Drafting in the 21st Century?

[beav]

It's a numbers game

Sunday, June 17, 2007

BY BRIAN HENDRICKSON

The billboard stares at drivers headed west into Portland on I-84.

"Oden, honk once," half the sign says.

"Durant, honk twice," urges the other.

The Portland Trail Blazers' marketing campaign fuels the debate over who the team should select with the No. 1 pick in the NBA Draft on June 28, but it also illustrates the division on the issue. A month of arguing has only produced an unending wail of honks for each side on radio shows and in the media, with no consensus near.

On one end is Ohio State's powerful center Greg Oden, who at age 19 is already worshiped as a legend on the hallowed courts in his home state of Indiana. At the other is Texas' Kevin Durant, a superstar-in-waiting whom some believe will revolutionize the game. Each camp will argue passionately for their favorite. And each will say the Blazers would be foolish to pass on either.

Honk once for Oden - if only the decision could be that simple.

But Kevin Pritchard believes he has the formula that will produce a straightforward solution. And it won't rely on emotional opinions, debates or honking horns to uncover the answer. No, the Blazers general manager's approach will delve deeper, breaking the issue down into its unbiased components.

The numbers.

Pritchard's formula, which he guards with a thick veil of secrecy, is an algorithm designed by a Massachusetts Institute of Technology-educated whiz. And he believes so highly in its application - and is so concerned about other teams getting it - that Pritchard will only discuss the method and its results in generic terms. But the potential of the technique Pritchard terms "quantitative analysis" excites him as he prepares to use the formula in a draft for the third consecutive year.

"It's been right on with some guys that we wouldn't normally have picked," said Pritchard, who offered no details. "I feel comfortable with it, and it's going to become a bigger part of our evaluation process."

Beating bias

To understand why quantitative analysis gets Pritchard so excited, first understand the unavoidable limitations of traditional methods.

For nearly as long as there has been professional basketball, teams have relied on swarms of scouts to research prospects and offer opinions on which players a team should draft. Those scouts fly to all areas of the country and watch dozens of games in person in addition to hundreds on film, during which they sort out what they feel are the strengths and weaknesses of players the team has targeted.

But Pritchard feels human nature limits the success rate of traditional scouting methods because the data is contaminated with individual opinions and proclivities. The scout's preferred style of play could factor into their evaluation, or perhaps they'll watch a prospect on an off night and have their perceptions tainted.

For instance, if a scout who prefers a system that runs through a strong post presence were to watch Oden and Durant, they may see greater benefits to taking Oden, with his powerful 7-foot frame, and limitations with Durant, a wiry wing player.

Those biases make traditional methods a subjective analysis, and sports history is littered with examples of talent miscalculations (Michael Olowokandi, Kwame Brown) and overlooked warning signs (Sam Bowie's injuries).

That's what Pritchard sought to eradicate through the use of mathematics: A bias-free analysis that the team could use to counterbalance its live scouting.

A new approach

Pritchard first learned of the concept after reading a pair of books.

The first, "Moneyball," chronicled Oakland Athletics vice president and general manager Billy Beane's use of mathematical analysis to identify the best talents available at the salaries the small-market team could afford. Anchored around Beane's revolutionary approach, the A's have advanced to the postseason four times since 1998 and have had players win the AL MVP award (twice), AL Rookie of the Year (twice) and a Cy Young Award.

The other book, "Bringing Down the House," told the story of a group of six MIT students who used mathematics to develop a system that would allow them to efficiently - and very successfully - count cards at a blackjack table to maximize their bets. The group earned millions of dollars in winnings from that strategy and caused casinos across the country to develop extreme tactics to guard against card counting.

Pritchard turned to a member of that MIT group, Jeff Ma, after deciding to further pursue a mathematical approach to scouting. Ma had recently co-founded a new fantasy sports company, Protrade, whose foundation was based on mathematical principles similar to those chronicled in "Moneyball." With Protrade consulting with the Blazers, algorithms were developed which could predict a prospect's likelihood of success.

The process gathered readily available statistics - such as a player's age, height, weight, scoring and rebounding averages, and other numbers applicable to their position - crunched them through the algorithm to measure their perceived chance of success by their third season and ranked the results. Those figures would then be used to complement reports from the Blazers' scouts to form the complete analysis.

Pritchard and Ma are quick to point out that the system is not foolproof, though, and its findings will raise some eyebrows amongst casual fans.

For instance, Memphis point guard Kyle Lowry, selected with the 24th pick, was Protrade's top-ranked player in last year's draft, while Portland's all-rookie forward LaMarcus Aldridge - taken with the second pick - ranked 22nd.

But the algorithm successfully identified Utah's all-rookie forward Paul Millsap - drafted 47th, but ranked sixth by Protrade - as an undervalued prospect.

So while the system isn't an end-all solution, it gives the Blazers added information that can improve their decision making. It can lead them to reevaluate a well-regarded player whom Protrade ranked low, or look deeper into an overlooked prospect who rated favorably.

"I think that there's points and places where the numbers are going to lie, and there are places where your eyes are going to lie," Ma said. "So the beauty of using both of these is you get into a situation where you have all your bases covered. Essentially what we've tried to do is provide a situation where they get a lot of opportunity to sort of see, to get as much information as you can."

Putting it in play

The first draft was considered a test: Only 1 percent of decisions were based on Protrade's analysis, Pritchard said. But that 2005 draft produced Jarrett Jack, who became the Blazers' starting point guard last season.

Pritchard won't reveal how much the system has affected their decisions since, but last year's draft - with Protrade's algorithm playing a larger role - has been widely hailed as the most successful in franchise history. Brandon Roy and Aldridge were named to the All-Rookie First Team after being selected among the top six picks, while point guard Sergio Rodriguez - acquired late in the first round - quickly gained notoriety as a rising talent.

But there was another key decision in that draft: The Blazers elected not to select Adam Morrison, a regional favorite from Gonzaga, after acquiring the No. 2 pick in a trade with Chicago. Fans had rallied outside the Blazers' practice facility to express support for drafting Morrison, while analysts compared his game to that of Hall of Famer Larry Bird. The hype put the Blazers under pressure to take him.

But Protrade's algorithm? Ma said it projected Morrison as a late first-round selection at best.

"We didn't really like Morrison," Ma said.

Perhaps the Blazers still would have passed on Morrison without the algorithm's input - though he went on to an all-rookie season in Charlotte and justified his lottery selection. But the contrasting viewpoints illustrate what Pritchard believes is the system's greatest benefit: It can raise a warning flag about a potentially poor decision, or identify a player they may have otherwise overlooked.

"It always makes me reevaluate where we are," Pritchard said. "There are certain things that can't be measured. You can't figure out determination without looking in someone's eyes. I'm not saying this is foolproof. It's just about helping us steer away or towards someone to look at a deeper level."

A growing trend

Other NBA teams are now recognizing the benefits as well. New Houston Rockets general manager Daryl Morey has initiated similar work with quantitative analysis. San Antonio was beginning development of its own system when Pritchard left the Spurs in 2004 to join the Blazers as director of player personnel. And Ma said he has heard of several other teams experimenting with the concept.

But Tom Penn, who started last month as the Blazers' assistant general manager and was recently introduced to Portland's system, said the Blazers' approach is groundbreaking.

"The group here is definitely pushing the envelope," said Penn, Memphis' assistant general manager for the last seven years. "What I've seen is raw data, if filtered properly, doesn't lie."

But the true test has yet to be performed. The last two drafts may have produced encouraging talents, but they're hardly comparable to the expectations being placed on this year's decision.

Some believe the choice is a toss-up. Oden's size, athleticism and strength inside appear to hold the ingredients of a champion. But Durant's passion, athleticism and offensive skills are unique, electrifying and impossible to dismiss. The debate over who Portland should pick has raged for nearly a month among fans and in the media. The sides are starkly divided.

So there's the test: Can math answer a question that emotions, opinions and honking horns struggle to solve?

"Sure," Pritchard deadpans, revealing no cards. "Why not?"

And there Pritchard lets the question hang, preferring to let this numbers game play out in secret.

Protrade's Top 10

The players in the 2006 NBA Draft that the Trail Blazers' draft consultant mathematically concluded had the greatest perceived chance of success by their third season.

Rank, Player Draft Position Drafted by Notable

1. Kyle Lowry 24 Memphis Played in only 10 games after breaking wrist in November.

2. Tyrus Thomas 4 Portland* Averaged 5.2 points and 3.7 rebounds in 72 games.

3. Rajon Rondo 21 Boston Second among rookies with average of 3.8 assists.

4. Mardy Collins 29 New York Averaged 4.5 points and 2.0 rebounds in 52 games.

5. Brandon Roy 6 Minnesota** Named Rookie of the Year; started 55 of 57 games played.

6. Paul Millsap 47 Utah Voted to NBA's All-Rookie Second Team.

7. Shelden Williams 5 Atlanta Led all rookies in rebounding with 5.4 per game.

8. Cedric Simmons 15 New Orleans Averaged 2.9 points in 43 games.

9. Daniel Gibson 42 Cleveland Averaged 13.5 points in Eastern Conference Finals.

10. Shawne Williams 17 Indiana Averaged 3.9 points and shot 47 percent from the field in 46 games.

* Traded to Chicago for No. 2 pick LaMarcus Aldridge (ranked 22nd by Protrade)

** Traded to Portland for No. 7 pick Randy Foye."

[TexasPete]

I love Pritchard.

[txsting]

Very interesting...thanks for that post. As a software engineer that kind of stuff really intrigues me. The system will only be as good as the data that you feed it. Like they said, no algorithm could perform the task perfectly, but it's a good opinion to look at.

[mrhonline]

Considering that Portland traded the 2nd highest ranked player on that system, I'd agree that it's worth about 1% of their decisions.