Please PM me the new factors. I am very interested to see the changes, even if it is a work in progress.
Why do you think the Hockey Abstract calculations are bogus?
This is based on a discussion we had on here, with a few fans of the league. The NHLe calcs have the conferences in the following order: NCHC, H-East, Big 10. Some posters on here, I forget who and would have to find the thread, posted some pretty compelling evidence that the Big 10 is the strongest conference, despite it having the weakest NHLe. I am going from memory so I might not have that quite right, in any case, the NHLe's for the conferences were well out of whack with what the head-to-head conference records were, to a strong enough degree that it caused me to question them. I think this might have been in the Quinn Hughes thread. I really need to figure this out.
Rather than PM them, I'll post them later so that everyone can provide feedback.
EDIT: The discussion was here:
Player Discussion - #40 Elias Pettersson, Pt. VI
It seems pretty clear that H-East is the weakest conference and always has been at least for some time, yet NHLe frequently has calculated it as the strongest.
I'm unclear at to the point of your first paragraph. Isn't NHLe an attempt to adjust for the performances in disparate leagues? If true, isn't this to help the projection of a draft-eligible player's probability of playing NHL games? In other words, NHLe is a projection based upon probability itself.
It is a subtle difference but an important one. The point of NHLe is to adjust a player's points to an NHL context. It does this by looking at all players that switched leagues and doing a before/after analysis. This means that for example the strength of the KHL is influenced by players like Phil Larsen and Linden Vey and such.
I am not really trying to measure the strength of the league in that sense. I am merely looking at how the production of draft-eligible players translates to future NHL success. Therefore my numbers are only influenced by whether or not a player "made the jump." This is a reason why I have the KHL lower than the SHL, because even if it is a "better league," the fact is that 18-year-olds who produce in the SHL are more likely to "make the jump" and produce in the NHL than similar players in the KHL. IOW, my model implicitly has a built-in "Russian Factor" just because of how it was put together. There are other reasons though why a junior player's performance may be better than you might think due to the NHLe. Some of the pro leagues really really dislike giving ice time to younger players, so a teenager putting up points there is a much bigger deal than the NHLe might suggest.
So when I have one league at 1.5 and another league at 1.6, it's not necessarily saying that the latter league is "better" than the former league, just that a teenager putting up points in the latter league is slightly more likely to become an NHL player than the same player profile in the former league. It's a subtle distinction and hard to explain, so I hope that makes sense.
Ah, I see... fair point. I may be jumping the gun a bit. I incorporate the year over year NHLe, and the indirect strength of the USHL this implies. Perhaps I am moving too quickly.
This is a matter of personal preference. I don't like to change factors for a league unless I have an actual underlying reason for why the league quality may have changed. Has there been any important changes to the league structure, its rules or its schedule? Has there been an expansion or a contraction? Have they moved to a different system for relegation/promotion? Etc. Otherwise, year-over-year league quality is just a moving target and I feel more comfortable averaging it out and taking the larger sample size of multi-year factors.
Question(s): You said that you take the NHL success of players drafted out of the USHL and compare them to similar players drafted out of the CHL. What is this similarity? Also, how did you determine the relative strength of the USHL to the NCAA?
I don't compare any leagues to one another directly, my index is the OHL and all leagues are adjusted based upon their predictability as compared to a similar player from the OHL. Here are some of them for current league configurations:
OHL 1.00
WHL 0.95
QMJHL 0.95
NCAA 1.20 (see above re: conference-specific adjustments which I am not confident in right now.)
USDP 0.85
USHL 0.75
SuperElit 0.75
Jr. A SM-liiga 0.80
MHL 0.85
BCHL 0.50
NLA 1.90
Allsvenskan 2.70
SHL 3.5
SM-Liiga 2.65 **
Liiga 3.40 **
KHL 3.25
** This is a bit of a weird one. In 2013 the league re-branded as Liiga and while there is not clean split in terms of quality, there were a number of changes to the league around this time period which did have a major effect on the quality. Since there was no easy way to split them all up, I just decided to put two different factors for the older branding and the new branding. This could probably be optimized and split a little bit more evenly.