corsi – McKeen's Hockey

Rested/Tired B2B: The Differentials

Gus Katsaros — Fri, 31 Jan 2014 21:33:08 +0000

I’ve been looking at the rested/tired scenario for years, but I like this approach Tyler Dellow introduced using Corsi ratings after various periods of rest. The analysis is housed here. I like how it provides intricate insight as to what really transpired during those games instead of a straight end result as a win/loss as a factor.

Dellow broke games down to the amount of days between games (rested). I wanted to get into the specific of how tired teams and rested teams actually performed in back-to-back sets.

To do this I needed to pull some data with a large enough sample to gauge true conclusions, and eliminating short term random fluctuations associated with games in a short periods of time.

I used timeonice.com to retrieve individual game data, running a script to obtain Corsi/Fenwick ratings from games ranges from 2007-08 to 2011-12.

I kept this current season out of the analysis due to the partial results and excluded the lockout shortened 2012-13 season due to restrictions of playing only within the Conference, which I feel takes away from some of the scheduling effects in some of the more high volume areas, like Anaheim-Los Angeles, teams in Alberta and Florida, stretching through to the New York area where any combination of a NY Rangers – NY Islanders – New Jersey Devils travel schedule can invoke a rested/tired scenario.

All in all, approximately 1620 games of data available which is a significant sample to normalize as will be shown below.

Using timeonice.com I pulled the data for 5v5 game situations, returning Corsi/Fenwick events and percentages at 5v5 excluding any empty net situations. I then ran the numbers for games ‘Close’ defined as games up or down by a goal in the first two periods, or tied in the third period to determine a true review of game situations reflected when games are close and overall.

A great resource for how to use timeonice.com is provided by Red Line Station.

My definition of what constitutes a rested or tired team, I’ll refer you to one of the many posts I’ve written about this subject, including Alberta’s little secret, housing a variety of links for a proper historical look using the available data of the time.

The definitions are below.

Rested – team that has not played the previous night facing a team in the second night of a back-to-back set on consecutive nights.
Tired – a team playing its second game on consecutive nights versus a team that is rested and not played the previous night.

I liked Dellow’s approach for the meaningful method in which it portrays the minutiae of the game situations and confirm/denies common knowledge of the expectations for each team in each situation. I lacked the data to provide as insightful analysis until running the timeonice scripts.

Generally, teams play to a .596 winning percentage as a rested team, and to .495 as a tired team

Game Type	Games	Wins	Win%
Total GP	17220	8607	0.558
Non B2B GP	14168	7222	0.569
B2B both teams	794	397	0.552
Rested B2B	2258	1270	0.596
Tired B2B	2258	988	0.495

Let’s look at each situation for tired and rested teams. The tables below have the following legend:

LEGEND

AvFF% - 5v5 Average Fenwick For %

AVFFclose – 5v5 Average Fenwick For % close

AvCF% - 5v5 Corsi For %

AvCFclose – 5v5 Average Corsi For % close

Av OI EV SV% - Average on-ice even strength save %

Av OI EVSH% - Average on-ice even strength shooting %

Av OIcl EVSV% - Average on-ice (close) even strength save %

Av OIcl EVSH% - Average on-ice (close) even strength shooting %

TIRED

The first look is at how teams fared as a tired team. Expectations would naturally be a tired team to be competitive from the start, but likely in association with an average .495 winning percentage, a Corsi rating under 50%. Here’s the table containing the results.

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.481	0.481	0.482	0.482	0.908	0.917	8.756	8.128

Results fall into expected ranges, with sub 50% all round, with very little differentiation between game results overall or close. This changes when filtering out game decisions.

Below are results of tired teams but only in Wins:

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.512	0.478	0.507	0.473	0.953	0.946	13.645	10.969

Judging by the data, it seems like teams come out with strong starts posting 51.2% FF% in close situations, and then likely having score effects determine the drop to overall FF%. The dramatic drop could be the combination of score effects and perhaps fatigue setting in, especially true for teams that had to travel immediately after the game the previous night and on to the next city, which is a more likely scenario than teams playing back-to-backs at home for any prolonged period of games.

Goaltending played an important part of generating wins in these situations, as did some timely shooting with a shooting percentage north of 10% and 13.6% in close situations.

Let’s look at the situation in losses. At this point, select patterns emerge with expectations being a mirror opposite of the situation in Wins.

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.457	0.484	0.463	0.489	0.872	0.894	4.896	5.890

Typically what would be expected in a loss is indicative in the data. Not only is goaltending affected with a sub .900 save percentage, along with accompanying deflated shooting percentage, but the close results indicate that goals weighed heavily on the rest of the contest as the rested team scored enough to engage score effects.

Possession stats follow suit. The indications are that with the game close, teams may be falling behind as indicated by the 45.7% FF% and 46.3% CF% after score effects kick in, boosting overall FF% and CF%.

As expected the data certainly supports the typical expectations of teams playing tired against a rested opponent on the back end of a B2B set.

RESTED

Once again the mirror effects are expected here as the tired team, and once again the data confirms the impact a rested team has playing a tired club, right down to the score effects. The first table encompasses a rested team overall regardless of game situation/result.

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.521	0.519	0.520	0.518	0.914	0.920	8.774	8.281

Data supports the probability. A rested team that seems to go up in score early and actively presses the play. Once attaining a lead, score effects take hold with an end result to decrease the overall FF% and CF%.

On-ice stats support these conclusions. The overall shooting percentage drops as getting pucks to the net and killing time take precedent over generating scoring chances, with the corresponding even-strength save percentage increasing due to the amount of shots generated by the trailing team looking to get back in the game, tie the score or putting the effort expected from coaching staff in situations that weight teams against winning the game outright.

Let’s break it down by game result/situation starting with wins.

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.549	0.516	0.542	0.510	0.959	0.946	12.878	10.915

Rested teams could take advantage early, scoring at an above average pace indicative of an inflated on-ice shooting percentage in games close. Couple that outcome with a puffed up save percentage and the quick start seems to signify the doom of a tired team, ending in an eventual loss.

The numbers fall back as the game overall, with score effects, lowers the overall average FF% and average CF% from the exorbitant high of close situations.

Losses also follow the typically anticipated results.

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.487	0.523	0.492	0.529	0.858	0.887	3.857	5.055

This is actually very interesting and led me to go back to recheck the data. Once a quality control was placed to ensure the data was indeed correct (it was), the next step was to break down the actual losses by extra time games played. In all, 742 games were decided in extra time, 288 in overtime and 454 in a shootout.

Score effects are clearly at play here, but the gap from Av FF% close to the overall 5v5 result is substantial at slightly less than 4%.

Let’s see the effect by isolating extra time games.

EXTRA TIME GAMES

The tables above incorporate extra time games without differentiation to shootout or overtime in overall losses.

Removing extra time games alters the end result of isolated losses with higher values as a result; in particular CF% jumps to 53.6% while the FF% stays fairly close to the original 52.9% as rested losses.

This leads me to believe a factor influencing the increase lies in the increased blocked shots boosting the rate stat slightly higher with FF% essentially similar.

AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
0.481	0.529	0.488	0.536	0.839	0.877	3.560	4.760

Extra time games results without the isolation between overtime and shootouts reveal consistency in results for both rested/tired teams.

	AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
R	0.504	0.507	0.504	0.509	0.917	0.917	4.774	5.943
T	0.476	0.473	0.478	0.475	0.926	0.925	0.063	0.072

Uniformity changes when introducing the separation of results from overtime or shootouts.

Rested extra time

	AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
SO	0.519	0.519	0.520	0.520	0.940	0.932	5.166	6.143
OT	0.511	0.517	0.508	0.515	0.917	0.915	7.220	7.429

Tired extra time

	AvFFclose	AvFF%	AvCFclose	AvCF%	Av OIcl EVSV%	Av OI EVSV%	Av OIcl EVSH%	Av OI EVSH%
SO	0.484	0.485	0.484	0.485	0.944	0.936	6.767	7.280
OT	0.488	0.484	0.492	0.487	0.921	0.921	8.357	8.436

With the above information providing the norms NHL-wide in various situations, the next step would be to gauge team’s performances relative to the norm.

That’s the next post.

**********

Follow the McKeen's team on Twitter:

@mckeenshockey
@KatsHockey

Senators Bruins and Leafs .. Oh My

Gus Katsaros — Wed, 17 Apr 2013 19:42:36 +0000

A short post today, just to show some of the contrast in playoff teams, in particular the difference between three teams, all from the Northeast division. Boston, Ottawa and Toronto all offer a different game-by-game corsi breakdown that offers an eye-opening glimpse into each team.

All three clubs are likely to secure a playoff spot soon, each having it’s own dynamic. The Bruins have been the class of the East since their Stanley Cup win. The Senators have been decimated by some devastating injuries, yet remain in the playoff hunt. Their game-by-game chart, however, doesn’t look remotely like a team decimated by injury. Without any prior knowledge of the club, just by looking at their image, one couldn’t quickly identify them as Ottawa.

The Leafs, on the other hand, are an interesting study and a different animal all together. I'm definitely not the first to point this phenomenon out as anyone following using advanced analytics can easily spot this problem.

OTTAWA SENATORS

This is the game by game Corsi breakdown for the Senators. Keep in mind the length of injuries to some key players among others and including the bumps and bruises from the group remaining in the lineup. Jason Spezza, Erik Karlsson and Craig Anderson, Milan Michalek, all key players that missed time, but the chart doesn’t seem to reflect that.

Here’s the explanation:

*** Note the numbers on the horizontal are the game numbers (I haven’t figured a way to capture the opponent yet, but that will come).

A quick reminder, Corsi is presented as a ratio, and is the differential of (shots, missed shots, blocked shots) directed at the opposition net minus the same criteria against your own net at even strength. Positive values indicate more pucks were directed at the opposition goal than at your own goal.

This chart below represents the calculated differential as part of the overall make up of their Corsi.

In other words, this represents the makeup of the individual components on a game by game basis.

When taking the injuries into consideration, it’s incredible that the Senators have as many positive Corsi games.

In this next image, the values above are represented as a percentage of their individual components.

BOSTON BRUINS

Let’s contrast that with the Boston Bruins.

The B’s should be considered an elite team in the East with a solid crease and blueline and depth up front. They have struggled lately. The image below captures that slumping over the past six games.

Boston hasn’t exactly been a powerhouse, but they are usually on the positive side of the ledger. At game 20543, it looks like the trouble started.

They were being outshot as per differential and at the same time, the differential of blocked shots is the lone positive component.

My interpretation would be that as the B’s are sending a lot of pucks in to traffic rather than through seams.

The fact they have a negative differential on missed shots is another indication that they likely aren’t missing a lot of shots, in conjunction with competition missing a greater number of shots on goal and pulling the differential down.

The percentage breakdown by component mirrors the six-game slump shows that most of the Corsi makeup has been dominated by the shots differential on the negative side, and the shot blocking component as the only positive. Boston has to get more pucks on net.

TORONTO MAPLE LEAFS

Let’s take it a step further with the Leafs.

This is the fifth overall team in the East Conference less than a handful of points from the division lead.

A closer look at their chart shows a clear difference between the Senators and Bruins. This is a team that should, for all sense and purposes be missing the playoffs, and considered a bottom feeder.

This is a telling image, especially when comparing to the images above.

As a side note, the last two games against the New Jersey Devils and the Washington Capitals on back to back nights as captured by this image is telling. Against the Devils (game number 20625) the Leafs clearly were outshot, had more shots miss the net and required more blocked shots, yet they came away with a 2-0 win.

In Washington, the raw numbers showed less of a differential, yet the Buds were blown out 5-1.

The percentage image also mirrors this breakdown.

While the Leafs have been on a roll and have likely secured a playoff spot, can bask in the achievement of finally making the playoffs for the first time since the NHL stopped recording tie games, but this breakdown shows a chink in the armor that could be fatal come playoff time.

Data is courtesy timeonice.com

Follow the McKeen's team on Twitter:
@KatsHockey
@mckeenshockey