36
Photo Credit: Sergei Belski-USA TODAY Sports

Let’s Go Streaking With Ryan Strome!

Strome is back! Two years x $3.1M.

Given the negotiator, that’s a decent number given Strome’s qualifying offer ($3M) and arbitration rights. These days, anything that isn’t a big overpay is a win in Oilersland.

That said, I have a lot of time for Strome. He’s a legitimate NHL player, and his signing helps clarify and solidify the third line while adding a player who has the skill to … on occasion … play higher in the lineup.

Hot Flashes

One of the interesting characteristics of Strome, though, is that he goes through these flashes that make you think he’s a much better player than he’s shown himself to be. (Arguably, an extended near-season-long streak of that sort fooled the Oilers into thinking Strome was on par with Eberle as a scorer).

It’s kind of funny – I first drafted this article when Strome had just had a really good run, with five goals in five games, leading some to declare that he’d finally arrived as a player. Then he went cold like Newfoundland in June, and the newsworthiness of the article fell. So I put it on hold.

A few weeks later, Strome once again heated up and scored five points in just three games, leading many at that time to similarly ask whether Ryan Strome has finally turned the corner. I started updating the article again.

And then he went cold.

Again.

Most recently, I heard a lot of people react to the contract signing by saying things like Strome had “found his game” and “arrived” late in the season. As I’ll show you later, that’s probably not the case. It’s more likely just a(nother) streak.

The season is over and Strome is signed, so dammit, let’s finish this analysis. Is Strome actually streaky? And if so, just how streaky is Ryan Strome?

Stop Making Sense

Embed from Getty Images

How streaky is Strome? It’s a surprisingly challenging question to answer.

The reason it’s challenging is that every player has hot and cold streaks, and the nature of those streaks is extremely hard to distinguish from randomness. In fact, for years statisticians argued that “the hot hand” in sports didn’t exist. Just like flipping a coin repeatedly inevitably leads to significant streaks of heads or tails, the points streaks we saw were borne entirely out of randomness.

That idea has been falling out of favour the last while, which makes sense to me – as much as we’re prone to being fooled by randomness, I don’t for a second believe that streaks and trends in human behaviour and performance can be modelled by memoryless coin flips. Coins don’t remember what happened last time, but we do, and it does affect us.

So rather than run run run away … oh oh oh … let’s think about how we might answer the streakiness question.

My plan here isn’t to write an academic paper, but just make a reasonable, understandable, and analytically light-but-sound argument as to the extent of Strome’s streakiness (if you’re really interested in the hardcore stats stuff, here’s a rigorous approach to testing for streakiness, knock yourself out!)

Mr. Connor Sistent

Let’s tackle the problem by starting with the other side of the coin – by looking at the steadiest player we can.

Suppose we had a player that scored just shy of the pace of Ryan Strome over the last couple of seasons (so about 0.4 points per game), but was absolutely not streaky at all. We’ll call this player Connor, last name Sistent.

What I’m going to do is ask you to think about what Connor’s point scoring would look like over time if you put it on a chart.

Got it?

OK, here’s a visualization of what it pretty much *has* to look like:

Given his total lack of streakiness, Connor’s 0.40 ppg means he’s getting a point every two to three games. In order for this to happen in a non-streaky way, we’d expect to see a couple of things:

  • First, as we watchpoints as they occur in games, we’d see a consistent one-on one-off one-on two-off kind of pattern. If the runs of points or not-points get longer than that, we’d accuse Connor of being streaky.
  • For that same reason, we’d expect to see few or no games where Connor scores more than one point … after all, if he puts up a two-point game, that means as a 0.4 ppg player, he’d then have to go four or five games without a point to compensate, and then he’s streaking again.

The chart above, from left to right, shows the games that Connor has played. Every time he scores points, we add a bar for that game showing how many points he got. No points, leave that spot blank. A player with no points would have an empty chart. A player who has points every game would have a solid blue mass.

Make sense?

So what we see with Connor’s game is nice, even, steady progress. There are no streaks, which would show as long periods of either blue space or blank space.

In addition to this game track type of chart, let’s count a few key ‘streaky’ items about Connor’s game history as well:

  • his longest streak of games where he scores no points
  • his longest streak of games where he scores at least a point
  • the number of games where he scores more than a single point

The more streaky the player, the higher these numbers will be. Less streaky players will have low(er) numbers. In chart form, Sistent’s chart looks like this:

As with the previous game track chart, this is pretty much what we’d expect. There are no long streaks, and no multi-point games. Connor Sistent is nothing if not … consistent.

Ryan “Yo Yo” Strome

Now, let’s run the same pair of visualizations, but this time for Ryan Strome. (Note: data sourced from hockey-reference.com). Here’s the game track:

Some difference, huh! Visually, lots of runs of blue and lots of gaps suggest Con and Ryan are distant relatives at best.

Of particular interest is the last section of the chart, where you can see a loooooong cold streak, followed by a fairly (for Strome) typical burst of points. That means if you think Strome “arrived” or “found his game” late in the season, you’re almost certainly being fooled by recency bias layered over Strome’s streakiness.

Now here’s the streaky chart:

We’ve got significant hot streaks (up to five straight games with points), some nasty cold streaks (up to 11 games, ouch), and a healthy dose of multi-point games … which is kind of what you’d expect given the length of the cold streaks.

Given that Con Sistent is so consistent, Ryan Strome might want to change his name to Riley Streakye.

Comparisons

But wait! I am forced to point out a fatal flaw in my own argument.

Sure – Strome is streaky compared to Con Sistent. But Con Sistent isn’t a living being. Every player is going to look like Flake City compared to him. What does Strome look like against peers who actually, you know, exist?

To answer that question, I pulled a list of players who’ve scored at a very similar pace (.42 to .44 points per game) and played a similar number of games (about 150) the last two seasons. The screen returned: Zach Hyman, Radek Faksa, Michael Ferland, and Boone Jenner.

Here’s how it looks when you directly compare the game tracks of those players with Strome:

-.– -.– –..

Embed from Getty Images

So in the context of living beings, the differences still kind of jump out at you.  With the exception of recently traded Michael Ferland (who is in the ballpark), Strome’s combination of bursts and empty spaces is quite noticeable, like he’s communicating to us in Morse Code. “Help me, Mark Messier has my cell number”.

And here’s how the streak tables look in comparison:

As with the game tracks, comparatively you can see that out of this peer group, Strome ranges from a tiny bit to noticeably more streaky. The main thing is that he has more multi-point games than anyone else. Remember – all of these guys score at very similar levels.

That Strome runs unusually hot when he’s on a scoring jag is exactly why people keep getting fooled!

Embed from Getty Images

The bottom line as far as Strome is concerned remains the same: yeah, we can be reasonably sure that he’s streaky. The difference isn’t extreme the way it is when compared to the legen (wait for it) dairy perfectly unstreaky player, but Strome is still visibly less consistent than several of his peers.

In other words, now we have a little bit of analytical confirmation – the runs of good play from Strome are not necessarily him “turning the corner”, “arriving”, or “finding his game”. More likely it’s just another hot streak from a player who has been consistently inconsistent and inconsistently consistent over the past two seasons.

It doesn’t mean he’s not a useful player. In fact, if you look at his comparables (in addition to the four I used above, you could roll in other useful NHL players like Ryan Hartman and Nick Bonino Bonino Bonino), it should be clear that Strome is a legitimate NHL player, better than many give him credit for.

And we should not hold the price Chiarelli paid to get him against Strome (or any player for that matter). The trade is done, it’s a sunk cost. Strome is a useful player nonetheless.

Just remember … now that he’s remaining an Oiler, we pretty much have to accept that we’re gonna go streaking with Ryan Strome.

  • AlexTheOilersFanSince2006

    I don’t mind Strome at all tbh. Other people don’t like him because of his stats, be he slotted into the 3C position nicely and I like that’s he back.

  • OriginalPouzar

    Strome was solid, not spectacular, in the 3C position. He somewhat solidified the line and played well with a few players such as Khaira and Puljujarvi.

    He also performed adequately on the PK and, I don’t think its a coincidence that the PK’s uptick last year coincided with Caggulia and Letestu being replaced by Khaira and Strome (not to mention McLellan taking over the PK in the 3rd week of January and reverting to a more standard “box”).

    I also think Strome has more offence to give – he is an underrated puck distributor in the offensive zone – I believe he has nice vision and has made some wonderful cross-ice passes that, unfortunately, have died on the stick of the recipient.

    Strome absolutely has the ability to play up the lineup, the numbers show that he did well with both McDavid and Drai last year and, in fact, both McDavid and Drai has better numbers with Strome than without (McDavid sample size likely too small – apx 50 minutes).

    He may be best suited to top 6 RW but I do think he is now pigeon-holed as the 3C as long as Nuge plays 1LW.

    Moving up when they shorten the bench should be an explorable option (and Brodziak showed in a fantastic 16 games with Berglund and Steen last year after Stastny was traded that he can play up the lineup with skill – at least in stretches).

  • VK63

    Interesting piece G. I dig it!!
    Of note, in the bibliography for the hot hand chapter in “Scorecasting” they (Moskowitz and Wersheim) reference this work.

    https://link.springer.com/content/pdf/10.3758%2FBF03206327.pdf

    I found it to be a solid effort on the popular fallacies, Hot Hand and Gamblers. They approach from a more behavioural economics slant but it is a decent contextual read when considering this streakiness thing.

    • Thanks V!

      Yeah, Scorecasting is a very interesting book. That and “Mathletics” were two of the first sports-specific books I read back when I first started dabbling in analytics for sports.

      One of the things I like to point out is that PDO (which has now become ‘controversial’ as a statistic mostly because people don’t understand what it is or how it should be used) is an indicator for avoiding the Hot Hand Fallacy.

  • hummanahummana

    I like the comparison charts. I always compared Strome to Boone Jenner which is why I am happy with Strome’s contract. He seems to have found his niche as a more talented Mark Letestu which I am totally okay with. Just needs to improve his FO% now! Khaira-Strome-PJ should start the season together.

  • Bills Bills

    Stop making sense? How about stop with the same old rhetoric? I stopped reading after 3 eye rolls and the part about the Oilers being fooled into believing Strome was on par with Eberle.

    Then I scrolled down and couldn’t believe how long you wanted to listen to yourself talk.

      • My Twitter timeline was packed with people who felt that Strome in future would deliver on his one-season peak in scoring i.e. near Eberle levels, rather than at the inconsistently consistent level that he’s shown over the entirety of his career.

        That the Oilers “brain”trust would comment that they were disappointed in Strome’s season when he delivered at exactly that demonstrated career level suggests that they thought along the same lines. You can quibble if that expected level was “Eberle”, “near Eberle”, “mostly Eberle”, whatever.

        It tells you a lot about both the Kool Aid people drink as well as the incompetence of the management team that they expressed disappointment.

    • This isn’t a puffy Oilers Rumours article, nor will you ever get those from me.

      ON asked me to write for them specifically to add to the stats/analytics base of their overall ‘portfolio’ of writing.

      I avoid diving into overly analytical work here, as that’s only going to be of interest to a hardcore few. But I do spend my time trying to find approaches and topics that aren’t covered elsewhere, and then make what can be heavily math-oriented and complicated into something more straightforward and digestible.

      Even simplified, those concepts I do write about here can rarely be walked through in a paragraph.

      All of that isn’t everyone’s cup of tea, and that’s perfectly OK.

      Suggestion: next time, just put “it was too complicated for me”, or better yet, just put tl;dr. We’ll understand.

      • Bills Bills

        Awesome retort dude.

        “Given the negotiator”
        Eye roll number 1
        “anything that isn’t a big overpay is a win in Oilersland”
        Eye roll number 2
        “fooled the Oilers into thinking Strome was on par with Eberle as a scorer”
        Eye roll number 3

        If you have taken to reading many of the bloggers on this site, you will soon realize that these little quips are constant among the non mainstream guys. It is a weak fall back and the moment you start your post with the exact same rhetoric as the others on this site, you are painting yourself with the same brush. I didn’t do that.

        As for your attempt to insult my intelligence by suggesting that I didn’t understand your post. Guess what genius, when I said I stopped reading, I did. The only thing that would have truly insulted my intelligence, would have been to continue to entertain your opinions.

        Maybe you are smart and have valid points, but unless you own the fact that your little pot shots are nothing but that. They will continue to detract from what could actually seperate you from the rest.
        Have a great day.

      • PJP

        Sir, I really enjoyed your article and I welcome more of your contributions.

        Kudos to ON for adding new voices to the site.

        BTW: really Bills?…”your article sux so bad I didn’t read it”…. But you engage with the author?

  • Redbird62

    An interesting perspective which shows that Strome’s outcome’s have been fairly streaky at least compared to the peer group. A lot of other factors would have to be considered though to draw the conclusion that all things being equal, Strome is streaky player compared to others. Would the conclusions be the same, if you could factor in by game the total time on ice, power play time, linemates, strength of competition position played et cetera? Looking at most of your comparables on Corsica.hockey, they all had significantly more consistent linemates than Strome did over the past 2 seasons. I have to believe that is a major contributor to Strome’s perceived streakiness. Hyman played almost 1200 minutes with Matthews and Nylander and Ferland played almost 1000 minutes with Gaudreau and Monahan. Strome’s longest full line combination in the last 2 season were Beauvillier and Nelson with NYI at 160 minutes; with the Oilers, he got 106 minutes on a line with Khaira and Draisaitl and every other complete line combination was less than 100 minutes.

  • aluchko

    It’s important to note that Mr Connon Sistent is by far the streakiest player in the sample.

    Every two or three games, he gets absolutely red-hot, absolutely guaranteed to score. And then the moment he ends up on the score sheet he goes ice-cold for 2-3 games. You could give him a breakaway with an empty net and he still wouldn’t score.

    A player who isn’t streaky, who has about the same chance of scoring every time he’s on the ice, is going to have a lot of streaks based purely on luck. Sometimes he’ll go a few games without getting lucky, and sometimes he’ll get lucky a few times in one game.

    I suspect a player who scores ~0.5 PPG is going to look the streakiest. Someone getting 1 PPG needs to be really unlucky to go without points for an extended period, and someone at 0.1 PPG is rarely going to get scoring streaks. But at 0.5ppg you can get an extended streak in either direction.

    • Hmm, I’m trying to parse what you’re saying, but I cannot. Con Sistent does not go hot then cold, he scores then doesn’t score for a game or two, in a perfectly even pattern for the level of his scoring, which is approximately one point every 2.5 games.

      You can have streakiness at any level of scoring.

      As far as I can tell, what seems to be the issue is that you aren’t accounting for the idea that you can have multi-point games. So a perfectly consistent 1 ppg scorer would score one point every game. A super streaky 1 ppg scorer would have three or four multi-point games in a row where he scores say 12 points, and then go silent for the next ~12 to 15 games.

      That’s the thing about Strome – the multi-point games that make you think he’s a better scorer than in actuality – that is almost the defining aspect of his streakiness.

      • aluchko

        I ignored multi-point games but it doesn’t really effect the math.

        Think about what Con Sistent actually does. If he didn’t score the previous 2 games, and it’s 59 minutes into the 3rd period of the 3rd game, then he’s absolutely guaranteed to score in the next minute because he has to maintain that pattern.

        That means Con Sistent can literally score at will, it’s just that he only chooses to do so every 2.5 games.

        I think this matters in practice. Consider a player who only averages 0.5 ppg but never goes more than 2 games without scoring during the season. At a glance he seems very consistent, but if his true talent level was 0.5ppg then the chance of a 3 game goalless streak is 0.5*0.5*0.5 = 12.5%, he should have a few 3 game droughts during the season.

        To go the season without any droughts, well what’s probably happening is gets a point, relaxes for a game, then realizes he hasn’t scored in a bit so starts playing like a superstar until he scores again. So that seemingly “consistent” player is actually hiding a ton of potential.

        Think of a coin, perfectly unbiased and it never goes hot or cold. But if you start flipping a coin and get H-T-H-T-H-T-H-T-H-T-H-T you know something is up because a random coin doesn’t do that.

        • Ooof, no, not at all. The fictional “Con” doesn’t score that way by choice, he ends up scoring that way because he’s just so damn consistent. That’s the whole point. And ignoring multi-point games doesn’t just affect the math, it pretty much defines the math.

          I’ll assume that it’s not something you need to ‘prove’ that pretty much every player in the NHL scores as often as they can.

          If any player can score at will, they would do so.

          If a player knows how to avoid slumps, they will never willingly allow themselves to have a cold streak.

          If a player can score at 1 ppg, and they go through a scoreless streak, they aren’t doing so because they’re relaxed.

          If a modest scoring player is on a multi-game multi-point hot streak, it’s a hot streak, not because they have the skill to do so all the time.

          Anything else is arguing something really really foolish.

          And yes, your argument about streaks in flipping a coin is exactly the ENTIRE POINT OF MY ARTICLE, I even made mention of exactly that point. Did you miss that part? That’s why telling a player’s hot/cold streaks from the natural game to game variation of any player is not very easy to do, nor is distinguishing the relative streakiness of players.

          If you’re interested in figuring out how to do that instead of discussing odd strawmen, I suggest you read the paper I linked to, it gives you lots of rationale and methodologies for rigorously identifying differences in streakiness levels, continuing on quite a bit farther than the fairly (deliberately) light approach I have taken here. And on that last point, again – something I specifically note in my article.

          • aluchko

            I think this might just be an argument over different use of terminology.

            You say Con Sistent is consistent because his scoring is consistent, but to achieve that consistency you require an extremely inconsistent level of play.

            Meanwhile the streaky scoring coin flip is delivering a perfectly consistent level of play.

            I wasn’t trying to make a strawman I just wanted to make the point a bit more directly since I feel players are often unfairly criticized for being streaky.

            It’s not that Strome is just as streaky as those other players, it’s that Strome and those other players might be delivering a perfectly consistent effort and not be streaky at all. And that ~ 0.4-0.5 ppg level of scoring is going to naturally generate a pattern that looks like streaky-ness more than other scoring levels.

      • aluchko

        For Strome the question is whether his scoring pattern matches what you’d expect for someone scoring ~0.4 ppg. If it doesn’t than he can produce better than he has, but if now we’re getting fooled by lucky streaks. I think that’s a mathematically answerable question, though I’m not sure how to calculate it offhand.

    • I’m confident that he’s a smart player who knows what to do without the puck, which is why the team does quite well defensively when he’s on the ice.

      But 40 to 45 points would be significantly above his historical scoring levels.

  • Abagofpucks

    The only reason i think 40-45 points for strome is probable is because of the way he was shuffled up and down the lineup last year . If they put him on a line that compliments his talents, And they let that line gel for a while it should reasonably be more attainable. Of course its not for sure but hes at an age where his career can get to the next level.

    • I think the issue is that if you put him higher in the lineup, he doesn’t score consistently (ha) enough to be a credible Top 6 guy. That’s ultimately the catch with all players – put them with Connor McDavid, and they’ll all score more. Even Drake Caggiula.

      But it hurts your team doubly when you push a player higher in the lineup than he should be in that you drag two lines down. That’s been a huge part of the story of the Decade of Darkness – too many players asked to play too high in the lineup.

      Strome’s also 25, which is past the age where scoring typically peaks. Improvement from here is usually in the two-way element, versatility, etc., not scoring.

      That’s a pattern of course, not a fait accompli, so it’s always possible Strome’s got some skills in his magic bag he’s yet to show, or maybe another high scoring season or two in his career yet to go.

      It’s just not a bet I’d want to make.

      • OriginalPouzar

        But Strome has shown not to be a drag on the top 2 lines when put there and, in fact, he’s shown to be a boost to Oilers when on the ice with each of Drai and McDavid.

        Leon’s GF% goes from 47% without Strome to 60% with Strome (in almost 200 minutes). Possession metrics stay roughly the same. There is no McDavid effect either (the trio only played 2 minutes together at evens).

        Similar with McDavid and Strome but the sample size is only 56 minutes – McDavid’s goal share goes from 56% without Strome to 75% with him.

        Unless, of course, I’m interpreting the numbers wrong.

        • The sample size is the problem (some would say it’s the use of GF% that is problematic, but really it’s the sample size).

          200 minutes is barely enough (150 mins is my absolute bare minimum, 300 minutes is generally the point of stability) even if you’re drawing conclusions from a high rate statistic like CF%.

          If you want to use GF%, you’d want to multiply that minute count by ~10 to 20. (Goals typically occur at 1/20th the rate of shot attempts – so 1500 mins minimum to 3000 preferred).

          GF% is great for assessing what happened and why coaches made the decisions they did (they react to GF% like crazy, no matter the sample size), but terrible for inference. Your McDavid/Strome example is a good demonstration of that – the total sample size is four. One single goal more, either way, and the percentages swing dramatically.

          In terms of possession, Strome is indeed roughly the same in terms of possession with Draisaitl, but to figure out why that’s happening, you’d want to look at rates. It won’t surprise you to learn that most of Strome’s benefit with Drai comes from the ‘against’ side of the column. Draisaitl’s shot creation falls when he’s with Strome, but his shots given up falls a little more, hence the ‘same or slightly better’ end result.

          As I noted elsewhere, I think Strome is a smart and defensively responsible player, so that makes sense to me.

          In terms of player deployment though, I wouldn’t mind seeing that combo if the Oilers have the lead, but it’s not great deployment when you’re looking for your top lines to score.

          • OriginalPouzar

            OK – fair enough.

            Personally, I do think that Strome is a good option to move up the lineup for RW when they want to shorten the bench and load up. Brodziak, has shown ability to produce with solid players when given a chance up the lineup (Steen, Berglund) – of course, Kyle up to 3C is only for short stretches.

  • Moneyball

    Strome is an effective third liner. His hands for the net are not great but he bring intensity and pushback to the third line. He and Puljujarvi could be an effective pair on the third line next season. Lets hope there is a bit more consistency out of Strome.

  • Eberle was analyzed as primarily only playing well against weaker teams. Does Strome only score against weaker teams as well?

    In looking at it based on end of year team point results and where the Oilers ranked, most teams were obviously better. Let’s consider playoff vs non-playoff teams.

    Of his 34 points last season, Strome scored 20 points against playoff teams and 14 points against non-playoff teams.

    0.43 PPG over 46 games against playoff teams.
    0.33 PPG over 42 games against non-playoff teams

    Strome is +0.1 PPG better against better teams.

    • This is a very good point.

      You could also argue that the Oilers haven’t had a true 3rd line centre in a very long time (See below). When Nuge was the 3rd line centre, that doesn’t count, because he’s not, especially at $6m AAV.

      2007-2008 Cogliano, Reasoner, Pouliot
      2008-2009 Cogliano, Gagner, Brodziak, Pouliot
      2009-2010 Cogliano, Poltuny, Penner
      2010-2011 Cogliano, Gagner, Fraser, O’Marra
      2011-2012 Belanger, Nuge, Gagner, Horcoff
      2012-2013 Horcoff, Belanger, Smyth, Nuge, Smithson
      2013-2014 Arcobello, Hendricks, Smyth, Gordon
      2014-2015 Draisaitl, Gordon, Lander, Roy
      2015-2016 Lander, Letestu, Nuge
      2016-2017 Draisaitl, Nuge, Caggiula, Letestu, Desharnais
      2017-2018 Nuge, Strome, Draisaitl, Khaira