Power Ranking

[EvanZ]

For betting against the spread, do you think picking the winner or RMSE is more important?

[mystic]

For betting against the spread, do you think picking the winner or RMSE is more important?

Is a Power Ranking a "betting tool" or something which should give us a prediction of the winners? I'm more interested in the game of basketball than betting, thus I prefer a system which can tell more about the possible winner. But well, if you want to beat the spread, RMSE is the better way to determine the better "betting tool".

How often did your system beat the spread?

[EvanZ]

I'm more interested in beating spreads. And I don't have a system that can do that consistently.

I don't think using the Vegas ratings will beat the Vegas lines consistently either (it's always going to be a step or two behind), but I think it does give us the best idea of "reality", simply because Vegas knows what Vegas is doing. Even if you beat Vegas, it will be just barely and maybe not for very long. And anyone who can beat Vegas regularly will likely not give you their ratings.

[mystic]

What does "beat Vegas" mean here? My system beat the spread 40 times out of 83 games. Well, I wouldn't actually call that "beating" at all.

Did you make any test with your current system? I would like to know how successful that is.

[EvanZ]

What does "beat Vegas" mean here? My system beat the spread 40 times out of 83 games. Well, I wouldn't actually call that "beating" at all.

You answered your own question. When you can consistently beat Vegas spreads 54% of the time, let me know.

Did you make any test with your current system? I would like to know how successful that is.

I have not done this for the NBA yet. I can tell you that my own ridge-regressed NFL ratings managed only 52% this season over half the season ( "Ridge Regression").

http://www.thepredictiontracker.com/nflresults.php

You can see it's not so easy. The best football rating out of dozens listed here managed 57% this season. Eric Hollobaugh got 58% last season, but could only manage 52% this season. (I looked up the winner from 2010 which got 60%, but they are not listed for 2011).

The moral of the story here is that it's very, very tough to beat Vegas. So if you can't beat, why not join them?

[EvanZ]

Here are the prediction tracker lines for the NBA. In theory, going to Crow's question earlier, one could use these predictions/rating systems to make an "ensemble" predictor. That might be interesting to compare to the Vegas ratings.

http://www.thepredictiontracker.com/prednba.html

[mystic]

You answered your own question. When you can consistently beat Vegas spreads 54% of the time, let me know.

Well, that's what I thought.

I thought about a new system.

Method:

1. Calculate average SPM for a team.
2. Calculate expected scoring margin via sos adjusted real win%.
3. Build the average of those two.
4. Adjust via a expected distribution going from 6 to -6 (not quite sure whether the expectation for a season should be 6 to -6, if someone has a better idea, please let me know).
5. Adjust for average pace.

http://bbmetrics.wordpress.com/vegas/

First test comes tonight.

[Crow]

Code: Select all

Home	Visitor	Line AVGPredtrac	SAG  EvanVPR     Mystic Betmarktanal RidgeRegTeam playerRAPM Equalblend Blend-Line
											
Boston	Chicago	-2.5	-4.95	-6.05	-0.77	-5.94	-1.5	-5	0.1	-3.33	-0.83
Charlotte	Detroit	3	3.23	5.47	1.4	2.49	2.5	-3.7	-2.2	1.52	-1.48
Dallas	Milw	9	6.79	6.9	6.8	6.29	7	10	7.2	7.50	-1.50
Denver	Miami	-3.5	0.94	2.29	-0.59	1.54	-2	0.5	-1.3	-0.27	3.24
Houston	Sac	8.5	6.74	8.36	5.9	3.42	12.5	8.6	14	8.50	0.00
LA Lakers	Clev	10.5	8.52	7.39	11.02	8.45	10.5	8.7	10.1	9.40	-1.10
New Orl	Minn	-1.5	2.96	4.38	-0.75	-0.57	-0.5	-2	-3.7	-0.21	1.29
Phil        Wash	13	17.23	19.96	11.44	16	8.5	-1.6	-2.5	10.25	-2.75
Phoenix	NewJ	6.5	9.98	9.29	7.5	8.17	8.5	-0.1	-2.4	5.93	-0.57
SanAnt	Portland	1	2.67	3.55	3.12	-0.29	2.5	3.4	2.1	2.26	1.26
Toronto	Indiana	-5.5	-2.55	-1.83	-1.92	-3.25	-2.5	6.2	5.8	-0.69	4.81

I am not adept at cut & paste code formatting.
But here is a 8 metric blend for tonight. Average from Prediction Tracker is a large blend itself so there are about 16-20 elements being used.

Denver-Miami, Phil-Wash and Toronto-Indiana are the games where the equal blend has the largest variance from the line found at Prediction Tracker. Were they good betting opportunities? Will check after the games.

Mystic= values from your site. Would be interested to see the values for your new blended metric.
For Evan VegasPowerRatings, Mystic,Bettingmarketanalytics, RidgeRegTeam and playerRAPM I assumed a 3 pt HCA, not knowing what they gave to individual teams.

[Crow]

If you made the bet suggested by the blend - the line in the 3 biggest variance games (Denver-Miami, Phil-Wash and Toronto-Indiana) you'd have won 2 of 3. This time.

[mystic]

Code: Select all

Home     Visitor     Line      AVG       SAG     EvanVPR    Mystic   Betmarkt   RRTeam     RAPM   EqualblendBlend-Line

Bos      Chi         -2.5     -4.95     -6.05     -0.77     -5.94      -1.5       -5       0.1      -3.33     -0.83
Cha      Det          3        3.23      5.47      1.4       2.49      2.5       -3.7      -2.2      1.52     -1.48
Dal      Mil          9        6.79      6.9       6.8       6.29       7         10       7.2       7.5       -1.5
Den      Mia         -3.5      0.94      2.29     -0.59      1.54       -2       0.5       -1.3     -0.27      3.24
Hou      Sac         8.5       6.74      8.36      5.9       3.42      12.5      8.6        14       8.5        0
Lal      Cle         10.5      8.52      7.39     11.02      8.45      10.5      8.7       10.1      9.4       -1.1
Noh      Min         -1.5      2.96      4.38     -0.75     -0.57      -0.5       -2       -3.7     -0.21      1.29
Phi      Was          13      17.23     19.96     11.44       16       8.5       -1.6      -2.5     10.25     -2.75
Pho      Njn         6.5       9.98      9.29      7.5       8.17      8.5       -0.1      -2.4      5.93     -0.57
Sas      Por          1        2.67      3.55      3.12     -0.29      2.5       3.4       2.1       2.26      1.26
Tor      Ind         -5.5     -2.55     -1.83     -1.92     -3.25      -2.5      6.2       5.8      -0.69      4.81

Just use Excel for this. You can save the file as formatted text (.prn) and then you can easily copy&paste.

Anyway, something looks wrong with some of the games, because I seriously doubt that RAPM or RigdeRegressed Team thought that Washington would win in Philly. In fact RRTeam had the 76ers winning by 16.3 and RAPM by 17.7 (assuming HCA of 3).

[mystic]

When you can consistently beat Vegas spreads 54% of the time, let me know.

Well, I probably counted wrong the last time. "Beating Vegas" means not being closer to the result, but correctly predicting that a team wins by more than the Vegas Line said. For example: Vegas says +6.5, I say +18, the result is +9, I won. Correct?

If that is the case, my Power Ranking beat Vegas 49 times out of 94 games.

[J.E.]

Does anyone know a good site for injury info? Specifically some up-to-date info whether a player will play in a game, or not.

What I have:
Yahoo Sports
http://sports.yahoo.com/nba/players/262 ... BetHAGPKB4

Sportsbookreview
http://www.sbrforum.com/betting-odds/nba-basketball/ (you need to click the info button next to the game, then injuries)

Unfortunately the info on both of these sites is often quite vague. You'll read things like "probable", "questionable", "doubtful". At SBR, Al Horford is listed as "day to day", I think that's not exactly up-to-date

These are probably a good starting point, but might not be the best. If someone knows better source, please post them

[EvanZ]

You can try Rotoworld, not sure it's better though:

http://www.rotoworld.com/teams/injuries/nba/all/

[EvanZ]

Crow, the usual convention is that home favorites have negative points spreads.

[Crow]

Thanks for the catch. I did copy a formula from a cell to new columns late last night where I should have copied the formulas from a whole column.

Here is the data corrected.

Code: Select all

Home	Visitor	Line	AVG Prediction tracker	SAG	Evan's Vegas Power rankings	Mystic	Betting market analytics	Ridge Regressed Team Rating	player RAPM	Equal blend	Blend-Line
											
Boston	Chicago	-2.5	-4.95	-6.05	-0.77	-5.94	-1.5	0.5	1	-2.53	-0.03
Charlotte	Detroit	3	3.23	5.47	1.4	2.49	2.5	5.5	0	2.95	-0.05
Dallas	Mil	9	6.79	6.9	6.8	6.29	7	7.5	2.5	6.60	-2.40
Denver	Miami	-3.5	0.94	2.29	-0.59	1.54	-2	2.5	-1.5	-0.04	3.46
Houston	Sac	8.5	6.74	8.36	5.9	3.42	12.5	5.5	10	7.62	-0.89
LA Lakers	Clev	10.5	8.52	7.39	11.02	8.45	10.5	4	9.5	8.74	-1.77
New Orl  Minne	-1.5	2.96	4.38	-0.75	-0.57	-0.5	-4	6.5	0.82	2.32
Phila      Washi	13	17.23	19.96	11.44	16	8.5	9	2.5	12.20	-0.80
Phoenix	NewJ	6.5	9.98	9.29	7.5	8.17	8.5	10	1.5	7.68	1.18
San Ant	Port	1	2.67	3.55	3.12	-0.29	2.5	4	1.5	2.26	1.26
Toronto	Indiana	-5.5	-2.55	-1.83	-1.92	-3.25	-2.5	-1	1.5	-2.13	3.37

I tried saving the file as .prn as you suggested but am having problems with that too.

"Crow, the usual convention is that home favorites have negative points spreads." I went with what that predictiontracker site had for the line. Was their line backwards to convention on that one or did I interpret it wrong? I did wonder about it but most looked right the way I was interpreting. I guess I'll have to get another source to sort it out. But I now am having a hard time finding a site that still shows yesterday's line.