BuzzerBeater Forums

*BuzzerBeater.com Simulation Prediction *
Toronto Raptors @ San Antonio Spurs December 20, 2008
TOR 84 – SA 102
By Forrest Collman and Charles Steinhardt

This will be the third article a series of articles in which we attempt to use Buzzerbeater.com's advanced game engine to predict the results of a real NBA game. Our first game was an outstanding success. Our second prediction had more mixed results. We predicted that Cleveland would beat Atlanta by 6, when in fact Atlanta won by 5, breaking Cleveland's 11 game winning streak. As we explained in our last article, running the same game produces different results, and in about 15% of our simulations, Atlanta did win by at least 5. However, two weaknesses of our simulation were also exposed. First, we simulate NBA players and teams based upon what they have been so far this season. In our simulation, the Celtics behave like a 76-win team, even though it's highly unlikely they will end up that way. Second, the key to our simulation is an advanced coaching AI that looks at the matchup, finds a best strategy, and uses it. Our simulations suggested that Cleveland would do better when Lebron took fewer shots, so our computerized Mike Brown used him that way and average our simulations showed Lebron scoring fewer points than his final tally of 33 and the Cavs winning. However, Lebron ended up taking more shots than he usually does and scored more than he usually does. On the other hand, we correctly predicted that Mo Williams (26 points) and Mike Bibby (24 points) would have big games.

The largest single factor our engine failed to predict was that Atlanta would have such a huge advantage from the free throw line. Atlanta shot 39 free throws in the contest, up from the 24 they average and the 25 Cleveland usually gives up. As Atlanta hit their free throws at a 80% clip, that factor alone makes up the 12 point difference between our prediction and the actual result. What was particularly bad about the abundance of fouls on this night for Cleveland was that Varejao picked up fouls early, leaving Cleveland severely undermanned in the frontcourt since Zydrunas Ilgauskas was injured. As a result, Atlanta ran up the score 47-38 at half time and interior players such as Marvin Williams and Al Horford scored more than expected. With Varejao playing more minutes in the second half Cleveland staged a comeback and had multiple chances in its final possession to tie the game. Whether the large number of fouls was due to referee bias, lazy defense on Cleveland’s part because they were playing back-to-back away games, or a random fluke is hard to say, but it was certainly a large factor in the game.

Today we will be focusing on the Toronto Raptors @ San Antonio Spurs. This is an interesting game for our game engine, because our game engine believes the Spurs are a much stronger team than their record would indicate. Our simulations of the NBA season up to this point consistently placing San Antonio as one of the top 3 teams in the Western Conference. Part of this might be explained by the injuries to Manu Ginobli, or some aspect of our methodology that is overrating San Antonio’s abilities. In particular the Spurs might be weaker this year defensively, as they have allowed 3 more points per game this year compared to last year, and more than 5 more points than 3 years ago. On the other hand, our simulations might be correct, and San Antonio has simply been unlucky so far this season and we should expect them to do better as the season continues. Of course, a human might explain this by pointing out that the Spurs have a habit of starting the season slowly on defense and increasing their defensive intensity as the playoffs approach.

Last edited by BB-Forrest at 12/20/2008 2:30:52 PM

So with that in mind, we are predicting that San Antonio is likely to beat Toronto handily, winning by 15 points or more 65% of the time, and by an average score of 102-84(http://www.buzzerbeater.com/BBWeb/reportmatch.aspx?MatchI...). Applying the “veteran team” correction we just discussed means that a final score of 102-89 might be a better prediction. Tony Parker has an excellent game in our simulations, who scores 25 or more 60% of the time, while shooting 56% from the field and handing out 6 assists. Manu Ginobli's scoring varies a considerable amount, as on average he scores 20 points but is under 10 points in over a quarter of our simulations, so he is a good wildcard if your fantasy team is an underdog, but a risk if you are favored. We think Tim Duncan will have a slightly below average night, scoring under 20 points 74% of the time, though he is likely to rebound well. The top fantasy prospect for Toronto is Jose Calderon, who scores more than 20 points in nearly half the simulations while shooting 35% from 3 and handing out 9 assists. Chris Bosh on the other hand is a Raptor we expect to have a below average night, scoring unde than 20 82% of the time. From a team perspective, a large factor will be in rebounding, where we expect San Antonio to have nearly a 10 rebound advantage over Toronto. Finally, as always these numbers depend upon how our computerized coach uses each team, which might not be the same way that Gregg Popovich or Jay Triano will use them.

Our computerized coach recommends that Triano make more use of Jamario Moon more than the Raptors usually do. In those games in which Toronto does pull out a win, Moon averages 17 points, instead of his normal 7. Also, he should really emphasize the importance of rebounding and sharing the basketball, because in simulations in which Toronto wins they keep the rebounding statistics even, and average 6 more assists per game than San Antonio. Since the Spurs are winning in nearly every simulation, we don't have any great tips for Gregg Popovich, but then again, he has four more championship rings than the entire BuzzerBeater staff combined so he probably doesn't need much help. When the Spurs lose, they typically shoot poorly and fail to gain a rebounding edge.

BuzzerBeater.com is a free online game that puts you in charge of your own basketball franchise, where you're given the opportunity to pick your players, build your arena, manage your roster, and trash-talk with tens of thousands of managers from around the world. We welcome all HOOPSWORLD readers who want to try us out, and hope that you continue to enjoy our simulations. We read all of the comments, even when we get it wrong.

I saw Bosh with 0 FT attempts and thought "no freakin' way". So I took a look, and the last time that happened for him... @SA, about this time last year: (http://beta.espn.go.com/nba/boxscore?gameId=271228024). So maybe you're on to something.

BTW, certain Italian fans are going to say you're still using Smitch as your Raptor coach model, with Bargnani only getting 12 minutes.

nice prediction,
i have a question
in the simulated matchs you put determinated skills in the players
but you use similar form for everyone? or you see the performance at this time and put the form to the player?

bad english 8-)


pd: skills of manu ginobilli please :)
footwork: legendary :P (?)

Handling: Insanely Disturbing Ü

manu ginobilli

Jump Shot: sensational Jump Range: proficient
Outside Def.: proficient Handling: proficient
Driving: prodigious Passing: tremendous
Inside Shot: proficient Inside Def.: average
Rebounding: average Shot Blocking: pitiful
Stamina: respectable Free Throw: phenomenal

How about Duncan and Parker?

I'll assume his flopping ability is modelled as a hidden ability like physicality.

So everyone wants to see the stars. Boring! How about someone like Matt Bonner?

manu ginobilli

Jump Shot: sensational Jump Range: proficient
Outside Def.: proficient Handling: proficient
Driving: prodigious Passing: tremendous
Inside Shot: proficient Inside Def.: average
Rebounding: average Shot Blocking: pitiful
Stamina: respectable Free Throw: phenomenal

From a your previous post:

In my most recent version Lebron was...

Jump Shot: prominent Jump Range: awful
Outside Def.: strong Handling: prominent
Driving: legendary Passing: prodigious
Inside Shot: prolific Inside Def.: proficient
Rebounding: respectable Shot Blocking: mediocre
Stamina: respectable Free Throw: prominent

Experience: legendary

Ginboili is stronger than lebron????and lebron's shot blocking is only mediocre?

Last edited by Steve Karenn at 12/21/2008 11:47:51 AM

stronger in what sense?

lebron is better than ginobli at driving and finishing around the rim... but ginobli is the better pure jump shooter.. and definitely the better 3 point shooter. (for the record, manu is shooting 40% from 3 this year, lebron 27%)

hmm.. shot blocking you might have a point.. i think i moduled that based upon position.. and so lebron got a lower ratings because he is a 3.. but upon reflection that doesnt really make a lot of sense.

Last edited by BB-Forrest at 12/21/2008 12:16:50 PM

So.. how did we do...

Prediction: Parker big game... 25 or more 60%, ~6 assists
Result: Parker 25 points, 10 assists, 8 boards
Check

Prediction: Manu... average night, but highly variable 20+/-8 points
Result: Manu 13 points
Check, but not really a strong prediction

Prediction: Duncan.. slightly under average scoring <20, solid rebounding
Result: 19 points, 7 boards
Check!

Prediction: Calderon leading Raptor, average around 20
Result: 16 points on 7-12 shooting
Hm... why didn't he shoot more he was having a good night, I'd say we missed this one slightly

Prediction: Chris Bosh below average night, <20 points
Result: Bosh 17 points on 6-14 shooting
Check

Prediction: Spurs win the rebounding battle by ~10
Result: Spurs win the rebounding battle by 10
Check!

Other notables..

Jermaine O'neal with a big night... we didn't predict this at all... i'd have to go back and look to see how rare this sort of performance was from the big fella.

My Ex-AAU teammate Matt Bonner going 4-7 from 3! I keep telling Charles this guy is good...

All in all I am very pleased.