19 responses to “ADVANCED BRACTON SCORES FOR 2013-2014”

1. 

   Don B.

   July 29, 2015

   I won’t even pretend I know what an Advanced Bracon is. Or whether you eat it for breakfast or it tastes better with an ice cold beer. But, perhaps you could explain why Boston went from an AB of 56 in 2013, to an AB of minus 4 ish in 2014, falling to a predicted minus 6 ish in 2015-2016.

   LikeLike

   Reply
   1. 

      freeforall201

      July 29, 2015

      Thanks for the note – I responded to your other comment. The big thing is having consistently high AB players stay healthy (like Krejci) and flashes in the pan (like Paille) not kill the team too much.

      LikeLike

      Reply
2. 

   Don B.

   July 29, 2015

   Second Question: Was there anything predictive in the AB model at the start of the 2014-2015 season that suggested the Bruins fall from grace was imminent? (I’m a Jets fan but the radical changes in Boston’s ratings are interesting.)

   LikeLike

   Reply
3. 

   ehpem

   July 29, 2015

   “we believe strongly that the AB was highly correlated” – believe? is this religion or statistics?

   How about verifying this statement with some statistical analysis?
   I don’t know much about stats, but if I recall correctly, something like Spearman’s Rho would be suitable for comparing your “Finish” and “AB Rank” columns, If you get strong correlations for several years of data, then you may be onto something.

   Even then, I suspect you aren’t going to get much interest in your AB stat if you keep the details secret. “Trust me, I know the recipe for the magic sauce” doesn’t buy confidence.

   LikeLike

   Reply
4. 

   ehpem

   July 29, 2015

   “we believe strongly that the AB was highly correlated” – believe? is this religion or statistics?

   How about verifying this statement with some statistical analysis?
   I don’t know much about stats, but if I recall correctly, something like Spearman’s Rho would be suitable for comparing your “Finish” and “AB Rank” columns, If you get strong correlations for several years of data, then you may be onto something.

   Even then, I suspect you aren’t going to get much interest in your AB stat if you keep the details secret. “Trust me, I know the recipe for the magic sauce” doesn’t buy confidence.

   LikeLike

   Reply
   1. 

      freeforall201

      July 29, 2015

      I agree in part. If I give it away and get nothing back, then I might gain credibility but lose my intellectual edge (if any). If I keep it and get nothing back, I retain the formula and am no worse off. If I keep it and get something, then I am way ahead of the game. Still debating what to do because I don’t see releasing it and getting something back as an outcome.

      LikeLike

      Reply
      1. 

         ehpem

         July 29, 2015

         A conundrum! Perhaps letting it roll for a season after you have made your “prediction” and I suppose updated it as the trades and injuries change the picture and then consider being more public?

         Also, what about my suggestion for some quantification of past correlations?

         Are your AB scores derived from past years based on rosters at the beginning or end of a season?

         LikeLike

      2. 

         freeforall201

         July 29, 2015

         It was a retrospective study over two years using end of season roster. I will make projections in September when the final rosters are out and player movement is largely over. I just want this idea out there, to be widely known that it is my original content (for better or worse), get feedback like yours, gauge the level of interest and continue to generate interesting articles is my present aim. I am new at the internet stuff, and am testing the waters slowly in different markets as transactions happen, I am shocked and humbled at the level of interest I have received on this site to date. The stats are exceptional for only doing this for 6 weeks. Thanks for reading and caring enough to post a comment.

         LikeLike

      3. 

         ehpem

         July 29, 2015

         You’re welcome!

         It would be interesting to see how things looked at the beginning of past seasons vs. final standings and final AB scores too. That way we can better judge the utility of an October forecast.

         Come to that what would a midseason correlation/forecast look like? Loads of work I expect but it might give a better idea of the optimal timing for a forecast with this stat. End of season correlations are interesting but ultimately of little value if they end up without sufficient predictive oomph before the final result is known.

         Good luck with it all!

         LikeLiked by 1 person

      4. 

         freeforall201

         July 29, 2015

         Duly noted – I can try that around the winter classic – maybe…..its a lot of work

         LikeLiked by 1 person

      5. 

         freeforall201

         July 29, 2015

         Could you please check my math but Spearmans rho suggests that the 2013 data has a rho value of 0.926581. That’s pretty high. Someone else did the 2015 data for r squared and came up with 0.77 FYI. I’m not a statistician but this looks interesting

         LikeLike

      6. 

         ehpem

         July 29, 2015

         Spearman’s rho using the numbers in your other post with 2015 data comes out to 0.8759.

         LikeLike

      7. 

         ehpem

         July 29, 2015

         I feel like a bit of a fraud doing this in public given my totally rusty knowledge, which was never much good in the first place. Maybe you have a reader that can come to the rescue with the right stats packages and so on:

         The Spearman’s Rho I come up with is 0.9119, also a high number. I used Excel with the CORREL function as described here: http://bit.ly/1MylTQG. I have not taken it further to see how significant that result is, though the eye test suggests a strong correlation between your AB Rank for the roster at the close of the year and the Final Standing.

         LikeLike

      8. 

         ehpem

         July 29, 2015

         Just to be sure my math was consistent (not necessarily correct, but consistent) I reran the calculation using the longer method, as found here: https://youtu.be/vOI5_LWRxrA. Same result – 0.9119

         LikeLike

      9. 

         ehpem

         July 29, 2015

         btw, I am using the 2013-2014 data from this post, not some other year.

         LikeLike

5. 

   Don B.

   July 29, 2015

   You deleted my questions?????

   LikeLike

   Reply
   1. 

      freeforall201

      July 29, 2015

      Didn’t see them sorry. Boston fell apart last year because of two things; the net power play margin went from 5.4 to 3.4 goals for 5×4 versus 5×5 net and their 5×5 play was atrocious compared to the year before. This was basically due to Krejci playing half the season and losing Iginla without a suitable replacement, That was 15 goals at the margin right there. Even though they had a similar number of shots, they scored 45 fewer goals, as the AB model would suggest without adjusting for the rest of the team (Paille produced +10 AB in 2013 and was -3 last year, so theres another 13). With a healthy Krejci they should be OK, but they still don’t have an Iginla player to pick up the slack. Look for the Bs to be around breakeven AB this year and finish 7th or 8th.

      LikeLike

      Reply
6. 

   Don B.

   July 29, 2015

   How often do you adjust your rankings? While Boston’s fall may not have been predictable at the beginning of the season, how many games would have to be played before somebody screamed “Fire!”

   LikeLike

   Reply
   1. 

      freeforall201

      July 29, 2015

      This is an excellent question. Since I did this retrospectively, it is tough to answer. When I adjust the model to predict 2016 based on final depth charts, we will take it for a test drive.

      LikeLike

      Reply