Time to Geek-out! A deep dive on BRHL draft probabilities - are we better than NHL GMs???
This is a full on geek-out on draft probabilities in the BRHL. I've done more data analysis on our draft history, which have revealed some very interesting little tidbits.
1. Draft Strengths By Year
We covered this a bit in the spring already, but we might as well take another quick look, with a few different perspectives. Below you see a graph showing the percentage of picks from each year 2007-2016 that have crossed the 99GP and 199GP thresholds. More recent drafts tail off significantly due to having fewer years to reach the game played thresholds. Even 2014 - 2016 I think are being impacted by this effect. Most notable are the 2008 draft, which absoloutely blows, and the 2013 draft which is looking very strong.
Next up, we can pefrom the same type of review, but rather than looking at the % of picks reaching the GP thresholds, lets take a look at Point production, Wins and GP on a per year basis. The below figure considers each player drafted, and looks at their Points, Wins and GP per Season. Doing this on a per season basis will help mitigate the downward trend in the above figure for more recent drafts.
Again, we see 2008 is a weak draft, but in this one we see 2010 and 2012 look the best. Maybe for 2010 and 2012 there are not as many players crossing the GP thresholds that we see in the top figure, but on aggregate for the entire draft class, we see a pretty strong performance from GP, Pts and Wins per Season for each of these two.
2. Draft Strengths By Pick
Here we look at the player performances based on where they were picked in the draft. Again, the top figure shows percentages of players crossing the Games Played thresholds at each draft slot. Based on what we saw in the first figure, I am only including 2007-2013 here. OK, I will be the first to admit, that is an ugly plot. We can see some general trends with close to 100% succes at the top of the draft, but it certainly gets muddy once we get past the 2nd round or so. Some anomalies at the end, as we only had one draft between 2007-2013 that had picks that lat, so they get hugely effected by the performance of a single player.
Next we will again look at Points, Wins and GP per Season at each pick level. For this one I will include drafts right up to 2016. For GP and Pts, we see the major drop-off after the early part of the draft, with a levelling out in the later rounds. I have also included an exponential decay model to give us a trendline for these. The Wins figure is actually super interesting, in that it demonstrates that there is essentially zero relationship between draft position and Wins for goalies. Very interesting. There are generally more "spikes" in the early part of the draft than the later part, so it does indicate some better performance there. Now that I look at this, I am thinking I may need to dive a bit deeper on the goalies here.
3. Draft Strengths By Round
Ok, this is quite possibly the most interesting part of the analysis. The first figure shows again, percentage of drafted players who exceed the GP thresholds, but now we are looking it them by round. I've only included 2007-2013 again here. I have also included the NHL percentages for >99GP, which you can compare against the green bar to see if we are actually better than the NHL GMs (with our 1 extra year of evaluation).
Finally, the anaylsis looking at Pts, Wins and GP per season on average for players drafted in each round. Again, we see the expected large drop off, followed by a levelling out. Interesting to see that on average a 1st rounder is going to produce about 20pts per season, and play about 40games per year.
I'd like to highlight one thing that you need to keep in mind when viewing the Points/Wins data. Take for example the graph directly above. For the average Points produced by a 1st rounder, the calcuation tallies the total points produced by all first rounders, and divides by the total number of first rounders. This is a bit unfair, considering that many of those first rounders will be goalies, who shouldn't be expected to produce points. This is seen more extremely with Wins - again for example, we take the total wins by all 1st rounders and divide by the total number of first round picks to come up with the data. This is again unfair, as many of those first round picks are skater who won't be expected to earn Wins.
It would be a great help to this analysis if we could tag every pick as being a Goalie or Skater, so I could do the calculations more accurately, but alas, that is a lot of manual work, especially if we go all the way back to 2007. So take these values with a grain of salt.
That said, I think the GP values provide a very good indication of overall performance, considering both Skaters and Goalies will "Play Games".