Last edited by temek; October 19, 2012 at 7:02 AM.
You know what's something that I find very interesting...
Most of the player's ADP is fairly close to the original Yahoo! rankings. I'm just wondering how much of a role that actually plays.
I wonder if you slotted someone fairly low say Stewart (176) and slotted him up say 100, if his ADP would naturally be at or around the 100 mark, or would it slip back down to the 170s...
Dobber Expert Pool Champion 2011-12
Code:Pro Campbell GarthButchers 91,5 Last Resort 48,5 Fergie's Leafs -16 Das Chicleteur -35,3333333333 expectoraters -44,6666666667 Serious Chinny -49,5 blammo -72,6666666667 Zorro's Howitzers -79,5 Loki -216,1666666667 Steam Roller -252,6666666667 rattus rattus -308,3333333333 Smoothfun -379,1666666667Code:Experts Mac's Militia 225,7666666667 Prospect Bits-Bitely 174,1666666667 Between The Lines! 155,1 Cage Match 142,6666666667 BizSmack 119,6 Amato 101,5 Angus Unleashed 91,7333333333 Maaaasquito Bites 48,8 Commish Office -8,5666666667 Lord of the Rinks -74,8333333333 Lupul on my Clitsome -93,5 Capped -106,5666666667 Beast from the East -121,2333333333 Legion of Dobber -138,1 GMG's Market Buzz -190,6333333333 Mess4Life'sWeeklyPix -196,9
There were changes in Entry leagues as well, but nothing too significant.
Last edited by temek; October 19, 2012 at 8:36 AM.
It's a pretty neat exercise.
Then - when I saw my team listed 4th last in the Smythe, I knew something must be flawed.
(LOL, not being cocky - but I'm 100% certain, barring injuries & goalie flops, that I have a top-half team. And I'd bet that an FHG projection of teams would also peg my team in the top half).
There is indeed a considerable flaw here, IMO. This is because true fantasy-hockey multi-cat value is all about relative difference in those actual statistical values. You need to get more value out of your players than what you paid for them. However, actual player "value" is not tied linearly (i.e. unpowered) to their draft position (which is reflected here), so much as it is tied to their statistics (which is not reflected here).
One of my favourite lists on this site is Dobber's Top 300.
Let me list a few players for sake of discussion.
4. Steven Stamkos 237.5
5. Claude Giroux 200.9
201. Tyler Kennedy 40.1
300. Mason Raymond 23.3
OK, consider there are only two teams and in a couple spots they've deviated from their pegged draft pick.
1st Round, #4: Team D selects Claude Giroux (200.9pts)
1st Round, #5: Team E selects Steven Stamkos (237.5pts)
14th Round, #206: Team E selects Mason Raymond (23.3pts)
14th Round, #207: Team D selects Tyler Kennedy (40.1pts)
By the actual draft position math - ADP:
Team D gets -1pts for the Claude Giroux pick (4 - 5)
Team E gets +1pts for the Steven Stamkos pick (5 - 4)
Team E gets -94pts for the Mason Raymond pick (206 - 300)
Team D gets +6pts for the Tyler Kennedy pick (207 - 201)
Now, if the DobberTop300 values are accurate to H2H categories (say), then:
Team D has two players totalling a value = 241.0pts (ADP score of +5)
Team E has two players totalling a value of 260.6pts (ADP score of -93)
***By the ADP math, Team D has the considerably stronger two players.***
***By the statistical value math, Team E has the considerably stronger two players.***
As another example:
In the Smythe draft, Marek Zidlicky (ADP 259.0) fell all the way to pick #336 (drafted 226, 227, 249 in other leagues). So this gives Zidlicky's team a +77pts in the ADP math. But how much more REAL statistical value does Zidlicky REALLY have over defensemen that went between 226-336. The range goes from Dennis Seidenberg to Francois Beauchemin. I mean, can you really look at those two players and say there is a 100pt value difference between those guys??? Shouldn't there be more difference between Erik Karlsson and Alex Pietrangelo who have an ADP difference of a 30pt value (ADP 21 vs ADP 59)?
Overall, I think the math is very skewed.
Fantasy hockey statistical-value in the NHL falls in a Gaussian distribution with the elite players exhibiting the most relative difference in, say, a 20 spot range.
Evgeni Malkin (2) should be consistently better than Zach Parise (19), and we know this.
But is it a lock that Martin Erat (223) is going to be better than Sam Gagner (240)?
I'd say no and the actual value of those middling players are going to be all over the place every year, maybe 170 this year, 260 next year, 243 the following year... but yet the statistics don't have to vary much for them to move. The difference in a 50pt player vs a 40pt player could be 100spots, because the Gaussian distribution of NHL statistics is populated greatly in those middling (30-50pt) players.
This ADP system suggests that there is a 77pt value in drafting Zidlicky at #336 (vs. his ADP = 259.0).
But, let's say somebody was able to draft Henrik Lundqvist at #79 (vs. his ADP = 2.5). That's also 77pt ADP value.
Should these be consider similar 77pt values?
As a post-logue
I think comparing projected FHG stats for an entire team of players is really the best forecast anybody can make. (assuming they also have a fairly solid set of projections). And even with projections, I believe that goalie statistics can be most misleading. For example, metaldude's team is forecasted as the #1 team in the DobberExperts league.... but if you go look at the project goalie rankings in the ROTO league, he's at the top of several - and goalie statistics are the ones most likely to NOT fall closest to predictions, IMO. When you've got 8 teams between 2.30 GAA and 2.40 GAA, those are pretty close. In a shortened season, if you have a goalie projected for 2.20 GAA and he puts out a 2.40 GAA, that'll move a team's goalies 0.10 GAA, which could be the difference between finishing 2nd in GAA & finishing 8th or 9th.
I actually considered doing the relative-ADP math myself, since it's set up very easily in the Excel sheet I have... but I recognized the flaw in draft position difference vs. actual value position (w.r.t. Gaussian distribution and separation of statistics).
ps. Sending you some REP anyways - because I love numbers.
Last edited by Pengwin7; October 19, 2012 at 9:57 AM.
I really wish my post wasn't so long... because it often comes off as negative.
I really did love the list - even if I don't feel it is accurate.
A few months ago I actually created a scoring system that gave players values in each category (0-10 value, pretty much).
The Base10 number is their actual statistical score.
This list actually matches up very closely with Yahoo!s "rank" and so I feel it's quite accurate.
It is a good representation of how closely the actual players value get together as you start nearing the 40-50pt players (say the 100-150th best forwards).
I am quite skeptical myself about the value of these kind of rankings and I would classify them more as interesting than useful as well. It's not totally out of question they might have some predictive power though so decided to use these leagues as a test of their usefulness. Can't be certain how useful these are without testing them and could be interesting to look at after the season.
Last edited by temek; October 19, 2012 at 10:52 AM.
As a thank you (temek) for your work, I've attached my absolute favourite spreadsheet that I made for the DobberEntry league. This shows the top 180F by value-scores based on the 7 categories (we include both PIM & Hits). It's really good. Each player gets a score (pretty much 0-10) based on the AVERAGE statistical count for top top 180F (since 15 teams x 12F) in each category.
For top 180F, average statistics:
Goals: 20. (0G = 0pts, 40G = 10pts, 50G = 12.5pts)
Assists: 30. (0A = 0pts, 60A = 10pts, 66A = 11pts)
+/-: +5. (-25 = 0pts, +5 = 5pts, +35 = 10pts)
PIM: 55. (0PIM = 0pts, 110 PIM = 10pts, 220 PIM = 20pts)
PPP: 13. (0PPP = 0pts, 25PPP = 10pts, 30PPP = 12pts)
SOG: 180. (0SOG = 0pts, 360 SOG = 10pts)
Hits: 100. (0 Hits = 0pts, 200 Hits = 10pts, 300 Hits = 15pts!)
Some people might think this scoring off/skewed, but there is justification in the summations at the bottom of the columns.
There is a top90 set of players, with total scores, and AVG player score per cat.
There is a 2nd90 set of players, with total scores, and AVG player score per cat.
The AVG player score for the top90 + 2nd90 should come out close to 10.0 (full score).
And if each category has the same total AVG score, THEN we know that the categories are all weighted evenly, since a 1st place in GOALS is worth the same as a 1st place in PIM.
The most notable thing I found in this spreadsheet is the key category to focus on in the first half of the draft: PPP. The top half of forwards (top90) account for 66% of PPP. (7.0/[7.0+3.6]). It's the greatest skew of stats in a draft. Anybody with experience knows, it is hard-hard-hard to find PPP at the end of a draft. Other statistics can still be found at the end of the draft... and even on the waiver wire right now. I won't get into them for fear of making my DobberEntry opponents too strong... but I do know that Ryan Ma has one of the greatest understandings of this finding. (and in truth, part of this spreadsheet was an inquiry into Ma's draft strategies, which I finally witnessed first hand in the RHRS).
Anyways, this is my favourite spreadsheet I've ever made.
Damn all these people analyzing my draft strategy to a T... I'm not going be able to win any more leagues from now on...
Dobber Expert Pool Champion 2011-12