### How the Stretch Run Predicts Playoff Success

## April 8th, 2010, 1:21 pm

This comes from an email we got for the podcast (make sure to listen, we actually have a guy with credibility this week), asking essentially, “Does the Sharks losing streak (or later, the winning streak) help or hurt the Sharks’ playoff chances? Does success in the last 10 or 20 games result in deep playoff runs?”

This is an important question, and let me be perfectly honest- Gabe Desjardins’ recent post on something similar is probably a lot better than this one is going to be. So I’d read this first with a sense of charity, then click over to Behind the Net and read the real stuff. To defend myself, I was planning this post since a week ago, so I’m not copying Gabe- I swear.

To answer this question, I compiled the record of the last 10 and 20 games of every playoff team since 2001. This gives us 112 teams, which isn’t a giant sample size, but it should smooth over some rough edges. I correlated this (Pearson product-moment if you must know) against the round where they eventually lost (or not), but then I decided to copy Gabe just a little and use playoff wins instead, because that gives us more granularity than just rounds achieved- we get 16 gradations instead of 5.

Here’s the chart. I’ll explain in a minute.

Ok, so this is what’s called an XY scatter plot, with the number of points in the last *n* games (red dots are last 10 games, blue dots are last 20) versus how deep that team went in the playoffs. 0 wins is swept in the first round, 16 is the Stanley Cup Champions. If there were a strong correlation, we would see something of a line going from the lower left to upper right in each color. That is, the teams that do poorly in the last bit of the season also bow out early. Or we might see something completely counter-intuitive- a negative correlation, in a line that goes from the upper left to bottom right. That would signal that teams that do well in the final stretch “peak too early” and are more likely to bow out in the first few rounds.

We have what we call in the mathematics world, a “mess”.

Of course there’s going to be some variance, and the graph would be more like a cloud than a line, but here we just have a plate of spaghetti. If you want numbers, the correlation for the last 20 games points and playoff wins is 0.12, and the last 10 wins 0.08. Correlation ranges from -1 (late points always means bowing out early) to +1 (late points means going deep in the postseason). 0.12 is *slightly* positive- it’s probably a tiny bit helpful to do well late in the season- but it’s certainly nothing predictive. It’s essentially a crapshoot- performing worse (or better) doesn’t mean much. Let’s look at some extremes:

- The team with the most points in the last 20 games among all teams I looked at is the 2006 Red Wings, who only lost one game in regulation in the last 20. That team lost in the first round to Edmonton.
- That very same year, the Carolina Hurricanes scored only 21 points in the last 20, barely .500. That’s the second worst record of any of the playoff teams that year (only the #8 New York Islanders were worse). For those that remember, the Canes won the Cup.
- Last year, the team that had the best late record of all playoff teams, the Pittsburgh Penguins, won the Cup. They played the Wings (again) in the Finals, who had 12 fewer points in the same number of games, good for 13th amongst playoff teams.

Does anything correlate better to playoff success? The answer is yes. The overall 82-game record correlates better: 0.32. And even better is the correlation of seeding to playoff wins: -0.37 (negative because a lower number (seed) correlates to a higher number (playoff wins)). This does tell us something interesting- it’s better to be lucky than good. It’s better to luck into a higher seed with a worse record (in a weak division) than score a ton of standings points and only get the #4 or #5 seed. With the Sharks guaranteed the #1 or #2, they are in the best possible position. Now the question is, can they transform this advantage, however slight, into real playoff happiness?

We’re going to find out. I’m nervous.

Amazing. Just the other night Bret Hedican was waxing poetic about how his Canes “were playing the right way” going into the playoffs and how the Sharks seem to be missing that attention to detail. Seems his memory was just a little off. Don’t get me wrong, I think Hedican has improved by leaps and bounds from the start of the season in the studio, but just another example of players sticking by cliches instead of facts.

I would love to see how record after the trade deadline correlates, but I would assume it is pretty close to your 20 game sample. I wonder if seeding helps more because you are playing a worse opponent? More home games? In any case, good stuff here Dudes!

They did get 12 points in their last 10 games that year, but that’s not exactly setting the league on fire. Worth mentioning that 6 of those points came against the hapless Ovechkin-less #14 Washington Crapitals. Some quirk of the schedule had them play them three times in a row, all of which they won, but two still went into OT.

I like Hedican too, but the conventional wisdom sometimes tells a better story than the truth. Somehow I doubt we’d hear a “playing the right way” comment if the Sharks lost their last two games to playoff teams, just like the Canes did in ’06. Actually, the similarities between those Canes and these Sharks is a little eerie. If the Sharks lose their last two, they’ll probably miss out on the #1 seed by a point or two, just like the Canes did. Let’s say the Sharks lose both of the last two, but one in OT. That would give them 21 points in the last 20 games. Spooky!

So before we all get going about tonights game…. My first thoughts are that I love the fact that the Sharks were INITIATING the physical play for once, instead of retaliating all the time.

I don’t mind an occacional slash to the back of an opposing player. 🙂

Mike

I know that you have an engineering background and I do too.

What is interesting about your graph is that I kinda see a bi-modal distribution from your scatter plot.

It looks like teams that only did average always went out early.

Maybe you should normalize the number of points to points/game for the last 10 and the last 20.

there may be a better correlation or at least a more telling pattern.

Wu

the red wings won in a shout-out tonight. do they have the tie-breaker over nashville? does that mean they will be 5th or 6th and play vancouver or phoenix? will the sharks play colorado or nashville?