Advanced Stats and Win/Loss Predictions

[Menzinger]

Hi all,

I'm a recent convert to advanced stats, as a leafs fan, last season I used to roll my eyes at all the analytical doom and gloom predictions, but let's just say since then I've become a believer.

I've recently came across this article though, and was wondering if some of the more quantitatively inclined of you would be able to deconstruct it? Or just have any general thoughts towards it?

http://rinkstats.blogspot.se/2013/12/why-popular-advanced-stats-are-bad-at.html

When you think about it, none of my findings should as a surprise to anybody. Fenwick and Corsi incorporate plays which, by definition, won't help you to win games. The best consequence of a missed or blocked shot is your team retrieving possession of the puck, putting you right back where you were before you attempted the shot. At worst, you're conceding possession to the other team. This doesn't mean that teams shouldn't take lots of shot attempts, since you don't know whether an attempt will be a shot, miss, or block. It does mean that analysts should think hard about using Corsi or Fenwick as an indication of how well a team is playing. You're taking a stat (shots on goal) that could directly influence the outcome of games, and adding plays (missed and blocked shots) to it which, by definition, cannot help a team win the game since they cannot directly put points on the board.

Click to read more...

 

[soireeculturelle]

The author does a pretty thorough job, but as the comments mentioned, he doesnt not account for score effects. essentially, when many teams go into a lead, they start getting outshot by their opponents 1) because they start playing safer and 2) their opponents become more offensively-minded in their play. lets say a team outshoots their opponents 10-5 and scores 2 quick goals in the first period, they may sit back, win 2-0 and end up getting outshot 20-30. stats like corsi/fenwick close only looks at 5vs5 shots when the score is tied or with either team up by 1 goal.

 

[Menzinger]

The author does a pretty thorough job, but as the comments mentioned, he doesnt not account for score effects. essentially, when many teams go into a lead, they start getting outshot by their opponents 1) because they start playing safer and 2) their opponents become more offensively-minded in their play. lets say a team outshoots their opponents 10-5 and scores 2 quick goals in the first period, they may sit back, win 2-0 and end up getting outshot 20-30. stats like corsi/fenwick close only looks at 5vs5 shots when the score is tied or with either team up by 1 goal.

Cheers!

 

[stephenp]

Thought you all might be interested in my new post about the predictive value of Corsi and Fenwick: http://rinkstats.blogspot.com/2014/05/corsi-and-fenwick-suck-or-why-we-should.html

 

[plusandminus]

Good points made. We just don't know what effect missed shots have. Sometimes a player intentionally shoots beside the goal, in order to make the puck bounce in front of the goal where a forward might put it into net. Sometimes missed shots ends up around the corner, and goes up into neutral zone where opponents can counter attack.

Correct also regarding shots taken. Hockey is about outscoring the opponents, not outshooting them. In itself, outshooting them means nothing of value. If a shot at goal results in an offensive face off, then it's good, but that can be tracked instead by the offensive/defensive zone finishes.

Previous studies also have shown that they more shots a goal faces within a game, the better save percentage he tends to have. Goalies usually has nothing against facing shots from certain areas or distances, and a good defence contributes to getting the opponents to shoot from those "easier" areas.
This probably also applies to teams leading, who can "control the game" and make the opponents shoot from "safe" areas. The opposite regards to the team trailing. The trailing team might shoot more, but when their own goalie is facing shots it might be on dangerous counter attacks (including two-on-one situations).

 

[rosemount289]

One thing that..............?

One thing that Corsi and Fenwick cannot and will not be able to produce..........is to predict a hot, hot goaltender.........that teams face in a regular season game or especially in the playoffs...........so those stats will go out the door.

 

[Bear of Bad News]

One thing that Corsi and Fenwick cannot and will not be able to produce..........is to predict a hot, hot goaltender.........that teams face in a regular season game or especially in the playoffs...........so those stats will go out the door.

No statistic (or combination of statistics) is ever going to predict a single game (or even a seven-game series) with certainty (or even near certainty). That doesn't mean that Corsi or Fenwick need to go "out the door".

And I'd like to see some evidence that "hot goaltending" is a real phenomenon, and not just something ascribed after the fact.

EDIT: The latter is something that I can now test with my database. I'll add it to my list, although there are some things ahead of it.

 

[The Imp]

Would "hot goaltending" not be identifiable through a significant positive deviation in the PDO numbers of the team with the hot goalie? I realize that the team's shooting percentage is in there, too, but there has to be some correlation.

And of course, there is a cognitive bias in PDO, about regression to the mean.

 

[Bear of Bad News]

Would "hot goaltending" not be identifiable through a significant positive deviation in the PDO numbers of the team with the hot goalie? I realize that the team's shooting percentage is in there, too, but there has to be some correlation.

What I'm thinking of doing would be based off of the framework that I've developed to measure above-average starts, average starts, and below-average starts.

For instance, if a goaltender has a game where he's 2 standard deviations above what was expected, what is his expected performance (and distribution of performance) in the following game? If he has two straight games like this, expected performance (and distribution of performance) in the following game? Things like that.

 

[BB1133]

You can make a small fortune using advanced stats for predictions.

 

[Thirty One]

Yeah that was a pretty silly argument made there.

Read this and tell me it doesn't matter. http://www.habseyesontheprize.com/2013/4/4/4178716/why-possession-matters-a-visual-guide-to-fenwick

 

[rosemount289]

Are you..............???

You can make a small fortune using advanced stats for predictions.

If you are serious...........show us please???

 

[Thirty One]

In case any one's interested, I developed a simple forecasting tool based on three variables (shots for/against per game, and average save percentage).

Excel file:

http://rangersunlimited.com/wp-content/uploads/2014/05/forecast.xlsx

edit: I'm not sure if this relates to the topic, but the title made me think of it.

 

[BB1133]

If you are serious...........show us please???

Just create better power rankings than Vegas does (not that difficult) and you'll see where to take advantage.

A good example is where they ranked the Kings/Rangers before the playoffs started.

LA was 14-1 and NYR 26-1. They were both the most underrated in their respective conference, and anyone paying attention to advanced stats this year would've seen the glaring discrepancy. Advanced stats don't make up all of a team's power ranking, but it's a big chunk.

During the regular season, there are opportunities on a nightly basis. Vegas is kinda dumb when it comes to hockey.

 

[Thirty One]

I don't have a foolproof method or anything, but I should end up pretty profitable at the end of the playoffs just making bets with the principles of advanced stats in mind.

 

[rosemount289]

Can you give us...............???

Just create better power rankings than Vegas does (not that difficult) and you'll see where to take advantage.

A good example is where they ranked the Kings/Rangers before the playoffs started.

LA was 14-1 and NYR 26-1. They were both the most underrated in their respective conference, and anyone paying attention to advanced stats this year would've seen the glaring discrepancy. Advanced stats don't make up all of a team's power ranking, but it's a big chunk.

During the regular season, there are opportunities on a nightly basis. Vegas is kinda dumb when it comes to hockey.

Can you give us a detail example?.......cause I still don't know what you are talking about?

 

[BB1133]

Can you give us a detail example?.......cause I still don't know what you are talking about?

Not sure what kind of detail you're looking for.

Basically Vegas creates odds on who will win the cup, and prices on each game/series, etc. They base it on their own internal power rankings, money wagered, home/road, injuries, public perception, and so on.

No one has a more vested interest in ranking teams accurately than Vegas. Be wrong on a team, and they pay for it. Fans and "analysts" can be wrong all day everyday with their predictions and nobody cares.

The problem for Vegas is they don't incorporate advanced stats into their evaluations of teams, so there is a pretty sizable mismatch between the prices they offer and the real price it should be. At one point in February/March they actually had the Leafs at 20-1 to win the cup, which is absurd.

The other major factor is that hockey is just a niche sport for Vegas and there isn't a whole lot of $$ being wagered on it, so their incentive to make accurate prices is low compared to other major sports like soccer, NFL, NBA, MLB etc. You can be sure that Vegas is aware of advanced stats/sabermetrics in those sports.

But advanced stats for hockey are relatively new and still developing, so the opportunity to make a lot of money is there. Sooner or later Vegas will catch on, but they haven't yet.

It's so easy to make money that even casual bettors can make money on basic advanced stats...

@TravisHeHateMe

Is it a bad time to remind everyone that Score-Adjusted Fenwick% last twenty is 9-3 picking playoff winners this year, 70-31 since '07?

 

[Bear of Bad News]

The problem for Vegas is they don't incorporate advanced stats into their evaluations of teams, so there is a pretty sizable mismatch between the prices they offer and the real price it should be. At one point in February/March they actually had the Leafs at 20-1 to win the cup, which is absurd.

A lot of that is because Vegas primarily cares about bettors' expectations, and less about what the actual odds are (of course, in many sports these align closely).

If the true Toronto odds were 50-1, but 20-1 balances the books (*), then it's going to show as 20-1.

(*) Hence the name "bookie".

 

[BB1133]

A lot of that is because Vegas primarily cares about bettors' expectations, and less about what the actual odds are (of course, in many sports these align closely).

If the true Toronto odds were 50-1, but 20-1 balances the books (*), then it's going to show as 20-1.

(*) Hence the name "bookie".

Yes and no. That Toronto line was definitely set with squares in mind, but for the most part they have to contend with sharp money on opening lines. That is where most of the discrepancies are.

 

[Bear of Bad News]

Yes and no. That Toronto line was definitely set with squares in mind, but for the most part they have to contend with sharp money on opening lines. That is where most of the discrepancies are.

Well, yes and yes. Part of balancing the books is understanding what you just described.

 

[BB1133]

Well, yes and yes. Part of balancing the books is understanding what you just described.

Yea, they make enough back on squares that they profit regardless. I imagine that the public will become more savvy in the next year or two as advanced stats become more and more mainstream. Next year will be interesting..

 

[nobuddy]

It's no wonder that the SCF is the #1 FenClose team against the #6 FenClose team, and that all the teams they beat on the road to the SCF were ranked behind them, I'll just say that. Predicting based on Fenwick would have given a fantastic bracket this year with the exception of Montreal.

Of course, this one postseason is a small sample size, though.

 

[Thirty One]

It's no wonder that the SCF is the #1 FenClose team against the #6 FenClose team, and that all the teams they beat on the road to the SCF were ranked behind them, I'll just say that. Predicting based on Fenwick would have given a fantastic bracket this year with the exception of Montreal.

Of course, this one postseason is a small sample size, though.

Long-term it's done well, too.

 

[Rebuilt]

If they had set the Toronto numbers at 10-1 they would have made more money. Most fans see the odds and back away from long shots. Leafs have a huge fanbase and putting it at 10-1 is actually stroking the Leaf fans egos.

 

[OCSportsfan]

I am not a 100% supporter of Advanced stats, but do think there is some validity to them, just like some other stats.

What i agree with
1) Over the long run if you are able to take more shots, you have more opportunity to score, and should be in possession of the puck for more time. You should have a better chance to win.
2) Zone starts and finish really show the good players in my opinion when taking shots into account.

What I don't agree with
1) Shot percentage is different with every team, and a high shot percentage can be sustainable, just like a low one. If for example 75% of your shots come from Forwards within 20 feet of the goal, you will have a higher percentage than if 50% of your shots come from Defenders throwing the puck at the net from just inside the blue line. If you could weigh the shots depending on the distance from the goal (or quality of player), it would help.
2) Save percentage is less sustainable, but to pretend every goalie and defense should have the same save percentage is just silly. There are better goalies and better defenses that prevent the high percentage shots.

In a few years, I am sure there will be some technology out there that will take this advanced stats to another level, but at the moment I just don't think it is as accurate as people hope. It is a Chicken and the Egg, are the Kings good because they take a lot of shots and don't give up many, or are they good because they are the deep down the middle and have a great Defense, and as a result have the puck most of the game.

I bet the Kings to win the SC because of how deep of a team they were (matching player to player), not because they out shot the other teams.