News Article: "Fun With Numbers" - Advanced Stats Talk Here

Discussion in 'Ottawa Senators' started by BonkTastic, Aug 5, 2014.

  1. Micklebot

    Micklebot Moderator

    Joined:
    Apr 27, 2010
    Messages:
    30,315
    Likes Received:
    7,477
    Trophy Points:
    157
    This quote is spot on, you shouldn't take a lack posting names of members of an analytics department on the Sens website at face value, particularly given how notoriously slow that site has been in updating information in the past.

    That said, I know Tim Pattyson is involved in the analytics side as well. Cookson is in charge of it. Not sure about anything else, but I imagine based on the way both Boucher and Crawford have talked about analytics in the past, they will want the info available, and will likely ensure that the means for them to get the info they want is there.
     
  2. DrEasy

    DrEasy Out rumptackling

    Joined:
    Oct 3, 2010
    Messages:
    6,487
    Likes Received:
    1,392
    Trophy Points:
    109
    It's funny how no matter the information (or lack thereof), the optimist will decide to interpret it one way, and the pessimist the other. Humans... :shakehead
     
  3. StefanW

    StefanW Registered User

    Joined:
    Mar 13, 2013
    Messages:
    6,286
    Likes Received:
    0
    Trophy Points:
    54
    Occupation:
    Research
    Location:
    Ottawa
    Home Page:
    The Senators do have an analytics person, and Dorion has made reference to him in interviews for about a year now. As is the case with every other team the exact types of things the person is asked to look at is proprietary info and kept secret.

    While the Sens have beefed up their analytics department over summer with Cookson and and Pattyson, there is valid reason to question what this means in terms of the team's use of analytics. For example although video analysis is really handy, and can provide insight for coaches to use, this is not what is typically meant by "analytics" unless it is done on a very large scale (such as what analytics firms like SportLogiQ do).

    From the outside it looks as though Ottawa is doing what a small number of teams such as Edmonton are doing, which is to emphasize old school (circa 1980s) approaches to collecting info such as coaches pouring through video analysis. There is nothing wrong with that per se, but it is not likely to appease fans who are looking for the team to hop onto the analytics bandwagon in a larger way. It is even less likely to appease fans who believe the team will be left behind as other teams are snapping up quality analysts as they search for a competitive edge.

    IMO all of this has nothing to do with optimism or pessimism. Either analytics make teams better or they dont, and it should come through in the results either way. Either teams make decisions to collect info in ways that lead to improvement or they dont, and that should also come through in the results. And, either teams use the info they have to good effect (i.e. knowing how to weight info so that low value things are not assigned high value during decision making) or they dont, and that should come through in the results. These standards never change.
     
  4. DrEasy

    DrEasy Out rumptackling

    Joined:
    Oct 3, 2010
    Messages:
    6,487
    Likes Received:
    1,392
    Trophy Points:
    109
    I agree with this, and actually that's why I'm very pro-analytics because it takes away subjective matters out. (it's a different debate as to whether hockey analytics are at a point where there's enough nuggets to be found yet in the very noisy data they have to deal with).

    I was only referring to the fact that people who are down on the org look at the evidence we have regarding the existence of an analytics team and conclude that we don't really have one, and the people who are pro-org will look at the same evidence and draw the opposite conclusion. It's very predictable. That's what happens when you have a polarized fanbase/electorate/etc. I'd love to be surprised more often by people's opinions. But I digress.
     
    NyQuil likes this.
  5. StefanW

    StefanW Registered User

    Joined:
    Mar 13, 2013
    Messages:
    6,286
    Likes Received:
    0
    Trophy Points:
    54
    Occupation:
    Research
    Location:
    Ottawa
    Home Page:

    For sure. IMO there are overlapping issues. First off, using public analytics is not going to create much of an edge because everyone has access to the exact same information. Proprietary data can give you the most bang for your buck in terms of gaining a real edge on the competition.

    This leads to the second point, which is that teams have to be able to identify which information will lead to the best results. A typical approach you see in industry is to hire people to run particular numbers that are of interest. With hockey that would mean hiring an analyst to run numbers on things the team already believes is there. So if a team believes a cross ice pass leading to a one timer leads to goals they may ask an analyst to measure cross ice passes in the O zone. This can be problematic if the things the analyst is asked to measure are meaningless or have low value.

    That brings up the third point, which is the skill of the analyst. If a team leaves room for an analyst to run the numbers he or she wants then it becomes a matter of ferreting out high value information that no one else has. Some analysts are more highly skilled than others and are much more likely to be able to use their imagination to come up with new ways of approaching problems. This is where a lot of online analysts fail, because they start to over-rely on data that is already available rather than figuring out what they would do if they had full access to the team and its resources. The worst over-reliance occurs when the data that is publicly available frames the question rather than asking questions and then figuring out how to get the data to answer it.

    Ok, I've rambled too much. I could go on and on with this stuff.
     
    Last edited: Oct 20, 2016
  6. Engineer

    Engineer Rustled your jimmies

    Joined:
    Dec 23, 2013
    Messages:
    6,143
    Likes Received:
    1,791
    Trophy Points:
    126
    Ever notice that the GAR and WAR stats that people love to throw around never display error bars.

    Or why is it that a full article that's dedicated to explaining why WAR and GAR stats are important, doesn't have a single use of either of the words "variability" or "standard deviation"

    How can one actually analyze data without knowing how much it varies?
     
  7. BonkTastic

    BonkTastic ಠ_ಠ

    Joined:
    Nov 9, 2010
    Messages:
    30,810
    Likes Received:
    9,458
    Trophy Points:
    156
    Location:
    Parts Unknown
  8. Sens of Anarchy

    Sens of Anarchy Registered User

    Joined:
    Jul 9, 2013
    Messages:
    36,325
    Likes Received:
    15,774
    Trophy Points:
    156
    Well no actually. :) I am just beginning to trying to understand the WAR stats. I've read a couple of articles on them and find it interesting.

    With any stats context is important. They provide an answer to a specific question but that is only a slice of the pie. The data quality / consistency / variability is obviously important to this type of analysis.

    Are you a WAR / GAR guru that could help me understand it better? or point me to something that explains the metrics and factors and calculations used to derive the numbers.
     
  9. Sens of Anarchy

    Sens of Anarchy Registered User

    Joined:
    Jul 9, 2013
    Messages:
    36,325
    Likes Received:
    15,774
    Trophy Points:
    156
    I am with Bob on this but I don't see it changing. In fact I think it will get worse as more astats are published and more sights allow the slicing and dicing to drive conclusions and provide the basis for argument. It is unfortunate... It is hard to model hockey players with a plethora of stats or enter into meaningful discussions where there is a common ground understanding of what they mean what is in context and what is not.
     
  10. BondraTime

    BondraTime Registered User

    Joined:
    Nov 20, 2005
    Messages:
    18,075
    Likes Received:
    7,191
    Trophy Points:
    186
    Location:
    East Coast
    Fully agreed with Bob. I'm open to "advanced stats", but I would never, ever, ever rely on them.
     
  11. DrEasy

    DrEasy Out rumptackling

    Joined:
    Oct 3, 2010
    Messages:
    6,487
    Likes Received:
    1,392
    Trophy Points:
    109
    I don't think anybody's saying you should manage a team just by looking at numbers. Stats are just there to help you make a decision. When you see some number that doesn't match the eye test, you want to find out why. Is that a problem with the stat, is the sample too small, or is there really something worth investigating?

    Teams are all looking to gain an edge. Sure, if you can have the best scouts out there, good for you (hmm, how would you know they're the best scouts? you probably look at some set of numbers?), but if you can't you need to find another way.

    We're all already using numbers all the time. The whole point of advanced analytics is to go beyond the basic numbers, because we know they're not good enough: W/L, a goalies GAA. We know that run differential is a better predictor of future performance than W/L. We want to find even better predictors if possible. What's wrong with that?

    More generally you should always strive to have a way to compare things quantitatively. The trick is to find the right metrics for that. So, sure, some people may jump on the latest fancy stat as if it's a miracle recipe, and then we find out that there's more going on. Well, that's how science progresses. If all you do is just rely on eye test you get stuck.

    Baseball has made huge progress thanks to fancy stats, but I feel there's way more variability in numbers in hockey (hence the need for error bars and standard deviations mentioned earlier by someone), and so it's going to be more difficult to find meaningful metrics. That doesn't mean we should give up.
     
  12. Fenix Rises 2026

    Fenix Rises 2026 Alfie forever!

    Joined:
    May 31, 2008
    Messages:
    7,519
    Likes Received:
    645
    Trophy Points:
    139
    Here's the rest of the team's top 10 closest comparable players. Probably have to open in a new tab as it's a huge image.

    [​IMG]

    You can read a bit more about the work here: https://datafuntime.wordpress.com/

    I'm going to pump out graphs for the other teams, but I'm still trying to tweak the way I build them a bit. Most of the comparisons are what you'd expect, but I don't like Zib and Ryan being comparables. Any help is appreciated. Specifically, is there a close comparable player that you would have expected to see that isn't on the list? If so, I can dig into the data and see what's going on.

    Also, I love how Boro is such an anomaly. :laugh:
     
    Last edited: Oct 21, 2016
  13. Sens of Anarchy

    Sens of Anarchy Registered User

    Joined:
    Jul 9, 2013
    Messages:
    36,325
    Likes Received:
    15,774
    Trophy Points:
    156
    Hi is there a sweat spot to the comparables.. on the virticle axis... IE closer to the center means close match , above means > than and below < than?
     
  14. Fenix Rises 2026

    Fenix Rises 2026 Alfie forever!

    Joined:
    May 31, 2008
    Messages:
    7,519
    Likes Received:
    645
    Trophy Points:
    139
    No, I probably should make it easier to read. :laugh: Closest to the bottom is the #1 closest comparable, closest to the top is #10 closest comparable. There's supposed to be a line and a circle on the player name at the bottom, but it seems to have got lost somewhere. EDIT: updated the image. Hopefully it's less confusing now. I'm going for more infographic and less scatterplot.
     
    Last edited: Oct 21, 2016
  15. DrEasy

    DrEasy Out rumptackling

    Joined:
    Oct 3, 2010
    Messages:
    6,487
    Likes Received:
    1,392
    Trophy Points:
    109
    The graphics look really nice. Hard to judge the results without knowing the metrics you used to compare them. For example, it is interesting that the data did not include the position they were playing, and yet it shows up. But if you have used data about where their shots came from, maybe things aren't so surprising.
     
  16. thinkwild

    thinkwild Veni Vidi Toga

    Joined:
    Jul 29, 2003
    Messages:
    9,223
    Likes Received:
    325
    Trophy Points:
    169
    Location:
    Ottawa
    Lots of cool new data coming out for hockey these days, keep it up. Something hockey fans yearned for for a long time. Data is never wrong, its just data. Our ability to interpret it however seems an art for that is still in progress. Which really just adds to the the fun.
     
    Do Make Say Think likes this.
  17. BigRig4

    BigRig4 Registered User

    Joined:
    Feb 22, 2014
    Messages:
    2,551
    Likes Received:
    584
    Trophy Points:
    109
    4th in shots for per game and 14th in shots against per game :handclap: surely this means our success is probably sustainable right??
     
  18. 18Hossa

    18Hossa And Grace, Too

    Joined:
    Oct 12, 2012
    Messages:
    6,582
    Likes Received:
    225
    Trophy Points:
    82
  19. DrEasy

    DrEasy Out rumptackling

    Joined:
    Oct 3, 2010
    Messages:
    6,487
    Likes Received:
    1,392
    Trophy Points:
    109
    Good stuff, thanks for the pointer and thanks to StefanW for writing the post! Wish I'd known about the event. Well, there's still the hockey analytics conference at Carleton to look forward too. :yo:
     
  20. StefanW

    StefanW Registered User

    Joined:
    Mar 13, 2013
    Messages:
    6,286
    Likes Received:
    0
    Trophy Points:
    54
    Occupation:
    Research
    Location:
    Ottawa
    Home Page:
    Glad you enjoyed the summary! I hope the Sens put on more events like this down the road.

    Along with OTTHAC in the spring there is another one in Vancouver this weekend. I dont think any announcement has been made about streaming it but it will probably be worth checking that day to see if any streams are available.
     
  21. aragorn

    aragorn #Boucherout

    Joined:
    Aug 8, 2004
    Messages:
    19,819
    Likes Received:
    1,896
    Trophy Points:
    231
    Finally, something you & I can agree on. :handclap:
     
  22. JungleBeat

    JungleBeat Abracadabramov

    Joined:
    Sep 10, 2016
    Messages:
    2,601
    Likes Received:
    1,296
    Trophy Points:
    94
    Location:
    Calgary, Alberta
  23. solidprospect

    solidprospect Borveetzky

    Joined:
    Sep 30, 2017
    Messages:
    4,422
    Likes Received:
    1,258
    Trophy Points:
    85




    Summary:

    -Novak is ranked extremely low in hf prospect rankings.
    -We currently have 5 forward rookies with a greater than %80 chance of making the NHL
    -these stats can change as the season goes along, could add Batherson and others to that group of forwards by next year or they could drop
    -Wolanin and Jaros are likely NHL defenceman eventually
    -Drake Batherson is very good right now compared to our other top rookies his war/82 projection is only lower than Logan Brown's
    -Tychonick and Bernard Docker should also be very good.
    -it's interesting that Batherson has the 2nd highest war/82 projection for the rookies but in the other chart he's only at %60 chance of making the NHL right now. Whereas the others have a greater chance of making it but a lower war/82 projection.
    -Some of these players should still get time in the AHL, this just indicates the chances of making the NHL and their projected production if they we're in the NHL next season.
     
    Last edited: Sep 12, 2018
  24. solidprospect

    solidprospect Borveetzky

    Joined:
    Sep 30, 2017
    Messages:
    4,422
    Likes Received:
    1,258
    Trophy Points:
    85
  25. solidprospect

    solidprospect Borveetzky

    Joined:
    Sep 30, 2017
    Messages:
    4,422
    Likes Received:
    1,258
    Trophy Points:
    85
    [​IMG]

    We're fun.
     
    Sensators likes this.

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice
monitoring_string = "358c248ada348a047a4b9bb27a146148"