[GDT] Kraken @ Flames 7pm MT

[Enoch Root]

Originally Posted by Bingo

Stats are an important tool to help eliminate bias in eye test.

Sample size is key for sure.

Also context ... stat selection.

What kind of player is it? If it's a fourth line player don't compare xGF% because they don't generate much xGF60 in their role. But how do they compare in xGA60 against their teammates knowing they have a sheltered role? How do they look against other fourth liners?

With the right context a stat isn't wrong. It's just important or not important.

Can you use one game sample sizes? Of course you can. If player x was on the ice for 8 HD chances against and no other player was on the ice for more than 2 it was likely a bad game.

The nature of the stat doesn't pin it on the single player though. So that's why longer data sets make more sense. If he's on the ice for more HD against over 40 games than any other player? Probably an issue.

We should try and think these things through, and hopefully land on something more constructive than "the stats are BS, I'll trust my eye test thanks!"

Such a frustrating post because some of the comments are good and worth discussing, and some of them are not.

Yes, stats can be wrong, and often are.

Can you use one game sample sizes? Yes, for some things. But first, you always have to be careful with small sample sizes, and second, many of the stats that are quoted on here with single-game samples, should NOT be used in samples that small.

You use one example as an argument on behalf of all stats, but it isn't - each stat needs to be analyzed and evaluated on its own. Just because one example can survive a single-game data set, does not mean others can.

And rebutting any criticism with the standard: it's better than the "I'll trust my eye test" is not only lazy but is the opposite of what you are trying to claim with "land(ing) on something more constructive".

We've been around this a hundred times. You have faith that the stats are good enough. I think stats need to be held to a higher standard, because most people don't know how to interpret them, and put far too much faith in the results. And since they do, 'good enough' and 'they're the best we're got' isn't actually good enough in many cases. It is completely reasonable - and accurate - to criticize them when they warrant it.

[Bingo]

Originally Posted by Enoch Root

Such a frustrating post because some of the comments are good and worth discussing, and some of them are not.

Wow ... thanks for the post grading! Sorry to waste your time with some of my dribble that isn't worth discussing

Originally Posted by Enoch Root

Yes, stats can be wrong, and often are.

Probably splitting hairs, but I wouldn't phrase it that way.

A stat can't be wrong. It's use can be wrong.

If player x was on the ice in one game for more HD chances against ... he was. It's set as a count based on parameters and he fell into those parameters more often than his teammates.

But if you concluded on 8 minutes of ice time that player X is the worst defensive player in the league you've overstepped a stat that isn't wrong.

Originally Posted by Enoch Root

Can you use one game sample sizes? Yes, for some things. But first, you always have to be careful with small sample sizes, and second, many of the stats that are quoted on here with single-game samples, should NOT be used in samples that small.

You use one example as an argument on behalf of all stats, but it isn't - each stat needs to be analyzed and evaluated on its own. Just because one example can survive a single-game data set, does not mean others can.

You know I said the same above right?

Originally Posted by Enoch Root

And rebutting any criticism with the standard: it's better than the "I'll trust my eye test" is not only lazy but is the opposite of what you are trying to claim with "land(ing) on something more constructive".

Which I don't do.

I push back on the "stats are crap I'll trust me eye test!" crowd for sure. But my push back is usually how I look to stats to see if they agree with MY eye test. Sometimes they do. Sometimes they don't. But most times they make me look (eye test round two) at the player a different way.

I'm never as simple as the laziness you suggest.

Originally Posted by Enoch Root

We've been around this a hundred times. You have faith that the stats are good enough. I think stats need to be held to a higher standard, because most people don't know how to interpret them, and put far too much faith in the results. And since they do, 'good enough' and 'they're the best we're got' isn't actually good enough in many cases. It is completely reasonable - and accurate - to criticize them when they warrant it.

Not sure I've said the stats are "good enough". I know I've suggested the can/should/will get better many times.

They are the best we've got and help reinforce or refocus eye test.

[Just a guy]

Originally Posted by Enoch Root

Such a frustrating post because some of the comments are good and worth discussing, and some of them are not.

Yes, stats can be wrong, and often are.

Can you use one game sample sizes? Yes, for some things. But first, you always have to be careful with small sample sizes, and second, many of the stats that are quoted on here with single-game samples, should NOT be used in samples that small.

You use one example as an argument on behalf of all stats, but it isn't - each stat needs to be analyzed and evaluated on its own. Just because one example can survive a single-game data set, does not mean others can.

And rebutting any criticism with the standard: it's better than the "I'll trust my eye test" is not only lazy but is the opposite of what you are trying to claim with "land(ing) on something more constructive".

We've been around this a hundred times. You have faith that the stats are good enough. I think stats need to be held to a higher standard, because most people don't know how to interpret them, and put far too much faith in the results. And since they do, 'good enough' and 'they're the best we're got' isn't actually good enough in many cases. It is completely reasonable - and accurate - to criticize them when they warrant it.

It seems to me that both can be right. A small sample measured against a larger data set tells you if its within a standard deviation. If outside (outlier) then likely either a good or bad game depending on the direction of the outlier.

So if I hear ER right, he believes a larger data set gives a better representation, which is true, but does not address a single games performance.

So using the single game against the larger season/career stats give you an indication how the player did in a particular game. This does not take in that quality of competition can vary greatly, which is another variable, but could give an indication of the level of the player (always plays well against lower quality teams, but badly against better quality).

Agree stats can be fun, but as Mark Twain said " there are lies, damn lies and statistics!".

[Enoch Root]

A stat can be wrong. Using your example, the definition of an HD chance isn't perfect. And a shot that took place from an HD location may not have been dangerous at all (due to time and space). So even the simple counting of HD chances for and against can result in an inaccurate number.

And that illustrates the primary problem with sample size: the quality, or accuracy of events is not consistent. Over larger sample sizes, that should correct, to an extent. But in small sample sizes it doesn't.

Then there is randomness, which is a factor in all things, and a very significant factor in hockey.

But let's use one of the stats that started this conversation: Parekh's individual PP contribution (which was on the chart as a small negative). What is included in that stat? Do we know that the things included are the right things? Even if they are the right things (which is of course impossible to define), were they all counted? Some items would be more nuanced, such as changing position to open lanes - were those all counted? How far do you have to move in order for it to count? What if you made a nice play, but at the same time , another player coughed up the puck and it went the other way - did your nice play get counted?

And it goes on and on and on. So yes, stats can be wrong, and can be garbage (and that's just one example of how)

[D as in David]

Originally Posted by Just a guy

It seems to me that both can be right. A small sample measured against a larger data set tells you if its within a standard deviation. If outside (outlier) then likely either a good or bad game depending on the direction of the outlier.

So if I hear ER right, he believes a larger data set gives a better representation, which is true, but does not address a single games performance.

So using the single game against the larger season/career stats give you an indication how the player did in a particular game. This does not take in that quality of competition can vary greatly, which is another variable, but could give an indication of the level of the player (always plays well against lower quality teams, but badly against better quality).

Agree stats can be fun, but as Mark Twain said " there are lies, damn lies and statistics!".

The eye test agrees, while, statistically, their arguments also stand up.

[PepsiFree]

Originally Posted by Enoch Root

Wow. This is all in your head. I am not being coy, or pretending anything, or trying to have secrets. I don't get upset over comments on CP (as you have tried to imply at other times). And I'm not acting in any way or trying to put on a show - I simply commented on some stats because I had an opinion to share. And I don't owe you or anyone else an explanation for my comments.

You don’t owe anyone anything, but this is generally how a normal conversation goes. You offer an opinion, or respond to someone’s comment, and they respond to yours, and it goes back and forth like that.

Much of these assumptions probably are in my head, but I am left to guess at these things since instead of offering a normal response in the casual flow of a conversation, you became cagey and started on some fantasy about how people asking you questions was weird and creepy, which is definitely in your head.

Originally Posted by Enoch Root

If you in fact don't know how small sample sizes can be wildly misrepresentative, and you are interested, I would suggest you look into it.

I know plenty about it, thanks. I’m asking you if you can tell me, not using truisms or generic comments about statistics in general, why single game statistics are not representative of that single game. A sweeping generalization about small sample sizes is useless to me, assume we’re past that. Get specific. Or get even more specific and tell me why Dom’s specific stat is broken.

Originally Posted by Enoch Root

But I have also found that bringing these things up is completely pointless and frustrating, and the rebuttals usually involve comments about eye tests and personal bias (which have some validity, but also completely miss the point)

I’ve asked you some pretty specific questions that you’ve totally ignored, so I’m not sure hand waving away response you aren’t actually receiving is relevant to this conversation.

So let’s get back on track, put away the “weird” or “creepy” or whatever you want to use as a deflection and just talk about the specific stat you called garbage and why you actually think that, with some level of detail.

[Bingo]

Originally Posted by Enoch Root

A stat can be wrong. Using your example, the definition of an HD chance isn't perfect. And a shot that took place from an HD location may not have been dangerous at all (due to time and space). So even the simple counting of HD chances for and against can result in an inaccurate number.

And that illustrates the primary problem with sample size: the quality, or accuracy of events is not consistent. Over larger sample sizes, that should correct, to an extent. But in small sample sizes it doesn't.

Then there is randomness, which is a factor in all things, and a very significant factor in hockey.

But let's use one of the stats that started this conversation: Parekh's individual PP contribution (which was on the chart as a small negative). What is included in that stat? Do we know that the things included are the right things? Even if they are the right things (which is of course impossible to define), were they all counted? Some items would be more nuanced, such as changing position to open lanes - were those all counted? How far do you have to move in order for it to count? What if you made a nice play, but at the same time , another player coughed up the puck and it went the other way - did your nice play get counted?

And it goes on and on and on. So yes, stats can be wrong, and can be garbage (and that's just one example of how)

Semantics I guess.

We both agree that stats should and will improve, but I wouldn't call a HD count wrong.

I'd just lean on what it says.

Player X was on the ice for a count of X of the following occurrence. In this case the HD stat is a shot from the home plate that came on a tip, rebound or pass into the home plate.

Once you meet that criteria of course there is a range of danger within that count, but it doesn't make the original count wrong or the stat wrong. It's better to not let tips, rebounds and passes into the most dangerous scoring area on the ice.

Players that give up more of those are probably not playing as well as players that give up fewer.

[Macindoc]

Originally Posted by Bingo

Semantics I guess.

We both agree that stats should and will improve, but I wouldn't call a HD count wrong.

I'd just lean on what it says.

Player X was on the ice for a count of X of the following occurrence. In this case the HD stat is a shot from the home plate that came on a tip, rebound or pass into the home plate.

Once you meet that criteria of course there is a range of danger within that count, but it doesn't make the original count wrong or the stat wrong. It's better to not let tips, rebounds and passes into the most dangerous scoring area on the ice.

Players that give up more of those are probably not playing as well as players that give up fewer.

I think that the problem here is that some stats don't measure what they claim to measure. For example, HD chances are determined only by the location from which a shot originated. They don't take into account where the puck was prior to the shot being taken (immediately antecedent lateral puck movement, for example), the position of defenders at the time the shot was taken (potential for blocking a shot or screening the goaltender), position of teammates when a shot was taken (were they screening the goaltender, or did he have to make adjustments for a possible tip or to control the rebound), or the position of the goaltender when a shot was taken. They also don't take into account individual players' characteristics (is it Ovie shooting from the faceoff dot, or Adam Pelech; is the goaltender particularly good at lateral movement, or particularly susceptible to back door shots, etc). So for HD shots to become a better representation of what it claims to measure, the definition of the statistic needs to be updated to take into account these other factors that affect the probability that a player will score on any given opportunity.

As you say, stats can and will improve as they take into account more of the factors that can affect outcomes.

[PepsiFree]

Originally Posted by Macindoc

I think that the problem here is that some stats don't measure what they claim to measure. For example, HD chances are determined only by the location from which a shot originated. They don't take into account where the puck was prior to the shot being taken (immediately antecedent lateral puck movement, for example), the position of defenders at the time the shot was taken (potential for blocking a shot or screening the goaltender), position of teammates when a shot was taken (were they screening the goaltender, or did he have to make adjustments for a possible tip or to control the rebound), or the position of the goaltender when a shot was taken. They also don't take into account individual players' characteristics (is it Ovie shooting from the faceoff dot, or Adam Pelech; is the goaltender particularly good at lateral movement, or particularly susceptible to back door shots, etc). So for HD shots to become a better representation of what it claims to measure, the definition of the statistic needs to be updated to take into account these other factors that affect the probability that a player will score on any given opportunity.

As you say, stats can and will improve as they take into account more of the factors that can affect outcomes.

I think the actual problem is that some stats measure what they claim to measure, instead of what people think they should measure.

High danger shots are, quite literally, shots taken from an area of the ice that has a higher probability of turning into a goal.

Adding something like whether it’s Ovie or Pelech shooting actually makes the stat more convoluted, and makes it so it no longer measures what it is claiming to measure.

Adding more complexity does not necessarily make a stat better. You’re not asking the stat to measure what it says, you’re asking it to measure something different entirely.

[Bingo]

Originally Posted by Macindoc

I think that the problem here is that some stats don't measure what they claim to measure. For example, HD chances are determined only by the location from which a shot originated. They don't take into account where the puck was prior to the shot being taken (immediately antecedent lateral puck movement, for example), the position of defenders at the time the shot was taken (potential for blocking a shot or screening the goaltender), position of teammates when a shot was taken (were they screening the goaltender, or did he have to make adjustments for a possible tip or to control the rebound), or the position of the goaltender when a shot was taken. They also don't take into account individual players' characteristics (is it Ovie shooting from the faceoff dot, or Adam Pelech; is the goaltender particularly good at lateral movement, or particularly susceptible to back door shots, etc). So for HD shots to become a better representation of what it claims to measure, the definition of the statistic needs to be updated to take into account these other factors that affect the probability that a player will score on any given opportunity.

As you say, stats can and will improve as they take into account more of the factors that can affect outcomes.

Completely agree.

And I look forward to the improvement in separating a simple count into shades of grey.

But can't we also agree that simple count of an event (shot from the most dangerous place on the ice) is a bad thing when it's in front of your goalie, and if you give up more of those than others you probably will get scored on more often?

[Macindoc]

Originally Posted by Bingo

Completely agree.

And I look forward to the improvement in separating a simple count into shades of grey.

But can't we also agree that simple count of an event (shot from the most dangerous place on the ice) is a bad thing when it's in front of your goalie, and if you give up more of those than others you probably will get scored on more often?

It's more often a bad thing than a random shot from elsewhere. But I would argue that a shot from a slightly less dangerous spot immediately following a cross-ice pass, or a shot taken from anywhere in front of the net immediately after the goaltender lost the puck behind the net, would be more dangerous.

It's hard to find the right balance of complexity with stats. Basic stats, like shots on goal, clearly miss too much. Highly convoluted and derived stats lose their connection with what is happening in real time, and represent increasingly abstract and hard to understand constructs.

[Enoch Root]

Looking at HD chances for and against, as a specific example of how a stat can be wrong, as well as being just not very useful:

1. Small sample sizes can be problematic (in this case, one, or a few games)
2. Small sample sizes of stats with very few data points can be very problematic (typically a player will have between zero and 6 to 8 HD chances for or against, in a game)
3. Converting very small numbers into percentages, is just flat out dangerous (example: player has 3 HD chances for, and 4 against in a game, so they are 43% - sounds like they had a pretty terrible game)
4. The actual events being counted are flawed - a tip is an HD chance, regardless of how weak the situation was. Conversely, Coronato's goal against the Kraken was not an HD chance because he was off to the side of the net, even though that chance was literally the best scoring chance in the game (open net play)
5. There is no ability in the stat to consider times and space, which, IMO is even more important than location (which is essentially the only variable considered). As an example, on the 3rd VAN goal last night, the opportunity wasn't all that dangerous, except for the fact that the shooter had all day to look at the goalie, pick a target, and get a good shot off, towards that target. The primary factor there was time, not location (as it usually is in hockey)
6. There is no ability to factor in the quality of the player and their shooting ability, or their ability to get a shot off quickly, or their ability to create open space (and thus limiting another player's ability to generate HD stats)
7. Bounces (randomness). A perfect play can be thwarted by bad ice. Conversely, a great chance can be the result of a lucky bounce. When the sample size is only a few events, random bounces can literally be the primary, or even only, determinant.

When you factor all these issues in, and then you consider that in any given game, a player is only on the ice for a few HD shots for and against, it is extremely likely for the numbers to be misleading - especially when presented as percentages.

As an example, a line has a pretty dominant night, having possession in the offensive zone for several shifts in a row. And because it is a play that has been working, they keep setting up the winger with one-timers from the side, none of which count as HD chances. They also set up multiple shots from the point, none of which get through and/or get tipped on net. So they only actually take 3 HD shots over the course of those shifts (despite prolonged domination, and lots of actual chances). Then, one bad turnover, and the other team has possession for 5 seconds, and gets a not-very-dangerous HD shot off (rushed), followed by a couple of (not very dangerous) rebounds, then a weak shot from the point that gets tipped, and suddenly the line that has been completely dominant, is sitting at 43% in HD chances for and against. "Hmm, I thought they played pretty well, but I guess I'm just biased because the stats say otherwise."

[Bingo]

I agree with a lot of the above honestly.

But can you agree it's better to have a lower count of shots against right in front of your goaltender than more?

[Enoch Root]

Originally Posted by PepsiFree

I think the actual problem is that some stats measure what they claim to measure, instead of what people think they should measure.

High danger shots are, quite literally, shots taken from an area of the ice that has a higher probability of turning into a goal.

Adding something like whether it’s Ovie or Pelech shooting actually makes the stat more convoluted, and makes it so it no longer measures what it is claiming to measure.

Adding more complexity does not necessarily make a stat better. You’re not asking the stat to measure what it says, you’re asking it to measure something different entirely.

Yes, sometimes, more complex isn't better (but sometimes it is)

And yes, sometimes it becomes an issue of what people want it to measure, as opposed to what it does measure.

But this stat measures HD chances. Yes, it defines that as shots from home plate. But that doesn't mean that is should. Sometimes the design of the stat can be flawed, or too simplistic. Again, Coronato's goal against the Kraken was a very high danger scoring chance in any rational person's mind. But it wasn't a stat, by the way it is counted. That isn't a problem of the viewer (wanting the stat to be what they want it to be), it is a shortcoming of the stat. Full stop.

[Bingo]

Originally Posted by Enoch Root

Yes, sometimes, more complex isn't better (but sometimes it is)

And yes, sometimes it becomes an issue of what people want it to measure, as opposed to what it does measure.

But this stat measures HD chances. Yes, it defines that as shots from home plate. But that doesn't mean that is should. Sometimes the design of the stat can be flawed, or too simplistic. Again, Coronato's goal against the Kraken was a very high danger scoring chance in any rational person's mind. But it wasn't a stat, by the way it is counted. That isn't a problem of the viewer (wanting the stat to be what they want it to be), it is a shortcoming of the stat. Full stop.

Is it though?

The origin of the HD stat is by percentages of goals on a shot attempt on all areas of the ice.

I'd guess that shots from where Coronato shot are lower in danger because of less net to shoot at, the goalie covering more of the space, and the higher likelihood of shooting wide.

They all contribute to a lower probability of a shot from their going in.

[Enoch Root]

Originally Posted by Bingo

Completely agree.

And I look forward to the improvement in separating a simple count into shades of grey.

But can't we also agree that simple count of an event (shot from the most dangerous place on the ice) is a bad thing when it's in front of your goalie, and if you give up more of those than others you probably will get scored on more often?

I look forward to a day where AI can track time and space, and the quality of the shot (vs rolling puck, for instance), and factor all those things into the event. And for it to continue to learn, and improve on itself to the point where it can actually measure HD chances in a meaningful way. And when that happens, I will love the stat.

But that day is not now.

As to the last paragraph, yes we can all agree that chances against your goalie are bad, and more of them are probably more bad. But that does not justify or validate the stat, which is doing a pretty lousy job of quantifying what it is trying to quantify, and representing how individual players faired, over the course of a game.

[Bingo]

Originally Posted by Enoch Root

I look forward to a day where AI can track time and space, and the quality of the shot (vs rolling puck, for instance), and factor all those things into the event. And for it to continue to learn, and improve on itself to the point where it can actually measure HD chances in a meaningful way. And when that happens, I will love the stat.

But that day is not now.

As to the last paragraph, yes we can all agree that chances against your goalie are bad, and more of them are probably more bad. But that does not justify or validate the stat, which is doing a pretty lousy job of quantifying what it is trying to quantify, and representing how individual players faired, over the course of a game.

Getting to the crux of it now.

I don't think a player that has too many shots right in front of their own net can conclusively be called weak defensively.

There are circumstances.

But for whatever reason more bad things happened to him than others when he was on the ice (others would be connected to those events as well) which meant defensively it likely wasn't a great game.

And to suggest you can have more bad things happen in multiple games and still be a victim of a bad stat doesn't wash for me.

[Enoch Root]

Originally Posted by Bingo

Is it though?

The origin of the HD stat is by percentages of goals on a shot attempt on all areas of the ice.

I'd guess that shots from where Coronato shot are lower in danger because of less net to shoot at, the goalie covering more of the space, and the higher likelihood of shooting wide.

They all contribute to a lower probability of a shot from their going in.

You defend the stats in every situation - it is okay to criticize them, in fact that is an important part of the process of deriving better stats.

Above, you are continuing to argue for location as the most important factor (as the stat does). But people keep giving examples of how that is not always the case - and in fact may not be the most important variable at all. Coronato's chance was a very high danger chance because Frost's cross-crease pass got through, and put the goalie out of position. That left Coronato to shoot at an empty net. The important variable here was the set-up. Location, in this case, was irrelevant - especially for an NHLer.

And has been stated many times, time and space can also be more important than location. As can the skill of the shooter.

So yes, more shots from close in is better (or worse). But HD scoring chances is a comparative stat that is trying to quantify how well a player performed (and that is how people use it). And it simply doesn't do a good job of doing that.

[Enoch Root]

Originally Posted by Bingo

Getting to the crux of it now.

I don't think a player that has too many shots right in front of their own net can conclusively be called weak defensively.

There are circumstances.

But for whatever reason more bad things happened to him than others when he was on the ice (others would be connected to those events as well) which meant defensively it likely wasn't a great game.

And to suggest you can have more bad things happen in multiple games and still be a victim of a bad stat doesn't wash for me.

Larger sample sizes (and this stat needs many games to achieve that) will compensate for randomness for sure. But they won't compensate for everything - some players are better passers (more HD chances for, for all players on the ice with them). Some players are better at creating time and space for themselves (which will help their line-mates' stats). Some players like to shoot from outside home plate. The point being, even with larger sample sizes, the stat is severely limited, and (I would argue) flawed. But this stat is rarely used with large sample sizes, it is almost exclusively used on single game discussions.

[InternationalVillager]

Originally Posted by Enoch Root

Looking at HD chances for and against, as a specific example of how a stat can be wrong, as well as being just not very useful:

1. Small sample sizes can be problematic (in this case, one, or a few games)
2. Small sample sizes of stats with very few data points can be very problematic (typically a player will have between zero and 6 to 8 HD chances for or against, in a game)
3. Converting very small numbers into percentages, is just flat out dangerous (example: player has 3 HD chances for, and 4 against in a game, so they are 43% - sounds like they had a pretty terrible game)
4. The actual events being counted are flawed - a tip is an HD chance, regardless of how weak the situation was. Conversely, Coronato's goal against the Kraken was not an HD chance because he was off to the side of the net, even though that chance was literally the best scoring chance in the game (open net play)
5. There is no ability in the stat to consider times and space, which, IMO is even more important than location (which is essentially the only variable considered). As an example, on the 3rd VAN goal last night, the opportunity wasn't all that dangerous, except for the fact that the shooter had all day to look at the goalie, pick a target, and get a good shot off, towards that target. The primary factor there was time, not location (as it usually is in hockey)
6. There is no ability to factor in the quality of the player and their shooting ability, or their ability to get a shot off quickly, or their ability to create open space (and thus limiting another player's ability to generate HD stats)
7. Bounces (randomness). A perfect play can be thwarted by bad ice. Conversely, a great chance can be the result of a lucky bounce. When the sample size is only a few events, random bounces can literally be the primary, or even only, determinant.

When you factor all these issues in, and then you consider that in any given game, a player is only on the ice for a few HD shots for and against, it is extremely likely for the numbers to be misleading - especially when presented as percentages.

As an example, a line has a pretty dominant night, having possession in the offensive zone for several shifts in a row. And because it is a play that has been working, they keep setting up the winger with one-timers from the side, none of which count as HD chances. They also set up multiple shots from the point, none of which get through and/or get tipped on net. So they only actually take 3 HD shots over the course of those shifts (despite prolonged domination, and lots of actual chances). Then, one bad turnover, and the other team has possession for 5 seconds, and gets a not-very-dangerous HD shot off (rushed), followed by a couple of (not very dangerous) rebounds, then a weak shot from the point that gets tipped, and suddenly the line that has been completely dominant, is sitting at 43% in HD chances for and against. "Hmm, I thought they played pretty well, but I guess I'm just biased because the stats say otherwise."

A lot of these nuances are addressed in public vs private models. So unless you have access to private models or pay to access some, it is always best to look at multiple data sources.