Alex Ovechkin: 50 in 50?

ovechkin-goal

Could Ovi join Maurice Richard as one of five players to score 50 in 50?

Could Ovi join Maurice Richard as one of five players to score 50 in 50?

Alex Ovechkin kept his pace for an 82-goal season intact against the Islanders which got me thinking: is 50 in 50 even close to a reality?

50 goals in the first 50 games is a rare feat indeed. It has only been done by five players: Maurice Richard in 1945. Mike Bossy in 1982. Lemieux in 1989. Brett Hull did it twice, and in back-to-back seasons no less (1990-1 and 1991-2). The Great One accomplished the feat three times: 1982, 1984 and again in 1985. So why not Ovechkin?

To estimate his chances I will use something called a Monte Carlo simulation. It starts with me using the shots on goal and shooting percentage for Ovechkin since he made his debut in the league 5 years ago to the last game of the regular season in 2009-10. Only regular season games were counted. For instance, he has only been held to zero shots on goal 3 out of 396 games, or .76% of the time. He has shot 20% in 32 of his regular season games, or 8.08% of the time.

Using this data, and Excel’s random number generator, I simulated goal-scoring totals for 50 games. Then, I ran this 50-game season 10,000 times and calculated the results.

Before I reveal the chances, take a guess at what you think they are. I’ll wait….

Ok, now that you have your guess, let’s look at the numbers.

Ovie50in50

Based on 10,000 50-game seasons in my cold, heartless math universe, Ovechkin will average 26 plus or minus 6 goals. He only hit the magical 50+ mark two times out of 10,000 – once with 58 goals and another with 53. He just missed the mark once with 49 goals in 50 games and 48 goals 4 other times.

For kicks I also ran this for an 82-game season when I saw NHL Hot Stove predicted Ovechkin would only score 37 goals. For an 82-game season, I have Ovechkin scoring on average 42 goals, plus or minus 8.

82gms

So ~68% of the time he should be between 34 and 50 goals. A 37-goal season looks light. I would take the over.

  • peachy

    math nerd…:)

  • peachy

    Don’t take it the wrong way, I love math and the devotion you show to the Caps fans!!!! Just kidding fellas!

  • Abhi

    If I wanted to hear cheap jokes about liberal arts students, I would read a certain webcomic that totally sucks.
    Great analysis, though, and it’s okay to use the words “standard deviation” on the internet :)

  • Tim

    I doubt he’ll get 50-in-50. I just don’t think it will happen anymore, with the vast improvement in goalies since the last time it happened. However, I still think he will get 50 over the course of the season – and probably 30 or more in 50.

  • bobby

    Ah, but his goal-scoring rate increased in his last few seasons. What if you ran the simulation based on just his past 2 or 3 seasons–which would be more accurate?

  • http://redlinestation.blogspot.com RAL

    I agree with bobby. Ovechkin looks much better under Boudreau (in large part because the team around him plays better and helps him be more dominant offensively).

    If there’s a new set of rule changes that gets Ovechkin tons of power play time, with all-around high PP%–like in 2006–then it’s feasible, I think.

  • Vlad

    How can you arrive at “Ovechkin scoring on average 42 goals, plus or minus 8″, given that historically he averages 50+? Seems to me that something is way off.

  • Neil – RMNB

    @bobby: interesting suggestion, so I did it.

    Using his numbers from the last 2 years, out of 10,000 seasons he hits 50+ goals 6 times, 49 goals 3 times and 48 goals 4 times. Averages 27 goals, plus/minus 6.

    Using last 3yrs: out of 10,000 seasons he hits 50+ goals 14 times, 49 4 times, 48 8 times. Avg 28, plus/minus 6.

    Better, but not enough for me to pack our bags and head to Vegas. :-)

    Again, great suggestion!

  • Neil – RMNB

    @Vlad: Good question.

    These are averages and the plus/minus is the standard deviation. Meaning, we expect 68% of the time for Ovechkin to score 34 to 50 goals. We expect him to score 26 to 58 goals 95% of the time.

    So when I say “Ovechkin scoring on average 42 goals, plus or minus 8″ what I am really saying is “If we had an Ovechkin playing in each of the 10,000 alternate universes there would be different results from each. Some Ovechkins will see longer/shorter shooting slumps, some will see injury, some will see his teammates get injured/traded, etc. So when we look at the 10,000 Ovechkins as a whole, 65% should score on average 42 goals, plus or minus 6 goals.

  • KHtaD

    If you’re simulating that Alex Ovechkin is going to score an average of 42 goals over the next 82 games based on the past five years of data (in which he’s dipped below 50 only once), I suspect there’s something wrong with the simulation.

    I’d be interested to see what the results were with his career shooting percentage applied to his distribution of shots on goal, rather than the very chunky per-game shooting percentages.

  • KHtaD

    Another thing I forgot to mention; to make it clear, when the mean is well below all previously observed data used to make the simulation parameters, it’s a red-flag; something’s wrong.

  • http://twitter.com/ngreenberg Neil – RMNB

    @KHtaD: Sorry there is so much confusion. Let me explain it a different way, but first a response to your statements: “If you’re simulating that Alex Ovechkin is going to score an average of 42 goals over the next 82 games based on the past five years of data (in which he’s dipped below 50 only once), I suspect there’s something wrong with the simulation.”

    If you expect Ovechkin to score 50 goals a year forever, that is flawed logic. Just because someone “always” scores x goals doesn’t mean the following season they won’t score y goals. And I am not saying he “won’t” score 50 goals. In fact, I am saying with 95% certainty he will score somewhere between 28 and 58 goals.

    “when the mean is well below all previously observed data used to make the simulation parameters, it’s a red-flag; something’s wrong.”

    So players don’t age? They don’t get injured? Fluky things don’t happen? NOTHING happens that could cause even a dynamic player like Ovechkin to have an off year? The chances of that happening are ZERO? I am curious: why is Ovechkin immune to the laws of nature and physics every other person seems to be influenced by?

    I am sure people in 1987 thought the odds of Gretzky scoring less than 50 goals were just as far fetched, yet he came in with a “clunker” season of 40 goals. Things happen.

    Now another explanation:

    1. I am not saying he is going to average 42 goals over the next 82 games. I AM saying if we took 10,000 simulated seasons, each with 82 games in them, we can expect Ovechkin to score on average 42 goals, PLUS OR MINUS 8. In some of these alternate universes he scores 34, maybe due to “injury” or a teammate getting hacked to death by the Boogeyman – whatever – but things happen.

    2. This is a 82 game SIMULATED season. Ovechkin cannot be expected to play every game every season – and even if he did chances are he will not score a goal in every game, nor even get a shot on goal every game. Granted, it is rare when he is held without a shot, but it CAN and DOES happen. The model takes this into account. And in reality, if he doesn’t get a shot on goal or doesn’t play the game, the net result is the same: 0 goals, so it works out.

  • KHtaD

    You can stop with the strawmen at any point. I said almost nothing of the things you accuse me of in your rebuttal.

    My objection has nothing at all to do with the results of the projection – 50 is a very difficult plateau to reach and you need either unprecedented volume (Ovechkin) or absurd accuracy (Kovalchuk) or some of each and some luck (Crosby). Any number of factors, including age, injury, or bad luck, can prevent a player from hitting 50.

    Rather, the problem is this; if you’re running a simulation of Alex Ovechkin’s next season BASED ON THE DATA FROM THE PAST FIVE YEARS and you’re arriving at a mean of 91% of the prior low-water mark, you’ve screwed up your simulation somewhere. Your mean is 42. The prior low in the data is 46. If the simulation is based on the past data, then the simulation is wrong.

    If you’ve included correcting factors for age and/or injury in your simulation, then the result makes sense, but I still think it’s a little strong. If you haven’t, and what’s in the article, a Monte Carlo based on observed shots-in-a-given-game and observed-shooting-percentages-in-a-given-game is all there is, you probably have a binning error and the first place you want to start looking is the shooting percentage. I think you’d be far better served with a Bernoulli simulation – you could still use the shots-in-a-game methodology, but you’d use a set shooting percentage for each shot; a weighted coin flip, in the parlance of Vic Ferrari. Figure out what you think Ovechkin’s true shooting percentage is, keep the shots/game, then apply that shooting percentage to each event. Much cleaner, no binning problem.

    You could also use a reverse binomial distribution, the distribution of the number of trials it takes to get to the Kth event, to figure out what AO’s chances of making 50 in 50, or 50 in 82 are in a symbolic way.

    In sum, I have no problem with the conclusion, but I do have a problem with how you got there, and I definitely have a problem with all the strawmen.

  • Breaklance

    While the simulations are done 10,000 times each set is only 50 or 82 times which is very small and allows for skewing. It is akin to flipping a coin 20 times you may get a 60-40 break and you do the same test thousands of times getting similar results. But if you do just one test 10,000 times you’ll probably get 50.5-49.5 and as you go out to infinite it will draw to it’s actual break.

    So what I’m saying if anything at all is that individual seasons will see big changes from one to the next but if you were to look at his career it would look more normal

  • Sean

    First off, before getting to the nitty gritty of the data itself, fix your labeling on the graph. According to your labels, your intervals are overlapping (which I certainly hope is not the case). For instance, a 10-goal season would apply to both the 5-10 and 10-15 intervals. I suspect the intervals are actually 5-9 and 10-14, but you’re the only one who really knows.

    Secondly, KHtaD is correct. If your simulation is deviating from the observed results that much, something is most likely wrong. Both KHtaD and I understand standard deviation, but the mean is just too low for this to be correct. You didn’t describe any factors you used in your Monte Carlo to account for regression to the league mean, injuries, or aging (though, seriously, Ovie is still in his prime, he’s not going to age negatively this season).

  • Neil – RMNB

    I will say one final note on this and then move on.

    Let’s not forget to see the forest through the trees. The purpose of the post was to have some fun with Ovechkin getting 50 in 50. Agree or disagree with how my numbers were reached, the bottom line is: The chances are slim to none. Maybe it is .0002%, maybe it is .0016%, maybe it is (n+1)^(ln(y)-e^3), but we should all agree that unless the odds are 14 trillion to one we should not take the bet.

    As for projecting an 82 game season, again, let’s not lose sight of the message: “A 37-goal season looks light. I would take the over.”

    You know what would be refreshing? A comment that starts with “I did my own simulation and came up with…”