first faceoff props posted and looks like our fears are true

everything -115

we got a model just gonna see what happens for the first couple days

todays goal is to see if the lineups appear in the API so we can automate the process

Opening puck drops today, which means I spent this morning puzzling over an old spreadsheet continually asking myself "Wait, so why did I do that again?"

Based on how things are scheduled this season, I expect time on ice to be spread out more than usual creating value on the unders. Today's plays.

PIT Sidney Crosby u2.5 -103
TOR William Nylander u2.5 +103
TOR Auston Matthews u4.5 -131
TOR Mitchell Marner u2.5 -114
TBL Victor Hedman u2.5 +100
EDM Ryan Nugent-Hopkins u2.5 +124
COL Nazem Kadri u2.5 -115
COL Mikko Rantanen u2.5 -113
STL Brayden Schenn u2.5 -202

9 plays. Risking 10.78u.

Last season I was continually tweaking my model and got stuck huge early on. I lost something like 30u, settled upon a model, clawed back to even, and then moved onto other things. All lines are taken from Pinnacle.

I’ll be battling with you this year TomG. Re tooled my model from scratch for the playoffs last year and I’m really liking it. See you at the window, guy.

Here are my opening faceoff power ratings for all skaters. I take all neutral zone faceoffs from the last three years, weight them, and regress based upon sample size to a mean of 0.50. Standard monkey projection system stuff. It's useful for those hammering F5 refresh on nhl.com right before the start of game. As soon as you see who will be in the opening faceoff plug this into your handy Log5 calculator to get expected win%.

https://pastebin.com/WvmZhMwe

Happy to hear you're around and doing these again. I really liked the methodology you outlined in your blog last season. I would love to hear more about any improvements you came up with and also any special ways to deal with this unique season.

Ben Simmons getting traded to the Rockets might kill the Ben Simmons u3.5 3 pointers in the season prop. Thats all they shoot.

Already a lot of shots on goal steam revolving around Malkin

Thanks guys for the hard work and these picks, GL!

Just in case you have not considered and if the unders dont go well, the following rule change for offside my have an impact on goals and shots on goal:
https://www.nhl.com/kings/news/know-...op/c-320216632

I've read on NHL.com in above link that this rule change would mean 14 more goals overall for all of last season.

I'll add this as an output in my bot, will let you know how it goes. Thank you.

Question re: using a mean of 0.5. We have evidence to suggest that there is a difference based on which hand the person is using. Should we scale differently and have two projections for each person

One for same hand opponent, one for opposite?

here's what I did the past week. I tried to bet the faceoffs from last season pre-bubble at hypothetical -115 lines

using log5:
without adjusting for hands it predicted 54.9%
adjusting for hands 55.7%

using the method outlined in this paper for people who had n >= 10 against each other and regular log5 if not: 59.6%

https://storage.googleapis.com/plos-...67d780c58076cf

I'm going to calculate all 3, but I think this Bayesian log5 approach is superior.

All 3 were profitable against -115.

What do you mean, hands? Some guys do better vs leftys or rightys? Like in baseball? Is this something that could be just random noise or is there some techniques certain guys have that make them better vs a same handed guy?

Yo sigs, that paper link is broken.

whoops. here's the link. it's open source and i guess the link has a time on it

https://journals.plos.org/plosone/ar...l.pone.0204874

and yea some people are betting against other hands. e.g., look at mika below



you might look at his aggregate face off % and be like wow mika's good. 53.42%

but then if you look. 54.75% of his wins are against lefties and 50.76% against righties.

not all players have such extreme splits, but if we adopt the methodology outlined in the paper we can avoid grouping all lefties vs. righties together, etc.

or just use regular log5. have an advantage but not as big but who cares because its a $250 market and the time spent increasing our edge isn't worth it

MODEL starting out good

Crosby 13-29 lifetime against Couturier coming into the game.

got fair line as coutuier -152 so i can't resist -115

now 14-29

i ****ing hate the nhl

I have fair line Crosby -120. I usually pass on those smaller edges so 0-0 so far for me

i gotta get my game up if -120 is good

i got couturier 57.83% lifetime and 59.84% against lefties

crosby 49.53% lifetime and 49.73% vs. lefties


and since the beginnign of time this is their h2h

I didn't find any face-off lines on Pinny or Bovada. Where are you guys seeing these lines?

Cris


Sent from my iPhone using Tapatalk

Looks like I need to update my stats, last season Couturier was 60% vs left and Crosby was 56%. Whoops.

Careful if you're going back before 2017, they changed the rules and enforcement standards.

i'm only doing last 2 years excluding the bubble

I will update accordingly

so nhl.com didn't decide to update for the second game. 5 mins in and they're still on the pre-game and BM still got the faceoff line up (Tor won if you want to get banned by betting it)

classic

btw if anyone wants to track just pinnacle NHL props

http://crowdsourcesyndicate.info/props/nhl/

slow but will eventually load

It was a past post on the early game too for a while

FD limited me to $11 on CBA ML, how am I supposed to feed my family?