Clearing up some questions about scoring chances

Read this today by mORRganRielly over at Maple Leafs Hot Stove:

at the 36 game mark, the Leafs did actually out-chance their opponents 474-469. Edit: I erred on my interpretation of these numbers — the original included special teams. The actual numbers available to us are 392 against and 355 for on even-strength. However, this isn’t the chasm inferred via shots total against. But it does leave me wondering what Cam’s final scoring chance counter was.

The Leafs out-chancing the opposition doesn’t necessarily mean anything. But it does give credence that the coaching staff actually knows what they want out of their line-up, how to get it, and what they are tracking. Oddly enough, the narrative that Carlyle and Co. don’t have a clue what they are doing come from the lowest denominator of an internet arm-chair general manager from that other place.

It’s wonderful that people are using my data, I guess. Since the Greg Cronin interview came out last week, a lot of people have asked me if I have a final tally for the Leafs scoring chances numbers for and against on the season. They came out well-below even at even strength, having a brutal final 12 games of the season in chances. I don’t have an exact number because the other computer is in the other room and there’s some element of laziness because I’m trying to vacation and stay away from hockey for a week. I can’t quit you guys, but the 36-game update is all we have for some time.

For those not in the know, every Leafs game this year except for two, I sat in front of my TV and noted scoring chances. It’s a simple process. After the game, I go to, check to see who was on the ice for each chance, and have a table at the bottom of games that gives each player’s “plus” and “minus”. It’s a good way to get people interested in analytics that’s more accessible than shot statistics.

Other than the fact that they could be score effects-neutral compared to shot statistics, there’s really not much value in these numbers. While it may be counter-intuitive, there’s no current evidence to suggest that players affect the save percentages of their goaltender. It is true that certain players allow a disproportionate number of shots in the slot or odd-man rushes when they’re on the ice, but that doesn’t mean that it’s sustainable. As Eric. T put it on NHLNumbers when this was a debate last summer, “Whatever tendency certain players might have for driving their team to get more scoring chances than a simple shot differential predicts is small and swamped by random noise”. If you want to draw any conclusions from my scoring chances data, make sure you read that post first.

Phil Kessel may take 20 shots, but only 5 of them are scoring chances, compared to, say, Nik Kulemin, who takes 10 shots but 4 of them are scoring chances. If you extrapolated the data, you would infer that Kulemin takes better shots than Kessel, but over the long run, they’ll probably even out to a cleaner percentage. Kessel also scores on a lot of shots that wouldn’t be considered scoring chances by the strict definition published at the Copper and Blue as do a lot of other players. I’d say about 25% of goals scored by both the Maple Leafs and their opponents this past season would come on plays that I didn’t mark down as a scoring chance. While some people love talking about “quality” shots when other people talk about Corsi statistics, you’d be surprised at just how many times a goal is scored not off an odd-man rush, a puck in the slot, or a rebound.

Don’t get fooled by noise, and don’t make conclusions about the Leafs’ appearing to have a higher rate of shots turning into scoring chances than their opponents last season. If you look at the 36-game update, you’ll find that a lot of players that perform well in scoring chances also perform well in Corsi, the table at the bottom of the post. Since there are much, much more Corsi events than scoring chances in a given game, I much prefer looking at those. Scoring chances are rarer, and I’m not fully convinced that there’s any good practical application of the data at this point.

  • In the long run, the team that controls the puck will have more shots. The team that has more shots will have more scoring chances The Leafs coaching staff did a lot of good things this season but their thought process on shots and scoring chances as revealed in the Cronin interview is flawed.

  • Badger M

    Glad you’re posting again Cam.

    I read the Eric T post, and his post in the comments was interesting:

    “But I’m looking at Daniel Sedin being on the ice for 57.5% of the shots and only 54.8% of the scoring chances while Keith Ballard was on the ice for 49.4% of the shots and 53.1% of the scoring chances; I’m looking at Jarome Iginla being on the ice for 49.4% of the shots and 48.4% of the scoring chances while Nigel Dawes was on the ice for 53.2% of the shots and 56.6% of the scoring chances. And I’m thinking that while there is in fact a statistically significant difference from player to player, it is small enough at the 1-2 year level that the thousands of hours being put into collecting scoring chance data are resulting in negligible corrections to the freely available shot data”

    I’d say a 3% difference between Fenwick and Scoring Chance % is fairly significant. I’m still a bit confused as to whether scoring chances are a strong indicator of goals scored or not, but if they are then there could be significant differences in how teams are evaluated.

    Fenwick currently says that a team like Ottawa (52.7 FF%) is much better than a team like Anaheim (48.5 FF%). However, what if Ottawa had a Scoring Chance % of 49.7% last year and Anaheim’s was 51.5%?

    • MaxPower417

      I think where the “negligible” difference comes in, is A) the sample size of those findings and B) the room for error in having different people using slightly different subjective criteria to count these chances.

      • Badger M

        I agree the sample size is too small, but that’s an indication that MORE scoring chance data should be collected in order to see if the correlation holds up.

    • You’ll see greater discrepancies for players than teams, but generally they aren’t repeatable.

      It all comes down to sample sizes. While Nigel Dawes may be on the ice for more scoring chances than shots… well, he’s only playing a small percentage of the team’s overall minutes. You could get +3% for just Dawes, but in the end, that means a 49% team jumps up to a 49.5% team, and those margins are realistically just too small to be worthwhile in the long run, and there’s no proof that Nigel Dawes’ “talent” is repeatable.

  • MaxPower417

    #Urgent…Does Anyone know whether or not the Leafs understand we probably have as bad a center combination of players than any team in the league and will get creamed again if we do not obtain, whether in a trade, free agency or draft.

  • MaxPower417

    Do you track quality scoring chances like the leafs do or just regular scoring chances? Else it seems to be a comparison of apples and oranges and I’m not buying that sort of juice.

    • MaxPower417

      How are “quality” scoring chances defined? I assume more narrowly than standard scoring chances.

      They would have a ludicrously small sample size in addition to adding error due to subjectivity.

      I’m not saying that you’re wrong; the data Cam collected may have different criteria that the data the Leafs collect, and it might be an apples to oranges comparison. However, there’s really no point in evaluating (never mind making organizational decisions) based on quality scoring chance data.

  • jasken

    This is rather a good read

    You got a nice collection of data I guess this season, and betting scoring chances all come in the form of quality is anything inside home plate or house everything outside is not. When I read the responses and see well there is not enough data to make a valid inform decision. (hilarious)

    How long has getting players in front of goalie and causing screens, and trying to distract goalies been part of hockey. What do you mean not enough data are you kidding history is full of it.

    There is no exact definition of a scoring chance, its actually pretty open different teams and coaches have different opinions and views.

    This was basic form, it’s evolved now a little to suit me but this would be a good starting point for monitoring scoring chances and how they are registered. I will find out next season how it matches up to Carlyle’s. I lazed around and just watched hockey after 10 games so didn’t really get a good read on it.

    Shots for line changes, or stoppage of play. always a quality of 1 for poor there is really no attempt to score just basically if you get lucky.

    Percentage shot location of ice in which where shots were from. 1-10 based on location and percentage of that location. Percentage decrease/increase as quality adjusts it.

    Quality shot would be the variables (screens, tip-in, goalie cutting angle, etc..)that would increase or decrease quality of percentage shot. These would be numbered 1-5 depending the chance of scoring. 1 poor chance 5 being high chance.

    So player shooting from blue line on a scoring attempt ps1 qs2 sv as it might have left wing partially screening it was saved.

    If you understand shot percentage, and how variables contribute you can do this. Adding positions, and numbers for better accuracy of the variables, knowledge of who was involved. You can evolve the basic form pretty much to what you want or need.

    What I came to understand shots in game never reduced the quality shot did. You could get 22-25 good quality shots a night. In today’s NHL you be really lucky to hit 20.

    That’s enough noise for now I think I bored enough of you.

    • There are tonnes of ways of looking at it. While point shots aren’t generally considered, I would sometimes count them if it was a hard shot and a well-placed screen and the point man was pretty close to the general “chance” area. There was a lot of subjectivity. Two people can count the same game and one person will get 15-14 and the other will get 11-10. Over a season the ratios will be pretty similar.

      But ultimately, if a team is doing better at generating scoring chances compared to shots, no matter how they do it, they will have higher shooting percentages from year-to-year. But teams generally don’t. The teams that have the highest shooting percentages are the teams that have the best players, the ones that can consistently shoot at higher rates than their peers. Pittsburgh has Sidney Crosby, Evgeni Malkin and James Neal. Crosby’s one of the few players that can raise the shooting ability of his linemates, and there you have Chris Kunitz and Pascal Dupuis. You don’t often see teams with minimal skill maintaining high shooting percentages. You actually don’t. It doesn’t happen.

      Maybe the Leafs have taken a higher rate of scoring chances than their shot numbers. Actually, that’s probably true over the last couple of seasons. But the question is will they maintain it. History suggests they won’t, because teams don’t.

  • “Oddly enough, the narrative that Carlyle and Co. don’t have a clue what they are doing come from the lowest denominator of an internet arm-chair general manager from that other place.”