[SIZE=medium]I've looked into advanced stats a little and at the risk of sounding arrogant, I think there may be better ways to statistically analyze hockey games. I've been thinking about this a lot in the past couple of months, so I would like to hear what you more knowledgeable fellows think of these ideas.[/SIZE]
[SIZE=medium]Instead of going into painstaking detail, I want to first layout the concepts I’ve been bouncing around and maybe start a conversation if this makes sense to people. My basic premise is that hockey could be more effectively analyzed through the prism of offensive zone possessions (OZP). The stat would not try to divorce offense from defense—hockey is entirely too fluid for such a baseball-like analysis. In theory, you would capture this data for a team’s OZP, as well as capturing how their opponents are performing against them, or Defensive Zone Possessions (DZP). Obviously many factors influence the effectiveness of an OZP, but the two I want to discuss are Zonal Order of Entry (ZOE) and Method of Entry [insert joke here] (MOE). I choose these 2 because I think they are potentially quantifiable and I think they would have predictive value if you could actually capture this data. [/SIZE]
[SIZE=medium]ZOE would seek to capture the order of the first few skaters into the offensive zone establishing an OZP. The reason I think this type of analysis is important might be better illustrated by way of example. [/SIZE]
[SIZE=medium]Scenario 1 – 1 skater on Team X is on a breakaway and is the first player on either Team X or Team Y to enter Team X’s offensive zone (for clarity this is Team Y’s defensive zone). Now that Team X has established an OZP, a second player on Team X is the next player on either team into the zone. For clarity, Team X now has a ‘2 on 0’ breakaway. The probability of a goal on this OZP should be relatively high.[/SIZE]
[SIZE=medium]Scenario 2 – 1 skater on Team X is on a breakaway and is the first player on either team into the zone. However, in this scenario, a defender on Team Y is the second player on either team into the zone. The third player into the zone is a second forward from Team X. This is still a very high quality opportunity, but should have a lower probability of scoring than Scenario 1.[/SIZE]
[SIZE=medium]Scenario 3 – Now imagine that a defender from Team Y is the first player on either team in the zone, but a skater from Team X then carries the puck into the zone. Following the Team X player with the puck, another player from Team X is the third player in.[/SIZE]
[SIZE=medium]Scenario 4 – A defender from Team Y is the first player from either team in the zone, then a player from Team X carries the puck into the zone. This time, a second defender from Team Y is the third player in the zone.[/SIZE]
[SIZE=medium]We can abbreviate the above 4 scenarios as follows:[/SIZE]
[SIZE=medium]X-X-Y[/SIZE]
[SIZE=medium]X-Y-X[/SIZE]
[SIZE=medium]Y-X-X[/SIZE]
[SIZE=medium]Y-X-Y[/SIZE]
[SIZE=medium]I think you’re starting to get the picture, but just to be thorough, I’ll give one more example.[/SIZE]
[SIZE=medium]Scenario 5 – 2 defenders from Team Y are in position and are the first two players from either team in the zone, then a forward from Team X carries in. This would be represented as “Y-Y-X”.[/SIZE]
[SIZE=medium]There are 10 non-goalie skaters on the ice at any given time, but I think that the 8th, 9th[/SIZE], and 10[SIZE=small]th[/SIZE] skaters to make it into the zone likely don’t have predictive value. In fact, the 7[SIZE=small]th[/SIZE] or maybe even the 6[SIZE=small]th[/SIZE] player into the zone doesn’t even matter that much. In other words, a ‘4 on 2’ and ‘3 on 1’ are scoring opportunities with potentially substantially higher statistically probabilities of ending in a successful OZP. Once you get to ‘4 on 4’ though, I’d be surprised if it mattered much. Let’s just be clear that this system would be difficult to maintain. Taking the seemingly simple situation of a ‘3 on 1’ break, the permutations could include:
[SIZE=medium]Y-X-X-X[/SIZE]
[SIZE=medium]X-Y-X-X [/SIZE]
[SIZE=medium]X-X-Y-X [/SIZE]
[SIZE=medium]Instead of going into more detail on ZOE, which you probably are already poking holes through, allow me to move onto MOE.[/SIZE]
[SIZE=medium]As I see it, there are 3 main ways a team can get the puck into the offensive zone: 1) carry it in (C), 2) connect on a pass to a teammate (P), or 3) dump it in (DI). I would classify the difference between a P and a DI as whether a player on Team X other than the passer touches the puck in the zone first (P) or whether a defender from Team Y does (DI). There may be a need for other categories, such as shots or plays along the board, but I digress. I think this is an important characteristic of an OZP because I would image that an OZP beginning with a player from Team X carrying the puck into the zone (a ‘C-OZP’) has a higher probability of success than an OZP started with a pass (a ‘P-OZP’), which in turn may have a higher probability of success than an OZP started with a DI (a ‘DI-OZP’). Because I think it's for another discussion, let's assume that a successful possession should include a shot, and a higher value given for OZPs ending in goals.[/SIZE]
[SIZE=medium]Combining these 2 metrics, I think a clearer picture would emerge as to what strategies are most effective, but it would also, at the least, show what type of strategy a team is employing. Generally, I'd want to know if C-OZPs result in more goals than DI-OZPs. More narrowly, I would guess that the Bruins have more DI-OZP than most teams. There’s a lot of interesting numbers you could crunch from this data. One example would be what percentage of a team’s OZPs are C-OZPs vs. P-OZPs vs. DI-OZPs. It may make more sense to have these stats separately at first, but if you could combine them to assign a value of a probability of success on each OZP, I think you could find some really interesting patterns potentially. Moreover, if a team is susceptible to DI-DZP because of having an undersized defense, the data may bear that out.[/SIZE]
[SIZE=medium]I acknowledge that there are problems with these metrics. For instance, the ZOE may be difficult to assess—many times, there’s just 3 guys (in some combination from both teams) steaming into the zone. But, I’ll let you guys hash that out if you so choose. [/SIZE]
[SIZE=medium]Instead of going further, I’ll just post this now and see what you all think of this. I know I’m kind of all over the place, but hopefully you see what I’m thinking and we can have an interesting discussion about how to look at the game.[/SIZE]
[SIZE=medium]PS – this is what happens when you watch 15 days in a row of world cup soccer, then have a day off.[/SIZE]
[SIZE=medium]Instead of going into painstaking detail, I want to first layout the concepts I’ve been bouncing around and maybe start a conversation if this makes sense to people. My basic premise is that hockey could be more effectively analyzed through the prism of offensive zone possessions (OZP). The stat would not try to divorce offense from defense—hockey is entirely too fluid for such a baseball-like analysis. In theory, you would capture this data for a team’s OZP, as well as capturing how their opponents are performing against them, or Defensive Zone Possessions (DZP). Obviously many factors influence the effectiveness of an OZP, but the two I want to discuss are Zonal Order of Entry (ZOE) and Method of Entry [insert joke here] (MOE). I choose these 2 because I think they are potentially quantifiable and I think they would have predictive value if you could actually capture this data. [/SIZE]
[SIZE=medium]ZOE would seek to capture the order of the first few skaters into the offensive zone establishing an OZP. The reason I think this type of analysis is important might be better illustrated by way of example. [/SIZE]
[SIZE=medium]Scenario 1 – 1 skater on Team X is on a breakaway and is the first player on either Team X or Team Y to enter Team X’s offensive zone (for clarity this is Team Y’s defensive zone). Now that Team X has established an OZP, a second player on Team X is the next player on either team into the zone. For clarity, Team X now has a ‘2 on 0’ breakaway. The probability of a goal on this OZP should be relatively high.[/SIZE]
[SIZE=medium]Scenario 2 – 1 skater on Team X is on a breakaway and is the first player on either team into the zone. However, in this scenario, a defender on Team Y is the second player on either team into the zone. The third player into the zone is a second forward from Team X. This is still a very high quality opportunity, but should have a lower probability of scoring than Scenario 1.[/SIZE]
[SIZE=medium]Scenario 3 – Now imagine that a defender from Team Y is the first player on either team in the zone, but a skater from Team X then carries the puck into the zone. Following the Team X player with the puck, another player from Team X is the third player in.[/SIZE]
[SIZE=medium]Scenario 4 – A defender from Team Y is the first player from either team in the zone, then a player from Team X carries the puck into the zone. This time, a second defender from Team Y is the third player in the zone.[/SIZE]
[SIZE=medium]We can abbreviate the above 4 scenarios as follows:[/SIZE]
[SIZE=medium]X-X-Y[/SIZE]
[SIZE=medium]X-Y-X[/SIZE]
[SIZE=medium]Y-X-X[/SIZE]
[SIZE=medium]Y-X-Y[/SIZE]
[SIZE=medium]I think you’re starting to get the picture, but just to be thorough, I’ll give one more example.[/SIZE]
[SIZE=medium]Scenario 5 – 2 defenders from Team Y are in position and are the first two players from either team in the zone, then a forward from Team X carries in. This would be represented as “Y-Y-X”.[/SIZE]
[SIZE=medium]There are 10 non-goalie skaters on the ice at any given time, but I think that the 8th, 9th[/SIZE], and 10[SIZE=small]th[/SIZE] skaters to make it into the zone likely don’t have predictive value. In fact, the 7[SIZE=small]th[/SIZE] or maybe even the 6[SIZE=small]th[/SIZE] player into the zone doesn’t even matter that much. In other words, a ‘4 on 2’ and ‘3 on 1’ are scoring opportunities with potentially substantially higher statistically probabilities of ending in a successful OZP. Once you get to ‘4 on 4’ though, I’d be surprised if it mattered much. Let’s just be clear that this system would be difficult to maintain. Taking the seemingly simple situation of a ‘3 on 1’ break, the permutations could include:
[SIZE=medium]Y-X-X-X[/SIZE]
[SIZE=medium]X-Y-X-X [/SIZE]
[SIZE=medium]X-X-Y-X [/SIZE]
[SIZE=medium]Instead of going into more detail on ZOE, which you probably are already poking holes through, allow me to move onto MOE.[/SIZE]
[SIZE=medium]As I see it, there are 3 main ways a team can get the puck into the offensive zone: 1) carry it in (C), 2) connect on a pass to a teammate (P), or 3) dump it in (DI). I would classify the difference between a P and a DI as whether a player on Team X other than the passer touches the puck in the zone first (P) or whether a defender from Team Y does (DI). There may be a need for other categories, such as shots or plays along the board, but I digress. I think this is an important characteristic of an OZP because I would image that an OZP beginning with a player from Team X carrying the puck into the zone (a ‘C-OZP’) has a higher probability of success than an OZP started with a pass (a ‘P-OZP’), which in turn may have a higher probability of success than an OZP started with a DI (a ‘DI-OZP’). Because I think it's for another discussion, let's assume that a successful possession should include a shot, and a higher value given for OZPs ending in goals.[/SIZE]
[SIZE=medium]Combining these 2 metrics, I think a clearer picture would emerge as to what strategies are most effective, but it would also, at the least, show what type of strategy a team is employing. Generally, I'd want to know if C-OZPs result in more goals than DI-OZPs. More narrowly, I would guess that the Bruins have more DI-OZP than most teams. There’s a lot of interesting numbers you could crunch from this data. One example would be what percentage of a team’s OZPs are C-OZPs vs. P-OZPs vs. DI-OZPs. It may make more sense to have these stats separately at first, but if you could combine them to assign a value of a probability of success on each OZP, I think you could find some really interesting patterns potentially. Moreover, if a team is susceptible to DI-DZP because of having an undersized defense, the data may bear that out.[/SIZE]
[SIZE=medium]I acknowledge that there are problems with these metrics. For instance, the ZOE may be difficult to assess—many times, there’s just 3 guys (in some combination from both teams) steaming into the zone. But, I’ll let you guys hash that out if you so choose. [/SIZE]
[SIZE=medium]Instead of going further, I’ll just post this now and see what you all think of this. I know I’m kind of all over the place, but hopefully you see what I’m thinking and we can have an interesting discussion about how to look at the game.[/SIZE]
[SIZE=medium]PS – this is what happens when you watch 15 days in a row of world cup soccer, then have a day off.[/SIZE]
