View New Posts
1. ## let's talk inefficiency of NHL markets

number one on my list: team to score first

intuitively this one is just flawed to me. naturally i have no data to back this up at this point, and would love alternate views proving me wrong. (aside: unsure if this is right, but i usually take 20+ Cent lines as another indication of inefficiency as the book insulates itself against posting risky numbers)

i might be to simplistic in my approach, which is based on actual first goal pct to deduce a probability and estimate the line.

logic as follows:
• tonight car hosting ana, they have matching % on scoring first.
• assumption: league average will be .500
• in such a league, when 2 teams meet that have equal first score %, the likelihood of either scoring first is 50%

i backed at Ana +137 based solely on that logic. want to refine it from there, but having a terrible time finding any site that breakdowns how H/A first goal stats.

maybe i should also deduct the teams average from the league average (meaning a team at 55% first goal scoring pushes the average of the rest of the league under .500, vice versa on 45%). goaltender must have an impact as well.

thoughts?

2. Good post...the Home Ice certainly plays into this. Goal tending also. Would factor days on the road/injuries also.

There must be somewhere to show you the H/A first goal stats.

\$20
Angelman
donation 02/17/2020

3. What if the expected score is 3-2? Would that have any influence on how you look at this? Joe.

4. Like the idea of, when comparing two teams, to compare them to the league average. I would use medians there.

Goal keeper would be like a strength of schedule adjustment, I suppose. Maybe more than goalkeeper, like overall defense, could be a refinement.

We need stats to take out the .500 assumption. Unfortunatley, it's hard to get anything inside of 1st quarter stats and end results.

Joe, we know that the team that scores first wins pretty much 2/3 of the time. So the final score may be irrelevant, unless you are converting the margin of victory to a percentage, then if you have a larger margin of victory, a more likely chance of winning, then perhaps there's a likely chance of scoring first.

What a mess, lol.

We may have to track some stats ourselves here.

\$20
Angelman
donation 02/17/2020

5. Of course it's not that simple because it's matchup dependent. Logically the favs will score first more often than dogs.

It's the same thing with totals, people think "Team A averages 6 goals, Team B averages 7 goals so together they should score 6.5" but obviously they're going to score less against the #1 team and more against the bottom feeders.

6. Originally Posted by Gaze73
Of course it's not that simple because it's matchup dependent. Logically the favs will score first more often than dogs.

It's the same thing with totals, people think "Team A averages 6 goals, Team B averages 7 goals so together they should score 6.0" but obviously they're going to score less against the #1 team and more against the bottom feeders.
This is why, when comaparing teams, we compare them to some sort of league average.

There's no doubt adjustments have to be made to account for those differences...

https://www.sportsbookreview.com/for...l#post25176977

It's a basic, simple take, but that link is what I mean when say comparing to a league average.

\$20
Angelman
donation 02/17/2020

7. thanks gents for the feedback.

i did find the H/A first goal stats and there having that final line change must really benefit the home team, as they score first 57%. reworked all of my numbers and it flipped my car probability on it's head and estimated that line at around -178 for car to score first.

however, also made me play the red wings to score first at +140 based on the numbers, and my own personal belief as a Pens fan that pit gives up that first goal much more often with Murray in net, they played a physical game last night in boston and howard is coming off a horrendous outing. unless crosby decides to be crosby and score 24 seconds in again, the jerk.

the drawback of having H/A numbers is it basically cut my data results in half, from 48 or 50 a team to sometimes as low as 22.

8. I don't like to use specific team Home vs. Away. It tends to be a small sample size, and I'm not sure any team that was significantly different from the normal would actual be a signal, but if you have a 95% confidence with 32, 2 or so will be just on random chance. I always just calc a constant on things like shooting, penalties, puck possession ect., based on the league average home vs away.

My understanding is that teams have specific combinations of lines, which would play in sequential order. If you were to say that, line one plays against line one for the first 3 minutes or so, followed by line 2 for 3 minutes, ect. You could get the plus/minus numbers, and figure out how often they should score in those intervals, and run it out, and add the percentages back.

Alternatively, you could just use the team numbers, and calculate goals per minute or something and figure it out that way.

It is an interesting idea, much more difficult to calculate than basketball, baseball or football. Keep it going, maybe we can figure out something here.

9. How did it go on Friday?

\$20
Angelman
donation 02/17/2020

10. My initial reaction to this was that it would be an interesting project to pursue. So I started down the logic trying to figure out an approach. But my logic led me to a simple approach. Follow along and tell me where you disagree.

1. First assumption is that the game money lines are efficient and give us some accurate probabilities of each team winning.
2. The winning team will score more goals than the losing team.
3. Goal scoring occurs somewhat randomly, and is distributed between the teams with the same proportion as the total goals.
4. The probability of scoring the first goal would be in the same proportion.
5. Use the computed probabilities from the game's money lines to evaluate each side of the first scores lines to compute the expected returns.

Unfortunately I don't know of a source of this particular props history, so can't back test it.

11. Originally Posted by SBR Drew
How did it go on Friday?
cashed out the ana play for a small loss and put down money on car. det did indeed score first.

my numbers are so far from the lines posted, any positive short term results are going to be blind luck i think. i hate posting after the fact, but my 'biggest' play today was Ott to score first because the math had them at minus odds but they were +115. when i say big play, it was for .25 Unit, with a Unit being 1% of my BR. so clearly i have limited confidence in this approach.

that is where the H/A is making a huge impact. coming in they scored 1st goal at H 57% of the time while Cal is at 38% scoring 1st goal A.

my numbers are so far off, that my approach is flawed.

12. thanks for sharing bsims! answers in BLUE

Originally Posted by Bsims
My initial reaction to this was that it would be an interesting project to pursue. So I started down the logic trying to figure out an approach. But my logic led me to a simple approach. Follow along and tell me where you disagree.

1. First assumption is that the game money lines are efficient and give us some accurate probabilities of each team winning. Not too much of an argument here.
2. The winning team will score more goals than the losing team. Agreed, naturally.
3. Goal scoring occurs somewhat randomly, and is distributed between the teams with the same proportion as the total goals. Interesting statement. I have no data to support or refute your claim. I guess we could look to scoring props that list Goal in 1st 5 minutes or 1st 10 minutes of a period.
4. The probability of scoring the first goal would be in the same proportion.
5. Use the computed probabilities from the game's money lines to evaluate each side of the first scores lines to compute the expected returns. And i think this is where a slight disconnect might be happening between game lines and team to score first. I always hear this stat thrown about that the teams that scores 1st wins 67% of the time. Logically it follows from there if a team is favored by a fair amount, and the market is efficient, the favorite will score 1st more often than not. That stat falls apart when we go team by team for sure. coming into today, Ottawa scored 1st in 18 games, but only won 9 of them. Good teams can give up the first goal, even 2, and still come back to win. Of course a 1st period 1st goal might have less value than a 3rd period 1st goal, but think my statement holds up in the long run. Continuing on with the Ottawa example, they gave up the first goal in 29 games, lost 22 of them. 76% is well above the smooth league wide results of 67%. Stands to reason there are teams out there that lose less than 60% of the time when giving up the first goal. This will be fun to track. If i start tracking i'll avoid posting it in HTT, but i'll throw a link in this thread. and by no means should anyone place any real money on this thing until we have about 90-100 results.

Unfortunately I don't know of a source of this particular props history, so can't back test it.

13. I find NHL markets to be efficient for the most part, if they weren't, they wouldn't offer the bet after all these years. Numbers books offer are pretty solid so what I try to do is find a team that has been un lucky or is better than looked facing a team that is inferior as far as match ups. Using saber numbers in hockey only goes so far as the books are using the same thing. NHL is profitable although plays are few that are +EV IMO.

14. Did quite some research. Where did you find the first goal stats? I can't find them using google anyway and every search term, etc.

15. EMT, interesting stuff , thanks for putting it out there

16. Originally Posted by oilcountry99
EMT, interesting stuff , thanks for putting it out there

Oil, good to see you buddy!

appreciate the kind words but a thanks is not needed, we're all in this together.

taking a long look at 2nd period totals now, think there might be something there, but need some time.

17. Originally Posted by hehfest
Did quite some research. Where did you find the first goal stats? I can't find them using google anyway and every search term, etc.
the tracking has begun over in the NHL Forum

found them on hockey reference dot com. select any team and then select 'situational'.

really wished i would have spent the time to finishing a screen scraping program back in the day. but this is where we are.

but now that i have a spreadsheet built out, can manually track and simultaneously work on my prob formulas and repeat YOY. (ASSUMPTION: there is an edge here for the time being that proves profitable)
Nomination(s):

18. Originally Posted by Believe_EMT
the tracking has begun over in the NHL Forum

found them on hockey reference dot com. select any team and then select 'situational'.

really wished i would have spent the time to finishing a screen scraping program back in the day. but this is where we are.

but now that i have a spreadsheet built out, can manually track and simultaneously work on my prob formulas and repeat YOY. (ASSUMPTION: there is an edge here for the time being that proves profitable)

Darn. I was at the reference site too (I mean before you posted situational stats), but looking under play index and only found corsi and fenwick. I've used fenwick with some success in the past. The situational stats is where I used to go for MLB pitchers all the time.

19. Wow. Doing good. So how are you figuring this out again? PM if you don't want it public. This just really interests me. I already applied my own theory but it had me scratching the Rangers pick. I think I see what you are doing here. If I'm right the Nyi/Car wasn't really off by much and I thought the Rangers wouldn't have been either, yet you still posted them....

20. Originally Posted by hehfest
Wow. Doing good. So how are you figuring this out again? PM if you don't want it public. This just really interests me. I already applied my own theory but it had me scratching the Rangers pick. I think I see what you are doing here. If I'm right the Nyi/Car wasn't really off by much and I thought the Rangers wouldn't have been either, yet you still posted them....
still too early to share as it needs refined moving forward. it isn't so much the computation that is the key, it is the philosophy.

once ran my own handicapping service in the mid 2000's. went well until recession hit. i was successful for 3 straight years in the nfl being tracked by 3 independent sites. point of that story is i was winning at that time because of a unique approach. everyone has access to the same data. developing a novel way to analyze or use the data is the key to winning.

in this case, it wasn't about developing a better model than lines makers or bettors, that is an arms race i have no chance of winning. this case is about considering HOW the line is being made. i think it does follow a similar logic set forth by bsims above.

even in an efficient market, they are doing it wrong if they are going:
A. favorites will win more than dogs (this is true)
B. team scoring first wins game 67% of the time (guess it is true, never tracked down the source but i read it all the time)
C. Home team does score first more often than visitor (this is also true, league wide, but very false at team by team level)
therefore
D. the favored team scores the first goal more often, and a very high rate at home.

naturally that is a severe over simplification on my part. i do believe they are building the derivative market based on the line and league averages and data gleaned over years of play, rather than using actual H/A 1st goal stats.

to this point, the pit at det game is the perfect use case.

pit was favored to win. overall they were at 50% first goal scored and det was an abysmal 35% or something in that neighbor hood. however, pens on the road score the first goal less often than det scores the first goal at home. why then would det be +140 to score first?

think you should be able to piece together the math from here:

what is the probability assigned to each outcome when team a (det) scores first 42% of the time at home and team b (pit) scores first 39% of the time on the road? that would be my approach.

what is the probability assigned to each outcome when team a (det) scores first goal 35% of the time and team b (the heavily favored pit) scores first goal 50% of the time? think current market follows some form of this approach, with much more data and formulas.

it really is that simple sometimes.

21. Shhh...you're going to ruin it for all of us

22. grunching...

how much \$\$\$\$ is bet on "score first" lines? my guess is not much........... so even as a "risk arb" type situation, it might be pretty limited.

23. All Star break now...time to crunch numbers for 2nd half...

\$20
Angelman
donation 02/17/2020

24. A few interesting facts from a friend who does rather well wagering on Hockey. First, while Boston leads the Atlantic with 70 points, they have played the most games of any team except lowly Detroit. Secondly, and more important, Boston has played 28 home games, more than any team in the NHL. That means that out of the 31 games they have remaining, only 13 are at home. While they do very well at home, they are only 12-8-3 away. This makes Boston a very good away favorite fade, especially in their next 20 or so games.

Other things he looks for are teams with large differences between home and away records. The Flyers are 17-4-4 at home, one of the best in the NHL, but a dismal 10-13-2 away. Buffalo is 14-6-3 at home but 8-14-4 away. Minnesota is 14-6-4 at home but 9-15-2 away. Obviously the wager here is to fade these teams as away favorites.

Another things he likes to do is keep track of streaks. Currently there are 9 teams with streaks (win or loss) of 3 games or more. He uses that information to determine wagers. He also looks for streaks by checking how each team has done in their last 10 games. There are 3 teams with 8-2 records in their past 10 games, all in the Eastern Conference.

Just a few things he looks for. I do not wager hockey competitively, so I yield to his knowledge. He normally ends up on the + side of the wager after the season is over.
Points Awarded:
 SBR Drew gave BigdaddyQH 2 Betpoint(s) for this post.

25. I think another big factor is rest days. Home teams are near 50% on back to backs and near 50% with 4 or more days rest. Much different than the overall avg of 55%. Interesting to look at how teams change when playing tired (ie. more defensively in the opening period or not). I agree there is likely some edge in this and wish you the best in finding it.

26. Thanks for sharing these insights...extremely useful!

\$20
Angelman
donation 02/17/2020

27. Originally Posted by gdon44
I think another big factor is rest days. Home teams are near 50% on back to backs and near 50% with 4 or more days rest. Much different than the overall avg of 55%. Interesting to look at how teams change when playing tired (ie. more defensively in the opening period or not). I agree there is likely some edge in this and wish you the best in finding it.
right on. the opposite is true is well. if anyone watched my beloved Pens fall to defeat in Phi the other night, they just didn't have it. think it was 11 games over a 20 day stretch. including 7 games in 12 days. that is a brutal pace and more games that teams used to play when playoff were strictly an every other night affair.

they were gassed in that game. Jarry played well, but that amount of games including the come from behind extra effort game sunday against Bos, it was just too much.

on the back to backs, the goalie is vital in that one. i've seen teams start the back up goalie on the first of back to back if that first game is a lesser opponent than the 2nd. that would be an interesting breakdown, back to back records with no. 2 guy starting 2nd game, the no. 1 guy starting 2nd game, the no. 1 guy starting both games

28. home and away is interesting.......... i think in hockey it makes alot of sense.

pounding terrible teams seemed to work very well at many times in recent past. not sure about this season. i'll check

i think i might have asked this in its own thread, and then never checked the answers.......... would people consider shorthanded goals (scored and allowed... allowed especially) as something sustainable? i..e if a a team is good/bad at this in 1st half of season, would you expect some continuation in 2nd half? OR, IS IT RANDOM?.......... a good/bad team at scoring/defending power play goals i'd expect pretty strong continuation

29. weird question.... the canucks are my local home team. doing very well this season..... but something i've noticed somewhat cursory is that quite often the game score is very inconsistent with the shots on goal differential.... i'm guessing i can check this fairly easily. but would this be quite common? i'm not talking about a team winning 3-2 being outshot 29 to 26. i'm thinking double take on shots after looking at score

30. wouldn't label that a weird question at all, i like it. my assumption has always been, in the long run, team racking up more shots will win more games. i don't know how far advanced stats have gotten in hockey. i'd guess someone has created a way to track 'quality' shots.be great to draw a correlation between zone time and quality shots both for and against.

pens are my home town team (how fun was that 8-6 game this season?!). remember watching a game vs flames earlier this season and cal threw a ton of shots on net, but it felt like most were from outside the face off circles, just randomly shooting from long distance. if shot quality or something similar is a stat, feel pens would be fairly high on the list.

point is teams just launching shots one and done style are different from teams controlling zone time and working for shots within 12 feet of the goal. the latter producing better scoring chances.

i guess a team shooting percentage would somewhat clue us in to which teams take smarter shots. it is interesting, got me thinking.

31. believe_EMT,

the quality shots of goal is very interesting....... i would think power play and how fast a team shoots makes a difference. i remember as kid going to canucks games at the pacific coliseum and fans would be yelling "shoot" on power play. it boggled my mind then how long the canucks took to take a slap shot they could have taken 20 seconds earlier.

another good possibility is that canucks games feature great or terrible goaltending very often by one team

i have been tracking this leaguewide since my comment, and it certainly doesn't seem rare at all...... and subconsciously, i feel like the other team i follow closely - winnipeg jets - have this happen very often too.

32. Originally Posted by gojetsgomoxies
believe_EMT,