RTSS stats: Giveaways, Takeaways, and Throwaways

Do Real Time Scoring System stats hold any value? I have followed them with interest for some years, mostly because they are there and do provide some context to different sorts of contributions players make. There has never been much doubt that the recording of said stats is deeply flawed, yet an actual count of things like Hits is surely better than simple anecdotal "seen him good" accounts. Many folks value Missed Shots and Blocked Shots for their contribution in fleshing out shots data into Corsi stats, which are a pretty good proxy for zone/possession time. Of course they too are subject to recording error and subjectivity bias of the scorer, as indeed are shots themselves.

Perhaps the canaries in the RTSS coal mine are the paired stats of Giveaways and Takeaways. These are opposite sides of the same coin, a turnover which is attributed to a negative action of a puck carrier OR a positive action of an opposing checker. Yesterday I was tooling around in this section of NHL.com and made some interesting observations. I was specifically looking at Oiler forwards with 20+ GP, and in the course of my investigation compared them against their peers in the Northwest Division. My methodology was to express a ratio of takeaways to giveaways (in that order, so that higher number = better). To my surprise I found an apparent correlation between Gv:Tk and the Northwest Division standings!  

(The last column, Pts% over league, is simply the ratio between a team's Pts% and the NHL mean, currently hovering around .559. A number of 1.00 in this column is the closest thing we have to the old standard of .500.)

Hey lookit! The highlighted columns are remarkably similar, ranking the teams in the correct order. The teams with the best success in giveaways v. takeaways are also enjoying the best success in the standings! Such a nice correlation; could this possibly be a coincidence?

In a word, yes. In this case the stats do co-align with reality, but it's a pure fluke of this particular (small) sample. Examining the team stats across the NHL reveals that the standard the NHL has seemingly established in counting these events is actually No Standard At All.

The NHL team leaders in Tk:Gv ratio are NYR, ATL, CHI, CBJ, CAR, and NYI. 5 of the top 6 are in the bottom third of the NHL. In other words, epic fail.

Particularly striking when looking at the these numbers on a league-wide basis is the enormous discrepancy in home/road splits. The scorers are not necessarily more "generous" with the home town team, but they are far more likely to notice them and attribute actions on the ice to them, be they positive or negative. League-wide in 2009-10, home teams have been credited with over 37% (!) more takeaways than visitors, and debited with nearly 42% (!!) more giveaways.

Moreover, some rinks simply count way more total events than others. The number of total events (Gv + Tk) for the home team ranges from a modest 281 in Phoenix to 1098 on Long Island, very nearly a factor of 4! On the road the range, as one might expect, is a much more constrained 390 to 552. This supports the theory of many, myself included, that RTSS road stats are much more valuable than full-season totals. They're a long way from perfect: some teams play three road games on the Island, others none (including the Islanders!), but discrepancies are far more likely to balance out.

The Islanders lead the entire NHL with 857 takeaways, a cool 500 ahead of the Stanley Cup champions from Pittsburgh who rank a rank 29th in this "important" defensive category. 99% of their advantage comes in the oh-so-friendly confines, where the Isles have recorded 624 of those takeaways, nearly 5 times as many as the Penguins' 129 on home ice; on the road, the Islanders' edge is an inconsequential 233-228.

The league leaders in giveaways are none other than your Edmonton Oilers, who have been charged with a scandalous 972 such turnovers, nearly triple the 359 charged to the Blue Jackets. Not surprisingly, most of this discrepancy can be found on home ice, where trigger-happy statisticians have charged the Oil with 737 giveaways, compared to just 235 on the road. They are also more generous with takeaways although in moderation; the Oilers have 307 TkA on home ice, 196 on the road. In all the Oil have recorded 76% of their giveaways and 61% of their takeaways on home ice.

Home/road splits are not readily available on an individual level; while they can be scraped from game sheets, NHL.com does not show them. So I concluded that any comps of individual Oilers against their Northwest foes would be an entirely useless exercise.

Are RTSS numbers valueless? Not quite. Players within an individual team will play their games in front of the exact same scorers from night to night, so in-team comps should be fairly valid. Let's take a closer look at those Oilers with 20+ GP, ranked in order of net turnover differential per 60 minutes of ice time (the right-most column):

Lo and behold, there's good old Shawn Horcoff at the top of the charts, challenged only by the up-and-coming two-way forward Marc Pouliot. Reliable Fernando Pisani ranks third, followed by Zack Stortini, whose low-event proclivities extend to having the fewest giveaways per hour on the squad. In the middle of the pack Ryan Stone has the highest events per hour in both categories, albeit in the fewest minutes of anybody on the list; he's good at getting the puck, not so good at hanging on to it.

At the other end of the chart are a couple guys in Ales Hemsky and Sam Gagner who handle the puck a lot and therefore have their share of giveaways, although both are disappointly weak on the defensive side of the puck, especially Gagner. Lagging behind them is J-F Jacques, as those who were subjected to watching him play might expect. Bringing up the very rear, though, is thecaptainethanmoreau, who is second worst on the club for giveaways per hour and dead last for takeaways. A checking line player who can't check? Ouch.

Conclusion: Takeaway/giveaway data is suspect at best, but has some validity in the context of a single team. With the exception of a couple of small sample size guys (notably Hemsky, Stone, possibly Pouliot), the above list of forwards does a reasonable job of identifying the guys who collect and protect the puck, and those who don't. Here's looking at you, Ethan.

* * *
EDIT: Here's the same list for defencemen. This includes the three guys who were traded at the deadline (shown in grey); the data supports some of the rationale for dumping at least two of those guys. I have lowered the threshold to 10+ GP so as to include early results of the new guys, although of course sample size is a particular issue.

On the blue giveaways are far more common than takeaways, to the point that a Tk:Gv ratio better than 0.5 is relatively decent, at least on this team. A little surprising to see Smid at the top, but he handles the puck both less often and more conservatively than some of the puck movers below him so has coughed the puck relatively infrequently. So far Johnson seems to be the Ryan Stone of the defence corps with high events per hour on both sides of the ledger; alarmingly so in the giveaway department, although it's a pleasant surprise to see anybody even close to, let alone ahead of, Sheldon Souray in the takeaway game. Early days of course, just 12 GP. Interesting that these two rank 1-2 on the club in both Gv and Tk, and that by eye both fit the mold of cycle-breaking "chaos" defenders.