posts: 5,892
Member is Online
|
Post by jwp91 on Nov 13, 2019 16:44:44 GMT -5
Can we talk about how to interpret this? I think I understand PPP differential which I interpret as the difference between PPP when on offense vs. defense. If that is the case, I am not sure I understand PPP offense differential. Differential between ? and ? Thanks in advance. My bad, I was just showing the underlying PPP on offense vs PPP on defense numbers behind the PPP differential figures from the article on Casual Hoya. My point was that based on PPP differential, Blair appears to have better results than McClung. However if you look at the PPP on offense and PPP on defense figures that go into the PPP differential you'll see that Blair's PPP on defense is lower than McClung's, which is something that should be considered before stating that Blair should get more playing time. My bad on reading comprehension. I thought I saw extra differential descriptions. Makes perfect sense now.
|
|
rhw485
Silver Hoya (over 500 posts)
Posts: 740
|
Post by rhw485 on Nov 15, 2019 14:04:09 GMT -5
I know it was mentioned in the game thread. But re-stating here (h/t Nolan from twitter):
1. Akinjo / LeBlanc / Yurtseven played two possessions together 2. Josh was even in +/- in his 20 min, which is more impressive considering fact 1.
Don't think we need an advanced analytics discussion to highlight the obvious issues with the rotation
|
|
hoyasaxa2003
Platinum Hoya (over 5000 posts)
Posts: 9,685
Member is Online
|
Post by hoyasaxa2003 on Nov 18, 2019 11:04:53 GMT -5
Does anybody know the difference in PPP between when LeBlanc is on/off the floor?
|
|
jwp91
Platinum Hoya (over 5000 posts)
Posts: 5,892
Member is Online
|
Post by jwp91 on Nov 18, 2019 11:44:01 GMT -5
Does anybody know the difference in PPP between when LeBlanc is on/off the floor? Offense 1.04. Vs. .95 Defense. .89 Vs. .96
|
|
jwp91
Platinum Hoya (over 5000 posts)
Posts: 5,892
Member is Online
|
Post by jwp91 on Nov 18, 2019 11:46:51 GMT -5
Only Alexander, Gardner, and Blair are negative vs. off the floor for both offense and defense.
|
|