Blowing Teams Out vs. Avoiding Being Blown Out
Posted by Neil Paine on November 29, 2010
Because the much-hyped 2011 Miami Heat have had such a bizarre season so far (they apparently have a knack for whipping bad teams but losing close games to good opponents), we've had to think harder about the nature of blowouts. For instance, are they really very predictive if they come against weak opponents?
Adapting an old Football Outsiders study to the NBA, I found that the answer is actually 'yes'. I also found that a team ranking system which gives a lot of weight to blowouts is more predictive than one which places less emphasis on lopsided games.
Now BBR reader "Anon x 2" asks another question:
"What if being blown out means a lot more than blowing someone out?"
Let's take a look using roughly the same methodology as the "Guts and Stomps" article.
For every game since the merger, I recorded the season winning percentages of the teams involved, and the game's home-court adjusted scoring margin (I added 3.5 pts to the road team's margin and subtracted 3.5 from the home team's, since handicappers typically give 3.5 pts to the home squad in any given game).
Wins of 10 or more points (after applying the HCA adjustment) were classified as "blowouts"; losses of 10+ pts were considered cases of being "blown out". I tallied these up for every team that made the Conference Finals, and I also broke them down into four subcategories: blowouts of good (.500 or better) teams, blowouts of bad (sub-.500) teams, times blown out by good teams, and times blown out by bad teams.
Finally, I looked at all 102 NBA and Conference Finals since the 1977 merger, comparing each combatant's regular-season totals in the aforementioned categories. Here are the series records of the teams that were superior in each category:
| Category | Conf Finals | NBA Finals | Total |
|---|---|---|---|
| More Total Blowouts | 39-27 (.591) | 21-11 (.656) | 60-38 (.612) |
| More Blowouts of Good Teams | 39-24 (.619) | 19-11 (.633) | 58-35 (.624) |
| More Blowouts of Bad Teams | 34-30 (.531) | 21-12 (.636) | 55-42 (.567) |
| Fewer Total Times Blown Out | 30-28 (.517) | 22-8 (.733) | 52-36 (.591) |
| Fewer Times Blown Out by Good Teams | 33-28 (.541) | 22-9 (.710) | 55-37 (.598) |
| Fewer Times Blown Out by Bad Teams | 31-23 (.574) | 13-11 (.542) | 44-34 (.564) |
In the Conference Finals, being able to blow away opponents, particularly good ones, is more predictive of success than the ability to avoid being blown out (although avoiding blowouts by bad opponents is apparently better than blowing them out, all else being equal). In the Finals, this phenomenon shifts a bit -- teams that avoid being blown out have actually fared better than teams that racked up a lot of blowouts. Overall, though, once you reach the NBA's Final Four, the team with more regular-season blowouts tends to win the series more often than the team that was better at avoiding blowouts.
November 29th, 2010 at 8:14 am
Have you ever looked at Head to Head record for this question? that is can we rely on regular season results at all to predict playoffs?
November 29th, 2010 at 9:54 am
The last two lines of the table should be "Fewer Times Blown Out" or something.
Since I'm going to claim to be the first person to actually ask this question, I'd like to thank you for actually showing that the championship teams can indeed be defined slightly more by their ability to avoid blowouts rather than blow out opponents.
How come the numbers don't add up row to row? Were there a lot of ties? Can't you just show the raw win total instead of this more/less comparison?
November 29th, 2010 at 9:54 am
Re: #1 - That's a great question. I'll have to take a look at that, hopefully this week.
November 29th, 2010 at 9:56 am
Also, why did you change it from 9+ wins to 10+ wins from your last analysis?
November 29th, 2010 at 9:56 am
Just a quick observation, the two subcategories that deal with bad teams ("More Blowouts of Bad" and "Fewer Blowouts by Bad") both have the lowest winning percentages (56.7% and 56.4%) and are not all that different from 50%. In other words, given the limited sample size can we actually say that the results of blowouts against bad teams (on either end) has any predictive value or are the results within the range due to just randomness.
It certainly appears that the results against good teams is more predictive than those against bad teams.
It would also be interesting to see the results not by the series win or loss, but by numer of games won or loss. Obviously a series going 7 games means that teams were most likely close to equal in quality where as a 4-0 sweep shows one team significantly more dominant. This current analysis would suggest that the team that won 4-3 is no different than the team that won 4-0. It is the same reasoning we use margin of victory instead of Wins and Losses when determining who the better teams really are.
November 29th, 2010 at 9:56 am
points*
November 29th, 2010 at 9:59 am
Re: #2 - Whoops, sorry about that, I forgot you had posed the question in that earlier thread. I knew there was a reason I was thinking about doing this before "Anon x2" commented, but I couldn't put my finger on it. Also, I fixed the table.
As for ties, yes, there were a lot of cases where the two teams had exactly the same number of blowouts/blown outs/etc. I'm not sure what you mean by "raw win totals" -- do you mean the total number of games won in the series?
November 29th, 2010 at 10:02 am
Re: #4 - Just for the sake of round numbers. People were complaining about the Heat-specific definitions in the first post, so I thought "double-digit wins" would be a more acceptable general definition.
November 29th, 2010 at 10:56 am
Cool question. Really makes you think outside of the box. I love this stuff.
November 29th, 2010 at 11:29 am
Awesome stuff Neil. As for Raw Wins, or Raw Losses, I mean for each winner and loser of the Conf Finals/Finals, total up how many of each type of game they had in the regular season, to get a better look at the percentages. So, lets say in a 2 Finals sample, Team A had 20 total blowouts versus Team B who had 10 total (Team A won), and Team C had 15 total blowouts and Team D had 16 (Team C won), then it'd show up in your table as 1-1, where it could be shown as 35-26.
May as well put an average in there too, that way, it can be viewed like "how many blowouts should my team have in an 82 game season, to look like a championship contender?"
November 29th, 2010 at 11:40 am
Ah, I see. So you're thinking more like this:
November 29th, 2010 at 11:42 am
I suppose this opens a whole 'nother barrel of comparing to Finals records (7 games? 4 games?), and pitting that against expected wins based on the blowouts gained in the regular season... ugh my head hurts now.
November 29th, 2010 at 11:44 am
Ah, thanks for the table
November 29th, 2010 at 12:58 pm
Great Post. Interesting results and analysis.
November 29th, 2010 at 2:06 pm
Thanks for the work, Neil.
But, in the end, this sounds like a cop-out but it really isn't, doesn't it just depend on a team-by-team basis?
Meaning, we're looking at historical patterns, but as the mutual fund advisory always warns "past performance does not guarantee future results". And, in the same manner as the stock market, that doesn't mean all quantitative analysis is out the window, it simply means that retro-fitting doesn't always give you the best insight into the future.
November 29th, 2010 at 2:17 pm
Nothing guarantees future results. Past performance is the best predictor, though.
November 29th, 2010 at 3:38 pm
nice work. Now the question becomes: "what about teams that blow out poor teams but struggle against good teams?" We need to separate teams not by "more blowouts" but by big disparities in whom is blown out.
Or better yet, what is the general Point differential threshold for "contenders" against winning teams? A team with a point differential of +6 might be +1.5 against winning teams and be a contender. But perhaps a team with the same point differential overall but -1.5 against winning teams isn't.
November 29th, 2010 at 5:49 pm
"Past performance is the best predictor" ... I'll have to chew on that for a while, I don't necessarily disagree, but I'll have to let that marinate in my soup for a little bit.
November 29th, 2010 at 7:57 pm
Lebron has such long fingernails.
November 30th, 2010 at 11:01 am
Actually the betting odds is the best indicator of future success, but I digress.
Thank goodness the season is long; the Lakers have plenty of time to get things back together. Those losses to Utah and Indiana were awful.
When it comes to predictive value,
Blowing teams out > Close wins > Close losses > Being blown out. That is my story and I'm sticking to it.