-
Hey, guest user. Hope you're enjoying NeoGAF! Have you considered registering for an account? Come join us and add your take to the daily discourse.
NHL Off-Season 2014 |OT2| - Dan Cleary
- Thread starter Marvie_3
- Start date
Good luck, hopefully in two years it'll get some recognition for all your hard work in this.
Could be, or could be a fruitful endeavor.
No one has said otherwise. It has to begin somewhere, and completely dismissing it while it has proved time and time again that is has merit is complete ignorance. You sound exactly like Simmons.
It is and it isn't. I can get basic data running without much analysis. (like 20-30 minutes per game)
And i still have crowd sourcing as an option... i know some people would love to do it because why not? Just want to see how much i could do alone (or at least with limited people) so that the info is at least as accurate as possible... at least until i can figure out ways to determine entry confidence
Are you going to be evaluating the 'quality of shot'?
You bet I am.
How will DopeyStats account for the fact that much of its data has been generated by the subjective perceptions of a biased hockey fan?
Unhandled Exception
Member
Clearly he's going to need to break down every single shot in video form.
Basically breaking down the offensive ice surface into zones and then bringing in a small set of variables (goalie in position, transition, or out of positions) type of shot, man in front of net, pass/one timer, whatever. Determine the shooting% for everything. Throw the shooting% back through a series of other calculations, determine missed shots with the shot quality, see the variance. Determine the average offense by adding all shots together, subtracting that figure from GF determines luck. The shot quality then can be rated across league. Shot quality then can be determined across players, and then you can see what the defense is doing. You can map the figures like corsi to get FAR better QoC and QoT figures as well as bring better context to the entire corsi suite as a whole. And for goaltenders? You can do some fun stuff like seeing their sv% against the upper half of quality shots to determine a more accurate view of their play.
He has a point there. A shot on goal and a shot that hits a defender 3 feet in front of the shooter should not be treated equally.
Christianity is used as fact throughout north america. Doesn't mean it's right either.
FUBAR McDangles
Member
Dat Minnesota Nice apology doe
You haven't addressed the problem. All you've done is break down one subjective decision into several.
I never said otherwise?
Minnesota sucks. Stop trying to make it look so gosh darn unique and wonderful.
A thousand apologies if that came across harshly!
The idea is not to determine what my opinion is. It's what to determine what is and what isn't happening.
So when I fill out the data for the shot, if the goalie is ready for a shot and not moving, the data is entered as such. Where the puck is where i place it. If there's a person in front of the net (of the shooters team), i will enter it as such. I am not determining the shot quality from opinion, but from the shooting percentage of variables. Data won't be perfect (some shots will be difficult to determine) but it paints a far more accurate picture than simply shots for and against.
KernelPanic
Member
The point of Corsi is (an attempt) to illustrate possession. The player has the puck which means the other team does not. He may take a weak shot from the point or he may keep it etc.
Sure you need to go deeper than that but it's a start and that's how far that stat goes.
FUBAR McDangles
Member
I'm just going to passive-aggressively ignore you until you realize and apologize for what you have just done.
Food for thought, the act of shooting is a player losing possession. Just because you generate more shots doesn't mean you have higher possession numbers.
Corsi isn't a possession stat, it's a shooting differential stat.
You're not a computer that can objectively evaluate reality, Dopey. You're a flawed human being with biases. In determining Yes/No for various events (eg, is there an offensive player in front of the net?) you will frequently encounter uncertain gray areas where a decision has to be made. If it's Gardiner taking the shot and it's JVR kind of in front of the net but not really, am I supposed to trust that your bias won't come into play? And that when you aggregate all of these decisions and data, that the picture you've painted is actually meaningful? You would say YES but I'm inclined to believe no.
he obviously is going to extrapolate the area the players covered between where the shot goes off and the net and then subtract the area of which the stick is for deflections, subtract an average area for where the stick COULD go in the time the puck takes to reach the player based on average reaction time of said player (based on... stats). He's going to do all this based on a 3d computing model that simulates what is happening on the ice based on a 2d recording.
Why would i have any bias determining if a player is in front of the net?
It's basically just a factor to determine the quality of that shot. It's balanced across all other shots with those same variables. The idea is... the player would have to be screening the goalie. Not to the left. Not to the right. But actively trying to block the goaltenders line of sight.
The only issue i have run across is zone selection, but it should have negligible impact across the data set.
The data that I'm forming is not very subjective. It's yes or no questions. No grey area except zoning.
I gave you an example of how you could be biased. ie, Gardiner quality of shot. You'll be the generator of some of the data and could, for example, make Leafs players look better than they actually are.
And of course there would be gray area. Many instances will be clear cut but many could also go either way and will require a subjective judgment call of Yes or No.
It wouldn't make leafs players better unless they did that one specific shot and no one else did
Basically when i'm detailing the shot, it's being added to all shots of the exact same type from the same area for all teams.
So if that one shot had a 5% shooting percentage, it's 5% shooting percentage for all. I don't rate on team by team basis except with the league wide shot quality, hence why the numbers change a year down the line because the data set is more dense and more accurate.
So the teams actual sh% with those variables will be different than the league wide data, but considering the amount of variables used, using a single teams data just wouldn't be feasible even for comparison sake.
KernelPanic
Member
The whole point of possession is to generate scoring chances(shot attempts). They're not playing keep-away. You don't win if you don't generate scoring chances. Teams with good possession metrics generate more shot attempts because they have the puck more than their opponent. Even if they lose possession taking shots(and not always you can get your own rebound etc) they are better at getting it back.
You're right, by definition it's a shooting differential stat but it's used to illustrate puck possession. If you have a positive Corsi rating it means you are generating more shot attempts than giving them up. Which means you have the puck more often than your opponent. Teams that have the puck tend to win more than they lose (over a large sample size -- ie. an 82 game season)
http://www.extraskater.com/glossary
Freyjadour
Member
"grit" is my stat of choice.
apocalidiot
Banned
TWTW
Completely false. Again. Just because you are shooting more doesn't mean you have the puck more than your opponent. It also doesn't mean they are better at retrieving the puck either. It just means they are generating more shots. That's all. Even shooting more doesn't mean you're likely going to win, either. Shooting more than your opponent also doesn't mean you're more likely to score. Allowing more shots against doesn't mean you're playing worse defensively.
It's not demonstrating possession because it's a shot differential stat. If you want possession stats... why not look at possession stats?
Just because you can see that the two (possession and shot differential) agree more often than not doesn't mean it's a stat that accurately illustrates possession. All it illustrates is that there's a higher chance that the scenarios above are true but by definition it doesn't mean that it is.
Corsi and +/- are of the same ilk.
Unhandled Exception
Member
Why couldn't you have put this at the top so I knew to stop reading then and there
You're still missing the point. I won't repeat myself again, though. No need. Let's just say I personally won't see any value in DopeyStats for the reasons I've mentioned. And so, I think it'll be a colossal waste of time for you.
On the bright side you'll get to watch a shit load of hockey. Hopefully you enjoy the process.
lol
Designed for a purpose, doesn't work for purpose. Ends up being a stat which kinda says something but really doesn't mean shit.
Despite the fact I already know it works and despite the fact that it's COMPLETELY IMPOSSIBLE TO RIG THE STATS...like holy shit, what part of league shared data set and both offensive and defensive stats derived from the same datasets(meaning there's a balance) do you not understand? I'm not going to be spitting out random stats... there's going to be game logs and breakdowns to the shots so you can sit there and whine about how accurate I keep being or something until you get bored and decide to troll someone else.
All he's saying is that things like "did he screen the goalie or not" are VERY subjective, calm your tits
is the goalie looking to the side of him? is the goalie looking over him? is he standing halfway covering the goalie? 3/4? Does the goalie's handedness matter? Is the goalie in butterfly? Is the goalie position wrongly because of the screen, or correctly despite of it, or unrelated?
Yeah obviously there are no variables involved in situations like that
you do realize that the following parts:
Zone (split into 5, 10 if not using mirrored), shot type, goalie in position (3 choices to limit "subjectivity"), goalie obstructed, one timer is effectively 120 (or 240) different possible shots?
the concept behind the obstruction was to have two sides. Goalie with a clear lane, Goalie without. you get into very miniscule details like that and you're entirely screwing the data set (limiting usable data) or you're not understanding the purpose.
that's Soc's and my entire argument.
How are you going to determine if the goalie's lane was clear or not. And don't fucking say "if a player is obstructing his view" because that also depends on a lot of things.
But I'll do what Soc did and just leave you to yourself now too.
Can I dump him on someone else? Please? I didn't know what I was getting myself into.
Wow, It's like I'm arguing with a bunch of preschoolers.
IS THERE A FUCKING PLAYER DIRECTLY IN FRONT OF GOALIE?
It doesn't matter if he's fucking dancing
It doesn't matter if the goalie is moving his head around like an owl
It doesn't. Fucking. Matter.
The whole purpose of that variable is 1) to make sure that those plays don't inflate clean shots so we get clearer picture 2) to demonstrate the value (if any) of putting someone in front of the net
It's the event that matters, not the minor details. That's why i am saying there is no subjectivity. Is there a guy in front of the goalie? Yes or no.
IS THERE A FUCKING PLAYER DIRECTLY IN FRONT OF GOALIE?
It doesn't matter if he's fucking dancing
It doesn't matter if the goalie is moving his head around like an owl
It doesn't. Fucking. Matter.
The whole purpose of that variable is 1) to make sure that those plays don't inflate clean shots so we get clearer picture 2) to demonstrate the value (if any) of putting someone in front of the net
It's the event that matters, not the minor details. That's why i am saying there is no subjectivity. Is there a guy in front of the goalie? Yes or no.
