The Most Interesting Thing I've Written in Some Time

by Jonathan Willis on Dec 30, 2008 3:14 PM PST Comment 37 comments

The official AHL website isn’t as easy to use as the NHL’s, but it does have some really interesting data. For example, looking back through the game scores, it allows us to see who was on the ice for every goal for and against. Now, this is a crude measure, but it should tell us what kind of opposition different players in Springfield are facing.

What I’ve done below is split the forwards that each Falcon played against into three categories: Strong, Medium and Weak. A "strong" player is anyone managing .75 points per game or more (Falcons who would qualify: Potulny, Schremp and Brule). A "medium" opponent is any forward averaging between .5 PPG and .75 PPG (Falcons: Trukhno, Lerg), while a "weak" opponent is anyone averaging less than .5 points per game. I’m welcome to suggestions; the idea here is just to see how good each player’s opposition is offensively.

To measure how often each Falcon sees this kind of player, we simply count each player on the ice for an event (goal for or against at even-strength). For example, on Sunday, Springfield played L.A.’s farm team, the Manchester Monarchs. O’Marra, Paukovich, Lefebvre, Young and Bina were on the ice for the first goal against, while Boyle, Martinez, Dravecky, Westgarth and Bagnall were on for Manchester. Each Falcon would have an event scored like this (Strong/Medium/Weak): 1/0/2. This is because Boyle (11pts in 13 GP) is considered a strong offensive player, while Dravecky (13pts in 33GP) and Westgarth (4pts in 30 GP) are considered "weak".

Here’s the raw data:

Bekar: 10/17/18
Benson: 4/3/5
Brule: 20/11/25
Corazzini: 20/22/32
Goulet: 4/3/11
Huddy: 0/1/2
Lefebvre: 16/15/30
Lerg: 19/18/32
McDonald: 16/12/16
O’Marra: 7/12/28
Paukovich: 7/9/17
Potulny: 27/26/44
Reddox: 11/12/10
Schremp: 17/16/39
Sestito: 18/15/14
Spurgeon: 22/27/22
Trukhno: 23/23/34
Urquhart: 1/3/11

Bina: 14/10/18
Bisaillon: 15/13/36
Chorney: 27/29/47
Constant: 3/6/13
Gabinet: 1/2/3
Hrabal: 12/11/14
Kinley: 2/0/0
Peckham: 48/33/51
Roy: 21/29/34
Taylor: 5/9/13
Wild: 20/15/27
Young: 6/8/13

The sample sizes on a lot of these players are very small; obviously, this exercise tells us nothing about players like Kinley, Huddy or Gabinet, but even a player like Goulet could have generated 18 marks in only 6 goals for or against. In other words, these aren’t definitive rankings, and common sense should be applied freely. Still, it gives us a starting point for discussion.

Now, it would be nice to break down this chart from three numbers per player to one number per player. For this exercise, I’m going to value an event against a "strong" opponent as being worth 2 points, and against a medium opponent as being worth 1 point. Players receive no credit for being on the ice against a bottom-six AHL forward. That number is then divided by the total number of events to give us a ratio of strong players to weak players. Also, for simplicity’s sake, I’m going to cut out all players with 12 events (4 goals) or less. Here’s the new list, starting with the forwards:

Sestito: 1.085
Reddox: 1.030
Spurgeon: 1.000
McDonald: .957
Brule: .911
Trukhno: .863
Corazzini: .838
Potulny: .825
Bekar: .822
Lerg: .811
Lefebvre: .770
Paukovich: .697
Schremp: .694
Goulet: .611
O’Marra: .553
Urquhart: .333

There are a few surprises here, but most of the list makes sense. Reddox and Sestito have both been called up to play a defensive role; it makes a ton of sense that they were deployed against the opposition’s best while in Springfield. Going down the list, there’s a nice group of veterans and players with solid two-way reputations, and at the tail end are players who have either a largely offensive game (Schremp) or are fringe AHL’ers (everybody else). The bad news here is that Schremp doesn’t appear to be generating his offense against tough AHL players; his results have been impressive but they’ve been coming against the soft parade, so it’s hard to imagine him playing against anything but the same in the NHL. The last thing worth commenting on is that Ryan O’Marra probably shouldn’t be on the radar anymore – he’s -6 with marginal offense, and it isn’t like he’s playing the toughs.

Peckham: .977
Hrabel: .946
Bina: .905
Wild: .887
Roy: .845
Chorney: .806
Young: .741
Taylor: .704
Bisaillon: .641
Constant: .545

Theo Peckham’s position here is very encouraging. He’s been even (on a team nine games below .500) against the best the AHL has to offer, and added some offensive punch to boot, and those are very encouraging signs that he’ll be able to keep his head above water in third pairing NHL minutes next season. Josef Hrabal (even through 17 games) also looks much better by this metric. Of course, the big negative is that Taylor Chorney’s team-worst -21 looks even more miserable in context. Roy’s -14 and Bisaillon’s +2 are also worse by this metric.

Final note: if anyone has any suggestions for improving this metric, please chime in below.

0 recs | Comment 37 comments

Story-email Email Printer Print

Comments

Outstanding stuff. Excellent work. And I should add that this is the first time Colin McDonald has looked +.500 in any way (so I thank you for that). :-)

I think your math makes sense, not thrilled with zero value for a goal for (even against weak opponents) but don't know how to make it better (perhaps move everthing to the left, as in "goals for and against weak" +1 and then +2 and on it goes) so look forward to the next installment.

by Lowetide on Dec 30, 2008 6:25 PM PST reply actions 0 recs

LT: Yeah, I was about half-way through when it occured to me that goals for and goals against would be a logical division; next time I'll be splitting it that way as well. I'm also curious whether my (somewhat arbitrary) boundaries would be better if they were altered somewhat.

by Jonathan Willis on Dec 30, 2008 10:09 PM PST reply actions 0 recs

Excellent work, Jonathan. LT beat me to the suggestion, so I'll just second it.

Question: Are you back?? I just happened on this post by clicking on the live link at the bottom of LT's game-day thread, must have missed the explanatory comment amongst all the Big Ben/Parliament garbage.

by Bruce on Dec 30, 2008 11:42 PM PST reply actions 0 recs

Bruce: Not really. I'll be posting some of my more elaborate stuff that isn't getting comments at ON here; I know quite a few people still use the blogroll on this site, and I'm really just fishing for commentary.

by Jonathan Willis on Dec 31, 2008 5:48 AM PST reply actions 0 recs

Rumour of your demise ...

Great stuff, Jonathan.

Peckham looks terrific and not surprising to see hrabel do well also considering his experience.

Wow is O'Marra ever terrible.

And the fact that Schremp is putting up numbers against iffy AHL opposition makes me wonder if he can even cut it in the NHL at all.

by Black Dog on Dec 31, 2008 9:11 AM PST reply actions 0 recs

My favorite thing about this list is that it help to explain the love the organization has for Tim Sestito and Liam Reddox.

by Jonathan Willis on Dec 31, 2008 9:30 AM PST reply actions 0 recs

Wow, this is an awesome post. Thanks!

by Coach pb9617 on Jan 1, 2009 6:22 PM PST reply actions 0 recs

I'm a bit encouraged by Wild's numbers here - or maybe I was overly discouraged by LT's top 20 take on him.

He's putting up the +/- while playing seconds. Bina, who has played barely better opponents is -8, while Wild is +8. Roy is -14 with easier minutes (though maybe Chorney is destroying him).

That...ain't bad for Wild is it?

It's possible that LT's position may change as this was a comment he made in the Wild entry:

I'm sure there is, but it's an open question as to where he is as a prospect versus Taylor Chorney. Wild isn't likely getting the tough minutes Chorney's getting and may not deserve them.

I wonder is Wild deserves a second look.

by Coach pb9617 on Jan 1, 2009 6:42 PM PST reply actions 0 recs

You did a great job on this Jonathan. Appreciate it very much. Am not surprised that this makes Peckham look even better. It confirms what I thought in tracking his boxcars. The kid is going to be ready whenever the team makes the call.

Am pleasantly surprised by Hrabel too. Maybe there really is some NHL depth in the system.

Two questions. What's your take on Spurgeon? Am not sure if he's getting any PK time and if he'd be an option for a bottom 6 wing one day but am curious if he'd had some heart and soul.

Also do you think that Wild could eventually be a 6/7 pairing (as a puck mover)?

by oilerdago on Jan 1, 2009 6:43 PM PST reply actions 0 recs

Wild's got size, is putting up decent numbers, and seems to be playing a decent level of competition.

Somehow, though, he's getting healthy scratched - there must be something we aren't seeing. Still, if he keeps this up, I've got to think he's a legit prospect.

As for Wild vs. Chorney, I don't know. Chorney's been hyped more all down the line, but this season's been so bad I almost wonder if we don't write it off (like we did for Plante a year ago).

As for Spurgeon, I've always pulled for him despite the indicators (size, injury issues, offensive challenges) because he seems like a real character, hard-luck guy. He's been a WHL captain before (IIRC) and Guy Flaming had him as his favorite for Springfield captain this year.

I think he's not a bad bet as a bottom-six guy down the line, if he can stay healthy.

by Jonathan Willis on Jan 1, 2009 9:27 PM PST reply actions 0 recs

the only thing I'd nit pick about would be that other teams may be trying to get their Reddoxes and Spurgeons out against the Schremps and Potulnys. As we've seen with the oilers, the guys the org considers to be most nhl ready are the guys with apparently solid two way games and these players may not score very high by this metric. having said that, this definitely does shed some light on what the org sees in these players. interesting stuff.

by relic on Jan 2, 2009 9:35 PM PST reply actions 0 recs

Relic: True enough, just like at the NHL level. Still, if Truitt is trying to run PVP, we should see it somewhat in these numbers. Brule, for example, has been used in that role a fair bit, while Schremp has not.

by Jonathan Willis on Jan 2, 2009 9:45 PM PST reply actions 0 recs

Just a few comments on the metric.

From a very coarse point of view what you're essentially measuring is how much time players spend on the ice, 5-vs-5, against high scoring opponents. I mean, technically that's not what you're measuring but I think there would be a very close correspondence.

This is a nice thing to know but in a sense it's also a measure of the coach's psychology -- what role they have in mind for that player.

It doesn't distinguish between players used in a power-vs-power role and defensive specialists, or whether that player is being killed or making a killing.

So this number is probably well complemented by the on-ice/off-ice +/- "Behind the Net" number. Are those numbers computed anywhere?

by delooper on Jan 3, 2009 10:19 AM PST reply actions 0 recs

Ryan Budney: That's pretty much exactly what I'm measuring. As far as I know, there isn't any kind of +/- number available (AHL play-by-play sheets aren't made available like NHL ones are).

This is an imperfect measure, to be sure, but it does give us a window into what kind of opposition each player's facing. I haven't found any information beyond the AHL gameday scores that helps us do that.

by Jonathan Willis on Jan 3, 2009 11:27 AM PST reply actions 0 recs

I think this number makes sense as a pure metric of events against X-quality of player, but also having a separate +/- of sorts attached to it (i.e. -2/-1/0 for GA, +2/+1/0 for GF) would tell us how well these guys are playing against quality. Obviously, it'd be nice to have shot data and ice times and faceoff zones and stuff like that, but if this is what's available, it's probably as good a proxy as anything.

And while I see what LT's saying about a null value for a goal, I'm not sure what you do to adjust it, unless you change the weighting altogether (4/2/1? 5/3/1?), because changing 2/1/0 to 3/2/1 just makes every number bigger by one.

And now I have to add this site back to the blogroll. Why does everyone keep changing things? ;)

by Doogie2K on Jan 3, 2009 1:04 PM PST reply actions 0 recs

Somehow, though, he's getting healthy scratched - there must be something we aren't seeing.

Have you read anything about the relationship with the Coach? I know you're looking for critique of the metric, but this is really interesting to me. These numbers are a bit stunning.

by Coach pb9617 on Jan 3, 2009 1:14 PM PST reply actions 0 recs

To address Lowetide's concern you could adjust the formula a bit -- currently he's using the "step function" with breaks at .5 and .75 ppg as the "weight" for each measured event, which in his case is a goal scored for or against. Instead of using a step function you could use a linear scale. Ie: for every goal scored you add up all the opposing players PPGs. It's simpler and deals with Lowetide's issue. Call it the "sum of opposing PPG goal event total" SOPPGGET for short.

by delooper on Jan 3, 2009 1:15 PM PST reply actions 0 recs

About the possibility of different values for GF and GA - I'm really not sure it's a good idea.

I'm going to post the two separately, but really I'm less worried about the nature of the event vs. the amount of events against who. I was just using the GF/GA data as a way of measuring who was on the ice against who; it should work like polling, with the GF/GA serving as a representative sample that should accurately reflect ice-time as a whole.

The only thing I really expect to see by posting the numbers seperately is that the quality of opposition will be higher for GA than GF.

by Jonathan Willis on Jan 3, 2009 1:20 PM PST reply actions 0 recs

I suppose my last post was a bit off. Using the average would be more interesting than the sum. AOPPGGET it is.

by delooper on Jan 3, 2009 1:22 PM PST reply actions 0 recs

Ryan Budney: That's a good idea, but a hell of a lot more work.

by Jonathan Willis on Jan 3, 2009 1:22 PM PST reply actions 0 recs

Spreadsheets are your friend. :)

by delooper on Jan 3, 2009 1:25 PM PST reply actions 0 recs

RB: Well, I'm using spreadsheets, but the data entry is still manual.

by Jonathan Willis on Jan 3, 2009 1:28 PM PST reply actions 0 recs

You need grad students. Ask them to write some python scripts to troll the AHL webpages and automatically convert all the data into an SQL database.

by delooper on Jan 3, 2009 1:30 PM PST reply actions 0 recs

What I need is a little more computer know-how. I'm rather bad at programming.

by Jonathan Willis on Jan 3, 2009 1:43 PM PST reply actions 0 recs

Maybe the Oilers blogosphere is big and data-sensitive enough to create a data/stats mining and analysis forum. We could teach each other how to mine webpages for data, and develop useful models for assessing player development. But then I think to myself "I'd want to be paid to do that."

by delooper on Jan 3, 2009 2:18 PM PST reply actions 0 recs

Fuck me, that's just ridiculously good stuff, Jonathon. Frankly, it's probably better than using ice times (if they were available), of course there aren't all that many goals in hockey, but besides that it's gold imo. Lowetide wasn't overstating with the 'watershed post' remark.

And poor Matthieu Roy, dude has really fallen into the abyss. It wasn't that long ago that dude was dominant at the AHL level. I liked a lot of his game as an NHLer as well, aside from the 'nearly getting killed once a fortnight' bit. You never know, but it does look like he's done as far as NHL hopes go.

The rest seems to generally mesh with the way that the organization appears to be valuing these guys. I would think that the guys near the top are, on average, playing something fairly close to NHL bottom-six level opposition, probably at least NHL 4th line level anyways.

by Vic Ferrari on Jan 3, 2009 5:46 PM PST reply actions 0 recs

Vic Ferrari: Thanks, that means a lot.

Roy looks done alright; I wonder how much of his game's disappearance is injury related - he was a pretty good prospect at one time.

I agree there probably isn't a world of difference between NHL 4th-line difficulty and top-line AHL difficulty. By that, Reddox was ready and Brule and Spurgeon are close. Sestito isn't far off, but he needs to generate some offense. He's off even last year's awful scoring pace.

by Jonathan Willis on Jan 3, 2009 6:19 PM PST reply actions 0 recs

Yeah, that tends to be the problem with guys still in the AHL at this age. They either play an honest game and struggle to create enough offense to outscore opponents, or they create enough offense but cheat a lot to get it, so it doesn't translate well to the NHL, where even the opposition's depth is just so strong.

by Vic Ferrari on Jan 3, 2009 6:29 PM PST reply actions 0 recs

I might be a bit slow and don't catch on to quick but will these results be off the mark slightly due to being based on a PPG metric. Wouldn't some of the better defensive players be matching up against say the Schremp/Polutny's. There PPG's would no doubt be pretty low.

Is this a concern or am I way off track?

by Ozzieoiler on Jan 3, 2009 8:52 PM PST reply actions 0 recs

Ozzieoiler: It's a bit of a concern. The biggest thing it helps us with is the +/- numbers; Chorney's -21 isn't coming against the offensive powerhouses, and Peckham's EV is.

It's not reliable for telling us who is playing against defensive specialists, but the best players (guys from the Brule/Reddox ballpark) in the AHL should be providing both offense and defense.

by Jonathan Willis on Jan 3, 2009 10:34 PM PST reply actions 0 recs

Fantastic post Jonathan, really and truly. One thing that might be helpful is tracking this stuff against EV offense instead of offense as a whole. Some guys may be getting good results at EV but no time on the PP and are then undervalued as a result, perhaps even given no value at all. It seems like that would be fixable but it may also be a lot of work. Regardless, thanks a lot for all of the work that you've put in.

by Scott on Jan 4, 2009 1:52 PM PST reply actions 0 recs

Scott: That's an excellent idea. Does anyone know where EV stats are? I'm thinking that the only way is to go through the AHL game sheets to get that information.

by Jonathan Willis on Jan 4, 2009 2:24 PM PST reply actions 0 recs

Great post, Jonathan.

Would multiplying these values by PPG to create an "adjusted PPG" be of value?

i.e. Rob Schremp would have (23/28)*0.694 = 0.56 Adjusted PPG

Brule would have (16/19)*0.911 = 0.77 Adjusted PPG.

Schremp fans might suggest that he is playing against the AHL's Pahlson/Sestito types and therefore getting the shaft, but I would probably suggest that if you can't create offense against Sestito you're probably not an NHLer.

Could also rectify the teams plus/minus somehow and make a similar stat. Maybe rectify the lowest +/- to 1 i.e. if -21 is the low, add 22 to everyone's +/-.

Then Chorney = (-21+22)*0.806 = 0.806 Adjusted +/-

Peckham = (0+22)*0.977 = +21.5

Not sure if that's of any value (or whether rectification in that manner is a valid way to do it), but it helps me compare the numbers.

by dubya on Jan 5, 2009 4:37 PM PST reply actions 0 recs

After letting this ruminate for a bit, a couple of thoughts:

I think the 2-1-0 point system works.
You could tinker with it, but the results would hardly differ.
You didn’t count opposition D-men & I know why.
Would giving a point for an opposition D-man over .75 make sense?
I’m not sure that it does, I’m just trying to get more data.

not thrilled with zero value for a goal for (even against weak opponents)

This goes whit LT’s oft repeated statement – "the AHL is a very good league".
It is.
But this process isn’t measuring that.
Giving zero credit for playing the Paukavichs of the AHL makes sense here.

Wouldn't some of the better defensive players be matching up against say the Schremp/Polutny's

As was noted,
Reddox can score in the AHL, and he’s a defensive player.
If you can’t score in the AHL, you won’t play in the NHL. Period.
Unless you can fight, of course.

by Mr DeBakey on Jan 7, 2009 10:51 AM PST reply actions 0 recs

Dubya: I've been thinking about multiplying that number by a player's scoring number, but I don't know. It makes sense intuitivly, but I'm worried it's a little too much apples to oranges.

by Jonathan Willis on Jan 7, 2009 11:04 AM PST reply actions 0 recs

You didn’t count opposition D-men & I know why.
Would giving a point for an opposition D-man over .75 make sense?
I’m not sure that it does, I’m just trying to get more data.

I thought about it, mostly to include them somehow, but in the end I didn't like the idea. From what I've seen, I think most coaches matchup against forwards; i.e if I see Hemsky, Horcoff, Penner, Gilbert and Grebeshkov out there, I'm not sliding my soft forwards out to get them away from Souray and Visnovsky, I'm sending out the A-team.

by Jonathan Willis on Jan 7, 2009 1:06 PM PST reply actions 0 recs

Vic just posted over at LT's site that he used this technique on NHL numbers and compared it to BehindtheNet with a .92 correlation; in other words, we can take these as a pretty accurate indication of the difficulty of opposition.

by Jonathan Willis on Jan 7, 2009 2:58 PM PST reply actions 0 recs

Comments For This Post Are Closed

Search

FanPosts

Community blog posts and discussion.

Recent FanPosts

+ New FanPost All FanPosts >

Northwest Standings

	GP	W	L	OTL	PT
Vancouver Nucks Misconduct	57	35	20	2	72
Colorado Mile High Hockey	58	33	19	6	72
Calgary Matchsticks and Gasoline	59	29	21	9	67
Minnesota Hockey Wilderness	58	29	25	4	62
Edmonton The Copper & Blue	58	18	34	6	42

(updated 2.9.2010 at 1:15 AM PST)

18 - 34 - 6

Streak: Lost 3

Oilers Schedule

Next 5 Games

@ Anaheim Anaheim Calling	Wed 02/10	7:00 PM PST
@ Los Angeles Jewels From The Crown	Thu 02/11	7:30 PM PST
Anaheim Anaheim Calling	Sun 02/14	4:00 PM PST
@ Nashville On the Forecheck	Tue 03/02	5:00 PM PST
@ Chicago Second City Hockey	Wed 03/03	5:30 PM PST

Last 5 Games

@ Phoenix Five For Howling	Mon 02/08	L 1 - 6
@ Colorado Mile High Hockey	Sat 02/06	L 0 - 3
@ Minnesota Hockey Wilderness	Thu 02/04	L 2 - 4
Philadelphia Broad Street Hockey	Wed 02/03	W 1 - 0
Carolina Canes Country	Mon 02/01	W 4 - 2

Oilers Stats Leaders

Stat	Forwards	Defense
TOI/G:	Horcoff (19:37)	Souray (22:46)
ESTOI/G:	Horcoff (14:59)	Grebeshkov (17:23)
Points:	Penner (44)	Visnovsky (28)
Goals:	Penner (22)	Visnovsky (9)
Assists:	Penner (22)	Visnovsky (20)
EV+/- /15	Penner (.190)	Visnovsky(.137)
Shots:	Penner (153)	Souray(111)
Corsi/15:	Penner(1.26)	Visnovsky (.945)
SCF/15:	Penner (5.347)	Visnovsky (4.751)
SCA/15:	Stone (3.364)	Gilbert (4.284)
SCDiff/15:	Penner (.688)	Visnovsky (.405)

The Copper & Blue: An SB Nation Community

Network Bar

The Copper & Blue -

An Edmonton Oilers Site - Where Steve Tambellini is still assessing Jeff Deslauriers