Quote:
Originally Posted by overpass
5 vs 5
I used behindthenet.ca's .xls files, with data from the 200708 season through the 201011 season.
Using the binomial approximation to a normal distribution, I calculated the zscore for each NHL player's onice shooting percentage over this time period, where zscore = (Onice GF  OnIce SOGF*LgSH%)/(Standard deviation calculated using binomial approximation)
Here are the players with the highest and lowest zscores
Player  Zscore  5on5 onice SH%  SIDNEYCROSBY  6.12  11.7%  HENRIKSEDIN  5.49  11.1%  DANIELSEDIN  4.98  10.9%  MARIANGABORIK  4.93  11.4%  BOBBYRYAN  4.91  11.3%  EVGENIMALKIN  4.88  11.0%  RYANGETZLAF  4.75  10.8%  ILYAKOVALCHUK  4.70  10.8%  COREYPERRY  4.37  10.6%  DANYHEATLEY  4.33  10.5%  JASONSPEZZA  4.31  10.7%  ALEXTANGUAY  4.15  10.8%  NATHANHORTON  4.10  10.5%  ALEXBURROWS  4.08  10.6%  JAROMEIGINLA  4.07  10.3%  RYANWHITNEY  4.02  10.8%  PAULSTASTNY  3.95  10.6%  MARTINST. LOUIS  3.93  10.3%  MIKERIBEIRO  3.90  10.5%  NICKLASBACKSTROM  3.88  10.2%  ALEXANDERSEMIN  3.78  10.5%  PAVELDATSYUK  3.68  10.2%  DARRYLBOYCE  3.62  14.9%  ALEXOVECHKIN  3.59  10.8%  JEFFSCHULTZ  3.59  10.3%  STEVENSTAMKOS  3.58  10.6%  NICLASHAVELID  3.37  11.0%  ALEXANDEROVECHKIN  3.32  10.5%  BRENDENMORROW  3.32  10.4%  JOETHORNTON  3.27  10.0% 
Player  Zscore  5on5 onice SH%  TRAVISMOEN  3.90  5.6%  SHAWNTHORNTON  3.73  5.5%  COLTONORR  3.71  4.5%  RADEKMARTINEK  3.62  5.9%  ADAMHALL  3.60  5.0%  BRENDANWITT  3.59  5.5%  DEREKMEECH  3.52  4.5%  CRAIGADAMS  3.49  5.4%  RYANHOLLWEG  3.46  2.6%  DONALDBRASHEAR  3.42  4.4%  STEPHANEVEILLEUX  3.33  5.6%  JEFFTAMBELLINI  3.29  5.6%  ANDYGREENE  3.27  6.4%  NATETHOMPSON  3.22  5.5%  RODPELLEY  3.18  4.6%  TODDMARCHANT  3.12  5.9%  SAMUELPAHLSSON  3.08  6.1%  RAITISIVANANS  3.05  4.3%  FREDRIKSJOSTROM  3.04  5.9%  TIMSTAPLETON  3.04  3.4%  DANIELWINNIK  2.97  6.3%  SCOTTGOMEZ  2.95  6.8%  ANTTIPIHLSTROM  2.92  3.8%  TOMWANDELL  2.92  5.2%  ANDREWMURRAY  2.91  5.5%  THOMASPOCK  2.77  4.1%  JAMIEMCGINN  2.76  5.4%  ANDREWPETERS  2.75  2.4%  MIKEWEAVER  2.73  6.6%  DALLASDRAKE  2.69  3.6% 
Generally speaking we see skilled players outperforming the league average and "less skilled" NHL players underperforming the league average in onice shooting percentage.
Taking the standard deviation of the zscores, I get 1.41. If random variation was the only factor, the standard deviation of the zscores would be 1.00.
Following the process from this blog post:
Step 7: 838
Step 8: 0.49
Step 9: 860
So, given a player's onice shooting percentage from 200708 to 201011, the best predictor of his 201112 onice shooting percentage is his 200708 through 201011 shooting percentage plus 860 shots at a league average shooting percentage. Because Sidney Crosby was on the ice for 1867 shots at a 11.7 SH%, his predicted 201112% would be 10.7%.
Best and worst predicted 201112 5 vs 5 onice shooting percentages
Player  Predicted 5vs5 onice SH%  SIDNEYCROSBY  10.7%  HENRIKSEDIN  10.4%  MARIANGABORIK  10.3%  BOBBYRYAN  10.3%  EVGENIMALKIN  10.2%  DANIELSEDIN  10.2%  RYANGETZLAF  10.1%  ILYAKOVALCHUK  10.1%  JASONSPEZZA  10.0%  ALEXTANGUAY  10.0% 
Player  Predicted 5vs5 onice SH%  TRAVISMOEN  6.6%  COLTONORR  6.6%  SHAWNTHORNTON  6.6%  ADAMHALL  6.6%  BRENDANWITT  6.7%  DEREKMEECH  6.7%  CRAIGADAMS  6.7%  RADEKMARTINEK  6.7%  DONALDBRASHEAR  6.7%  STEPHANEVEILLEUX  6.8% 

Nice work.
I can't remember if the data at behindthenet includes empty net goals.
I know the data is ostensibly 5 v 5 only, but 6 v 5 situations may have been included.
What is the aggregate onice shooting percentage in your sample?
If it's not almost exactly 0.08, then empty net goals haven't been removed.