[eaglesfan371]

NHL
Boston Bruins -113
New York Islanders +102

Boston has won the past 6 matchups vs Islanders including 2 wins this season. Thinking Boston would win 7 in a row would be recency bias but I argue there has to be some correlation between specific matchups using past performance.

Under Sagarin and eigenvector models the following are displayed:

http://sagarin.com/sports/nhlsend.htm
Sagarin:
NEW YORK ISLANDERS 0.05 0.07 0.04 0.07 boston bruins 104 51% 4.97
NYI being the favorite winning 51% here for an implied line of -104

Eigenvector:
@ New York Islanders 0.15 111 53% Boston Bruins

NYI being the favorite winning 53% here for an implied line of -111.


According to the model, it makes sense with home team receiving an additional 0.28 in rating with Boston at 4.34 vs NYI 4.11 on neutral field.


How do models take into account recency bias vs certain teams? Is there an value that can extrapolated from previous matchups instead of just calculating a point ranking system for each team, independent from matchup. While every game is different, I would still argue there is a small correlation between certain teams performance vs another specific team rather than as a whole league. For example, team A in X league wins 55% games over 3 years but against specifically team B, due to certain player matchups, wins 65% of the time. Is this already inputted into the team point rankings or can we further correlate this out? My thought is that the point system is a first level, with say a correlation of 0.8 of teams performing against an entire league, with 0.1 or less correlation specific to team matchup.





If I assumed Sagarin/Eigenvector were true models of probability (I do not at this time as I have not reviewed the models results), would that imply public bias has the wrong team favored here?

If the wrong team is favored here why have sharps not moved the line? Knowing that a large % of money will come in on NYI, is it expected that books will simply just leave the line as is even if sharps come on NYI?

At first glance of NHL games tonight, this screamed Boston win without looking at models. So I'm sure the public will be all over Boston too.



Sorry for repetition just trying to understand more of analytics and probabilities when it comes to sports betting. Work in finance modelling so it intrigues me.

[eaglesfan371]

I'm probably really overthinking this but IDK.

Just surprised the models have NYI favored here and trying to explain how those two league-based models can be wrong based on correlation of previous matchups of these two teams to imply the line -114 is actually correct. Or is it a line to attract public at -EV when the wrong team is favored.

Rambling I know.

[eaglesfan371]

With a goalie matchup that is nearly identical, 2.3 GAA vs 2.2 GAA, yes away matters here but offensively Boston is MUCH stronger than NYI offense. Raask at 2.33 GAA only has 91.6 save % vs Lehner 92.7% however NYI is generally outshot and Boston has a + shot differential while NYI does not.

With very similar defenses how can NYI be favored?

[vampire assassin]

One thing many models have problems with is adjusting for injuries, resting players, or starting goalies. If you aren't including these factors in your model (and your projection) you will get some wonky results.

[danshan11]

Boston has a ton of injuries

[danshan11]

Bruins avg line -151
Islanders -104
Islanders at home about 40-50 cents
so ISlanders -145
Bruins -151
makes bruins about a 6 cent favorite
figure a minus -108 line add 6 cents gives you -114 and Bruins are hot right now add a nickel gives you -119 or so
when I say Bruins hot I mean public and nhl tv are all on the nut sack of the bruins, so they will draw a few cents more

[danshan11]

its really that simple, you can pretty much add 40 cents on near +100 avg teams and +50 cents on -150 or +150 teams for HFA, the only trick in the NHL is goalies. they suck cause they switch goalies last minute and their overall line gets swayed if they have a starter goalie extended injury or time off. No need to actually model the NHL, sags, PR, and closing lines make it pretty simple