Wednesday, December 10, 2008

Week 15: Predictions

Hey everyone, be sure to check out NBA Black Box. NFL Black Box is great and all, but NBA BB is really knocking it out of the park. Over the past 7 days, it's 16-3 ATS (84.2%). On the season, the best picks, the 5-Star picks are 83-43 (65.9%). There are power rankings, too. It's totally free just like NFL BB, so take advantage of it.

May want to be careful with the spread, it's been over-estimating the probability of the underdog covering. In other words, teams that win aren't winning by enough. Which is very strange because the win % is virtually identical to reality (I'll go over that when I post the season's results) and O/U is not far behind. Just to give you an idea, without all of the details, based on confidence % and the number of games in each interval, the expected number straight up wins (oversimplified as the number of wins I should have, if prediction approximates reality), is 110, NFLSim hit 116, so that's pretty close. O/U expected wins is 100, actual is 94, again it's pretty close. Expected ATS wins is 105, though only 80 hit.

What's going on? NFLSim is perfect at assigning probabilities to straight up wins, very good at assigning probabilities to point totals, but the spread has been getting lost somewhere along the way. The right teams are winning, but the winner should be scoring more and the loser less.

Moral of the story: if your system's spread agrees with NFLSim's spread, go for it. Totals seem reliable. For straight up bets, if Vegas's payout is greater than the value of 100/win%, i.e. 100/53 = 1.89 vs. Vegas payout of 1.79, bet on that team to win, whether it's the predicted to win or lose (since win % are so accurate, mathematically, you will make money in the long run).

NO at CHI -3, 45
NO: 20.42
CHI: 23.24
Winner: CHI, 53%
Spread: NO +3, 55%
O/U: UNDER, 58%

BUF at NYJ -7, 41
BUF: 20.22
NYJ: 25.19
Winner: NYJ, 61%
Spread: BUF +7, 59%
O/U: OVER, 58%

DET at IND -17, 45
DET: 17.39
IND: 26.70
Winner: IND, 76%
Spread: DET +17, 74%
O/U: UNDER, 57%

GB -2 at JAX, 45
GB: 23.53
JAX: 21.97
Winner: GB, 56%
Spread: JAX +2, 66%
O/U: UNDER, 52%

SF at MIA -6.5, 41.5
SF: 17.04
MIA: 25.29
Winner: MIA, 73%
Spread: MIA -6.5, 53%
O/U: OVER, 51%

TEN -3.5 at HOU , 44.5
TEN: 22.37
HOU: 23.39
Winner: HOU, 52%
Spread: HOU +3.5, 62%
O/U: OVER, 52%

SD -6 at KC, 45.5
SD: 24.06
KC: 22.32
Winner: SD, 56%
Spread: KC +6, 63%
O/U: UNDER, 52%

TB at ATL -3, 44.5
TB: 19.81
ATL: 22.83
Winner: ATL, 59%
Spread: TB +3, 53%
O/U: UNDER, 54%

DEN at CAR -7.5, 48
DEN: 20.53
CAR: 24.29
Winner: CAR, 59%
Spread: DEN +7.5, 67%
O/U: UNDER, 63%

NE -7 at OAK, 40
NE: 23.34
OAK: 20.83
Winner: NE, 57%
Spread: OAK +7, 68%
O/U: OVER, 62%

WAS -7 at CIN, 36.5
WAS: 22.70
CIN: 17.12
Winner: WAS, 64%
Spread: CIN +7, 64%
O/U: OVER, 59%

PIT at BAL -2, 34.5
PIT: 15.31
BAL: 22.02
Winner: BAL, 68%
Spread: BAL -2, 65%
O/U: OVER, 54%

NYG at DAL -3, 45
NYG: 20.59
DAL: 20.17
Winner: NYG, 52%
Spread: NYG +3, 63%
O/U: UNDER, 65%

MIN at ARI -3, 47
MIN: 20.13
ARI: 22.68
Winner: ARI, 58%
Spread: MIN +3, 52%
O/U: UNDER, 66%

SEA -2.5 at STL, 43
SEA: 21.90
STL: 22.52
Winner: STL, 51%
Spread: STL +2.5, 55%
O/U: OVER, 55%

CLE at PHI -14, 39
CLE: 15.61
PHI: 26.12
Winner: PHI, 80%
Spread: CLE +14, 62%
O/U: OVER, 56%

Straight % Wins Games Win %

50-59 6 10 60.0%

60-69 1 3 33.3%

70-79 2 2 100.0%

80-89 1 1 ---

90-100 0 0 ---

Total 10 16 62.5%





Spread % Wins Games Win %

50-59 3 4 75.0%

60-69 4 9 44.4%

70-79 1 1 100.0%

80-89 0 0 ---

90-100 0 0 ---

Total 8 14 57.1%





Total
Wins Games Win %

50-59 5 11 45.5%

60-69 3 4 75.0%

70-79 0 0 ---

80-89 0 0 ---

90-100 0 0 ---

Total 8 15 53.3%

2 comments:

  1. Is the 66% confidence on Jacksonville covering the spread really correct? This seems inconsistent with GB winning 56% of simulations. Perhaps the line used in the (spread) calculation was more than 2 points...?

    I assume that you store the final game total for each of your simulated games and calculate the results ATS afterward from the ensemble of results; if so, this would be an easy mistake to make (and to fix).

    ReplyDelete
  2. I went back and checked it and it's correct. Odd, but correct. Out of 501 games, GB wins 283 (56%) but they only win by more than 2 points (cover the spread) in 172 (34%) games. Which means that they win by 1 or 2 points in 111 (22%) games.

    On average, GB outscored JAX by 1.5 points (23.53 to 21.97), so it makes sense, though it's pretty improbable.

    ReplyDelete