1. 01-06-13 07:21 PM #1
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Model for estimating MLB run totals. Now time to beta test.

    Hi all,

    To follow up with a previous post, I wanted to share the results for my run total prediction model. I spent about 6 months painstakingly processing, parsing, and integrating retrosheet data with historical closing lines from covers.com for the years 2004-2012.

    I know some people bash the covers lines as being "dirty," but I've checked them pretty thoroughly and have found that they are a pretty good reflection of closing lines. I'd also reason that additional value can be found from line shopping and tracking line movement, so if anything I think they represent a pretty conservative estimate of the overall range of lines that are offered prior to game time.

    I tried a lot of "sophisticated" modelling approaches - logistic/linear regressions, SVMs, HMMs, and a method that scaled previous offensive run counts according to the opponent starting pitcher, and applied a bootsrapping step to estimate run totals. They all failed (which was especially disappointing for the bootstrap method; it was actually pretty clever, and I was pretty proud/confident of that when I started coding it).

    The method that worked wonderfully was actually a lot simpler, and only relies on a few (very predictive) statistics. For each season it identified about 200-250 outliers. The results look too good to be true (2004 and 2005 were a little weaker than the rest), but I assure you it is not due to a coding error - I've quadruple checked it all. I've also checked my data files and run permutation tests for betting at random, to ensure that that distribution conforms to a negative expected value. So with that being said, this year will be my "beta testing" - I'm going to post my totals picks here and track the progress as the season goes along. For now I'll post up the running unit totals for seasons 2004-2012, a la the ones found on RAS. Feel free to comment - looking forward to April...

    RunningBalancePlot2004.jpg
    RunningBalancePlot2005.jpg
    RunningBalancePlot2006.jpg
    RunningBalancePlot2009.jpg
    RunningBalancePlot2010.jpg
    RunningBalancePlot2011.jpg
    RunningBalancePlot2012.jpg
    RunningBalancePlot2007.jpg
    RunningBalancePlot2008.jpg
    Points Awarded:

    Chi_archie gave matthew919 1 SBR Point(s) for this post.

    Reply With Quote Reply With Quote
  2. 01-06-13 07:35 PM #2
    greenhippo
    greenhippo's Avatar Become A Pro!
    Join Date: 02-15-12
    Posts: 9,091
    Betpoints: 723

    Will follow if you have a good first month. Nothing wrong with money.
    Reply With Quote Reply With Quote
  3. 01-06-13 08:02 PM #3
    tatehill2000
    tatehill2000's Avatar Become A Pro!
    Join Date: 05-17-07
    Posts: 187
    Betpoints: 11095

    Bro, 1 question, how did it fare in April vers july? I have to change all of my systems after the first month,,, Rather,, I use one set of tool in April, another from May on,,

    How many plays does it yield a day? at what time do the plays present themselves?
    anyway,,,
    cheers
    tater
    Reply With Quote Reply With Quote
  4. 01-06-13 08:05 PM #4
    Louisvillekid1
    LAMAR MVP!
    Louisvillekid1's Avatar Become A Pro!
    Join Date: 10-17-07
    Posts: 52,044
    Betpoints: 553

    good work,

    eyes open
    Reply With Quote Reply With Quote
  5. 01-06-13 08:08 PM #5
    VegasInsider
    Update your status
    VegasInsider's Avatar Become A Pro!
    Join Date: 12-12-10
    Posts: 14,593
    Betpoints: 18

    Interested for sure. Do you have a history of plays?
    Reply With Quote Reply With Quote
  6. 01-06-13 08:16 PM #6
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Tate - April is purely burn in; I only look at games from ~5/1 onward. That's because the model needs about 10 home/away games played to get a good read on each team. But no adjustment is done from that point on. There's an average of 1-2 plays a day from May - October, pretty consistently. Very rarely are there 4 plays on a single day. Some days there are none, though those are pretty rare too. Overall a pretty low variance in that distribution over the length of the season, which I was happy to see.

    The prediction can be calculated as soon as the starting pitchers for both teams are known. So you can have it in hand by the time the line opens. I actually plan to post my plays for the day sometime in the AM, since I'm anticipating those lines will be a little softer. But we'll see. Thanks for the comments.
    Reply With Quote Reply With Quote
  7. 01-06-13 08:44 PM #7
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Vegas - I'll post the 2012 plays below. I apologize if they look terrible, I'm not sure they'll make the transition well. If not, I'll edit this post in the future.

    You'll notice that I only play value bets - with odds better than -110. My criteria for betting an over play is that the mean of the predicted runs has to be at least one run greater than the vegas line, with a median strictly greater than vegas (and same idea for an under play). Let me know if you have any questions.

    Matt


    Date HomeTeam AwayTeam MeanPredictedTotal MedianPredictedTotal OverUnder OverLine UnderLine TotalRuns WonLost Profit
    20120429 CHA BOS 9.428571 9 8 103 -123 5 -1 -1
    20120429 MIN KCA 9.727273 9 8.5 100 -120 11 1 1
    20120429 MIA ARI 9.5 8 7 102 -122 12 1 1.02
    20120430 MIA ARI 8.804878 9 7.5 121 -141 14 1 1.21
    20120502 TBA SEA 11.14286 11 7.5 111 -131 9 1 1.11
    20120503 TBA SEA 11 11 8 -103 -117 7 -1 -1
    20120506 NYN ARI 8.578947 9 7 108 -128 4 -1 -1
    20120508 NYA TBA 6.8 7 9 -111 -109 8 1 0.917431
    20120513 CHA KCA 9.4 9 8 104 -124 10 1 1.04
    20120513 NYA SEA 12.5 11 10.5 -104 -116 8 -1 -1
    20120514 BAL NYA 10.4 10 9 -105 -115 13 1 0.952381
    20120514 BOS SEA 10.5 12.5 9 -103 -117 7 -1 -1
    20120515 BAL NYA 10 9.5 8.5 -105 -115 7 -1 -1
    20120515 SLN CHN 10.71429 10 8.5 -105 -115 13 1 0.952381
    20120516 ATL MIA 9.875 9 7.5 -105 -115 12 1 0.952381
    20120516 TEX OAK 7.4 7 9 -130 110 5 1 1.1
    20120518 NYA CIN 11.44444 11 9.5 -104 -116 4 -1 -1
    20120518 PHI BOS 8.888889 8.5 7 107 -127 10 1 1.07
    20120518 CHN CHA 6.090909 5.5 9 -111 -109 5 1 0.917431
    20120519 PHI BOS 8.521739 9 7.5 -104 -116 12 1 0.961538
    20120519 SDN ANA 9.222222 8 6.5 -103 -117 5 -1 -1
    20120520 PHI BOS 8.571429 8 7 115 -135 6 -1 -1
    20120521 BAL BOS 13.875 15 9 -102 -118 14 1 0.980392
    20120521 HOU CHN 9.307692 9 7 102 -122 12 1 1.02
    20120521 MIL SFN 9.45 8.5 7.5 -102 -118 7 -1 -1
    20120521 ARI LAN 7.846154 7 9 -129 109 7 1 1.09
    20120522 HOU CHN 5.714286 5 8 -113 -107 3 1 0.934579
    20120524 SLN PHI 11.375 13.5 8.5 102 -122 19 1 1.02
    20120525 BOS TBA 11.47619 11 9 109 -129 11 1 1.09
    20120525 MIA SFN 8.646341 8 7 -102 -118 13 1 0.980392
    20120526 MIA SFN 8.8125 8 7.5 104 -124 8 1 1.04
    20120528 ATL SLN 8.870968 8 7.5 102 -122 10 1 1.02
    20120529 TBA CHA 8.55 8.5 7.5 -102 -118 9 1 0.980392
    20120529 CLE KCA 7.291667 7 8.5 -113 -107 10 -1 -1
    20120529 CHN SDN 6.125 5 10 -122 102 8 1 1.02
    20120530 TOR BAL 9.761905 10 8.5 -107 -113 5 -1 -1
    20120530 NYN PHI 8.846154 8 7 -108 -112 16 1 0.925926
    20120601 CHA SEA 9.782609 10 7 -101 -119 11 1 0.990099
    20120602 DET NYA 8.25 7.5 9.5 -118 -102 7 1 0.980392
    20120602 NYN SLN 6.2 5 7.5 -131 111 5 1 1.11
    20120604 PHI LAN 8.25 7.5 6.5 103 -123 7 1 1.03
    20120605 CIN PIT 8.75 8 7.5 -108 -112 12 1 0.925926
    20120605 MIL CHN 8.7 9 7 -103 -117 10 1 0.970874
    20120606 BOS BAL 10.5 11 8.5 -109 -111 3 -1 -1
    20120606 CIN PIT 11.0625 11.5 7.5 -108 -112 9 1 0.925926
    20120606 MIL CHN 10.66667 10.5 7.5 112 -132 8 1 1.12
    20120607 PHI LAN 8.6 9 7 -101 -119 11 1 0.990099
    20120608 CHA HOU 10.37838 11 8.5 -106 -114 11 1 0.943396
    20120608 MIN CHN 12.75 11.5 9 100 -120 15 1 1
    20120609 MIN CHN 6.888889 7 8 -116 -104 14 -1 -1
    20120609 NYA NYN 8.074074 8 10 -116 -104 6 1 0.961538
    20120610 ARI OAK 7.9 8 9 -130 110 7 1 1.1
    20120613 MIN PHI 9.578947 9 8 -103 -117 17 1 0.970874
    20120614 CHN DET 10.25 11 7.5 102 -122 8 1 1.02
    20120614 SLN CHA 10.175 9.5 8.5 -105 -115 8 -1 -1
    20120615 DET COL 12.625 12 10 100 -120 16 1 1
    20120617 NYN CIN 8.571429 8 7.5 113 -133 4 -1 -1
    20120618 ARI SEA 10.66667 12 9 -106 -114 8 -1 -1
    20120618 HOU KCA 10.53846 9 8.5 101 -121 16 1 1.01
    20120619 HOU KCA 9.452381 9 8 103 -123 2 -1 -1
    20120621 DET SLN 8.691176 8.5 10.5 -115 -105 3 1 0.952381
    20120622 ANA LAN 9.733333 11 7.5 114 -134 13 1 1.14
    20120623 ANA LAN 5.75 5 7.5 -117 -103 4 1 0.970874
    20120625 PHI PIT 9.6 11 8 105 -125 11 1 1.05
    20120625 KCA TBA 6.625 6 9 -129 109 8 1 1.09
    20120626 BOS TOR 9.27027 9 10.5 -113 -107 6 1 0.934579
    20120626 MIA SLN 6 7 8.5 -120 100 7 1 1
    20120627 ATL ARI 9.404762 8 7.5 102 -122 10 1 1.02
    20120627 PHI PIT 9.25 11 8 -108 -112 18 1 0.925926
    20120627 CHN NYN 9.388889 9.5 11.5 -119 -101 18 -1 -1
    20120629 MIA PHI 9.240741 9 7.5 -101 -119 8 1 0.990099
    20120629 MIL ARI 11.33333 11 9 102 -122 12 1 1.02
    20120629 MIN KCA 7.857143 8 9.5 -115 -105 7 1 0.952381
    20120630 SFN CIN 8.888889 9 7.5 -101 -119 3 -1 -1
    20120630 TEX OAK 9.8 9.5 11 -122 102 9 1 1.02
    20120703 OAK BOS 9.090909 9 7.5 101 -121 5 -1 -1
    20120703 ARI SDN 10.40909 10 9 103 -123 14 1 1.03
    20120704 ARI SDN 11.38462 11 9 -103 -117 14 1 0.970874
    20120704 LAN CIN 8.695652 8 7 -106 -114 5 -1 -1
    20120704 CHA TEX 9.368421 8 11 -115 -105 9 1 0.952381
    20120704 OAK BOS 6.125 5.5 8 -127 107 5 1 1.07
    20120705 ATL CHN 10.33333 9.5 8 -104 -116 10 1 0.961538
    20120705 SDN CIN 9.222222 8 6.5 105 -125 3 -1 -1
    20120707 BOS NYA 7 6.5 10.5 -117 -103 7 1 0.970874
    20120707 NYN CHN 7.3 7 8.5 -120 100 4 1 1
    20120708 BOS NYA 11.71429 13 9.5 -105 -115 10 1 0.952381
    20120708 NYN CHN 9.666667 9 7.5 -105 -115 7 -1 -1
    20120713 MIA WAS 8.97619 9 7.5 -105 -115 6 -1 -1
    20120713 MIL PIT 10.16667 10 7.5 111 -131 17 1 1.11
    20120714 MIN OAK 10.29412 10 8 104 -124 12 1 1.04
    20120714 ATL NYN 10.42857 10.5 7.5 100 -120 15 1 1
    20120714 KCA CHA 6.625 5.5 8.5 -113 -107 9 -1 -1
    20120714 MIA WAS 6.333333 7 7.5 -114 -106 3 1 0.943396
    20120715 MIA WAS 8.657143 9 7.5 104 -124 4 -1 -1
    20120717 DET ANA 9.244444 8 10.5 -123 103 13 -1 -1
    20120717 NYA TOR 8.354839 8 10 -115 -105 7 1 0.952381
    20120718 LAN PHI 7.533333 6.5 6 -106 -114 8 1 0.943396
    20120718 NYA TOR 9.324324 8 10.5 -119 -101 6 1 0.990099
    20120719 ATL SFN 9 8 7.5 -104 -116 5 -1 -1
    20120719 CHN MIA 8.875 8 7.5 -107 -113 6 -1 -1
    20120720 PHI SFN 9.475 9 8 -103 -117 9 1 0.970874
    20120720 PIT MIA 6.375 7 8.5 -124 104 7 1 1.04
    20120722 BOS TOR 10.5 9.5 9 -108 -112 22 1 0.925926
    20120723 NYN WAS 9.052632 9 8 106 -126 10 1 1.06
    20120723 SFN SDN 7.857143 9 6.5 -101 -119 8 1 0.990099
    20120724 CLE DET 10.63158 11 9 107 -127 5 -1 -1
    20120724 MIA ATL 9.361111 8.5 7.5 103 -123 7 -1 -1
    20120724 PHI MIL 10.33333 12 7.5 102 -122 13 1 1.02
    20120724 BAL TBA 7.133333 6 8.5 -135 115 4 1 1.15
    20120724 TEX BOS 8.62069 8 10.5 -117 -103 3 1 0.970874
    20120725 NYN WAS 10.30435 10 7.5 108 -128 7 -1 -1
    20120726 ARI NYN 7.666667 7 9 -124 104 4 1 1.04
    20120727 ATL PHI 9.086207 8 7.5 -105 -115 7 -1 -1
    20120727 SFN LAN 8.666667 9 6.5 -101 -119 8 1 0.990099
    20120727 BAL OAK 6.65 6 8.5 -117 -103 23 -1 -1
    20120728 NYA BOS 10.08333 11 9 107 -127 14 1 1.07
    20120729 ATL PHI 8.731707 8 7.5 -103 -117 8 1 0.970874
    20120730 ATL MIA 8.846154 9 7.5 -102 -118 10 1 0.980392
    20120730 MIL HOU 11.66667 11 8.5 100 -120 15 1 1
    20120731 OAK TBA 9 8 7 117 -137 8 1 1.17
    20120731 CIN SDN 9.958904 10 8.5 108 -128 13 1 1.08
    20120802 BOS MIN 6.3 4.5 10 -113 -107 5 1 0.934579
    20120803 OAK TOR 9.363636 8 7 -107 -113 9 1 0.934579
    20120803 SDN NYN 8.111111 8 6.5 -103 -117 4 -1 -1
    20120803 WAS MIA 9.533333 11 8.5 -105 -115 11 1 0.952381
    20120804 WAS MIA 8.916667 8.5 7.5 115 -135 17 1 1.15
    20120805 NYA SEA 8.8 8 10 -117 -103 8 1 0.970874
    20120806 MIL CIN 9.826087 9 8 100 -120 9 1 1
    20120806 SLN SFN 10.45 9 8 -102 -118 10 1 0.980392
    20120808 PIT ARI 11.28571 10 8 106 -126 13 1 1.06
    20120809 CLE BOS 10.06667 9.5 9 -101 -119 8 -1 -1
    20120809 CHN CIN 9.25 8 7.5 -103 -117 8 1 0.970874
    20120809 SLN SFN 9.964286 9.5 7.5 -107 -113 4 -1 -1
    20120810 ARI WAS 10.81818 11 8 -102 -118 10 1 0.980392
    20120810 CHN CIN 8.205882 8 6.5 -105 -115 18 1 0.952381
    20120811 ARI WAS 11.64286 10.5 8.5 -105 -115 11 1 0.952381
    20120812 ARI WAS 10.70833 10 9 111 -131 11 1 1.11
    20120812 HOU MIL 9.625 8.5 7.5 -106 -114 8 1 0.943396
    20120812 ANA SEA 4.166667 3 7.5 -124 104 5 1 1.04
    20120813 ATL SDN 9.514286 9 8.5 -107 -113 5 -1 -1
    20120813 SFN WAS 7.625 7 6.5 107 -127 16 1 1.07
    20120814 SFN WAS 8.307692 7 6.5 115 -135 7 1 1.15
    20120814 BAL BOS 7.225 7 8.5 -122 102 8 1 1.02
    20120815 TOR CHA 10.375 11 8.5 -107 -113 14 1 0.934579
    20120815 CIN NYN 8.964286 9.5 7.5 -102 -118 7 -1 -1
    20120815 CHN HOU 6.555556 6 8.5 -113 -107 9 -1 -1
    20120816 TOR CHA 11.275 11 8.5 -101 -119 9 1 0.990099
    20120817 ANA TBA 9 8.5 7 102 -122 15 1 1.02
    20120817 NYA BOS 10.57895 10 9.5 100 -120 10 1 1
    20120819 DET BAL 9.153846 8 7.5 -107 -113 12 1 0.934579
    20120820 ARI MIA 11.25 12.5 9 104 -124 15 1 1.04
    20120820 LAN SFN 10.25 11 6 104 -124 3 -1 -1
    20120820 MIL CHN 11.8 11.5 9 104 -124 14 1 1.04
    20120820 SDN PIT 6.461538 6 7.5 -123 103 4 1 1.03
    20120821 LAN SFN 9.666667 9.5 7.5 105 -125 5 -1 -1
    20120821 NYN COL 9.5 8.5 7.5 -108 -112 8 1 0.925926
    20120822 LAN SFN 7.769231 7 6.5 102 -122 12 1 1.02
    20120824 BOS KCA 10.78947 11 9 103 -123 7 -1 -1
    20120824 CHA SEA 9.512821 9 8 -109 -111 17 1 0.917431
    20120824 TBA OAK 8.3 7.5 7 106 -126 9 1 1.06
    20120824 CHN COL 8.058824 7 9.5 -129 109 8 1 1.09
    20120825 SFN ATL 7.625 7 6.5 109 -129 10 1 1.09
    20120826 CHA SEA 10.51351 11 9.5 106 -126 7 -1 -1
    20120826 CIN SLN 9.6 9 8.5 -108 -112 10 1 0.925926
    20120826 PHI WAS 8.964286 9 7.5 109 -129 5 -1 -1
    20120826 SFN ATL 11.5 12 7 113 -133 8 1 1.13
    20120827 NYA TOR 12.5 12.5 9.5 103 -123 15 1 1.03
    20120828 ANA BOS 8.903226 9 7.5 -102 -118 11 1 0.980392
    20120828 ARI CIN 12.44444 12 8 -103 -117 7 -1 -1
    20120828 MIA WAS 10.125 11 7.5 106 -126 9 1 1.06
    20120828 PHI NYN 10.12821 10 8.5 -107 -113 14 1 0.934579
    20120828 KCA DET 6.444444 6.5 7.5 -121 101 17 -1 -1
    20120830 WAS SLN 9.052632 9 8 104 -124 9 1 1.04
    20120831 ATL PHI 9.478261 8.5 7 -102 -118 13 1 0.980392
    20120901 DET CHA 11.03846 10 8.5 -103 -117 6 -1 -1
    20120901 KCA MIN 7.105263 6 9 -124 104 4 1 1.04
    20120902 CLE TEX 10.23333 10.5 8.5 105 -125 11 1 1.05
    20120902 ATL PHI 9 8 7 -106 -114 15 1 0.943396
    20120902 CHN SFN 8.575 9 7 -108 -112 12 1 0.925926
    20120902 MIL PIT 9.548387 9 7.5 -106 -114 20 1 0.943396
    20120904 PIT HOU 11 12 7.5 -107 -113 8 1 0.934579
    20120905 PIT HOU 10 11.5 8.5 -105 -115 9 1 0.952381
    20120905 SFN ARI 8.258065 8 6.5 -107 -113 8 1 0.934579
    20120905 SLN NYN 9.307692 9 7.5 107 -127 8 1 1.07
    20120907 CHA KCA 10.64 9 8.5 -103 -117 12 1 0.970874
    20120909 MIN CLE 11.5 10.5 9 -108 -112 15 1 0.925926
    20120910 ANA OAK 9.416667 9 8 -107 -113 4 -1 -1
    20120910 CIN PIT 8.910448 8 7.5 -106 -114 7 -1 -1
    20120911 TEX CLE 11.46667 12 9.5 -109 -111 10 1 0.917431
    20120912 MIN KCA 11.45833 11 9 107 -127 15 1 1.07
    20120914 CLE DET 9.9 10.5 7.5 -106 -114 4 -1 -1
    20120914 ARI SFN 7.73913 7 9 -111 -109 8 1 0.917431
    20120916 ATL WAS 9.066667 8 7.5 107 -127 6 -1 -1
    20120918 CLE MIN 10.45455 10 9 -102 -118 11 1 0.980392
    20120918 KCA CHA 7.192308 6 8.5 -117 -103 5 1 0.970874
    20120918 MIA ATL 7.12 6 8.5 -117 -103 7 1 0.970874
    20120919 ARI SDN 7.315789 7.5 9.5 -115 -105 8 1 0.952381
    20120919 CHN CIN 9.47619 8.5 14 -115 -105 11 1 0.952381
    20120919 NYN PHI 5.727273 4 7 -120 100 5 1 1
    20120920 DET OAK 9.275 9 8 -105 -115 16 1 0.952381
    20120920 CHN CIN 7.181818 6.5 8.5 -130 110 8 1 1.1
    20120921 ANA CHA 10.15789 8 7.5 102 -122 8 1 1.02
    20120922 BOS BAL 10.53333 11 9.5 102 -122 15 1 1.02
    20120922 NYA OAK 10.42857 10 9 -106 -114 19 1 0.943396
    20120922 TBA TOR 8.235294 8 7 -102 -118 16 1 0.980392
    20120922 WAS MIL 5.0625 4 7.5 -111 -109 14 -1 -1
    20120923 NYN MIA 9.1 9 7.5 -102 -118 5 -1 -1
    20120924 DET KCA 9.181818 9 7.5 110 -130 8 1 1.1
    20120925 CHA CLE 11.48 10 9 -102 -118 7 -1 -1
    20120926 BAL TOR 10.80645 11 8.5 -108 -112 14 1 0.925926
    20120926 MIN NYA 10.125 9 8 -107 -113 10 1 0.934579
    20120926 CIN MIL 9.630769 9 8.5 107 -127 9 1 1.07
    20120926 SFN ARI 8.636364 7.5 6.5 -107 -113 6 -1 -1
    20120927 ANA SEA 9.346154 8.5 8 102 -122 13 1 1.02
    20120927 ATL MIA 9.066667 8 7.5 -108 -112 8 1 0.925926
    20120927 SFN ARI 9.615385 9 7.5 106 -126 10 1 1.06
    20120928 MIL HOU 11.28571 12 8 -105 -115 13 1 0.952381
    20120929 ARI CHN 11.33333 10 9.5 108 -128 10 1 1.08
    20120930 TEX ANA 10.2 9 8.5 102 -122 9 1 1.02
    20120930 ARI CHN 11.14286 11 9.5 -102 -118 9 -1 -1
    20120930 SDN SFN 9.071429 8 7.5 107 -127 12 1 1.07
    20121001 NYA BOS 9.828571 9 8 -105 -115 12 1 0.952381
    20121001 SLN CIN 9.3125 9 7.5 -109 -111 6 -1 -1
    20121001 OAK TEX 6.944444 7 8 -132 112 7 1 1.12
    20121003 LAN SFN 8.8 9.5 6.5 105 -125 6 -1 -1
    20121003 MIL SDN 9.924528 9 8.5 -108 -112 13 1 0.925926
    Last edited by matthew919; 01-06-13 at 09:34 PM.
    Reply With Quote Reply With Quote
  8. 01-06-13 09:12 PM #8
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    And here are the summary stats for all years. The OverUnderWinningTotal column is units won/lost. The PermXXX columns give the number of over/under bets and unit profit for the permutation testing, which bets at random. I tried to keep the number of bets in permutation tests pretty close to the average number of over/under bets from an actual season, for sake of comparison.

    Also, a note about the units - I don't believe in attaching 2, 3, or more units to a single wager. If you're following good bankroll management, there's only so much you can have in play at a given time anyway, so calling something a "3 unit pick" is nonsense IMO. I've sort of mimicked the way RAS presents their running totals, which in my opinion is the most "honest" way of looking at your record. Otherwise, what's stopping someone from calling every wager a 5 unit play and quintupling their profit?

    Anyway, my recommendation would be to have no more than 5% of the bankroll in play at any one time, and treat each of my plays as having an equal edge (if you're using a Kelly system, I'd play it ultra-conservatively, and call the edge 5% or something like that). So for instance, if your current bankroll is 10K, and there are two simultaneous games to wager on that day, put 2.5% = $250 on each. For 1 game: wager $500. Adjust the unit size daily.




    NumOverWagers NumUnderWagers OverUnderWinningTotal PermNumOverWagers PermNumUnderWagers PermWinningTotal
    2004 124 79 18.75714 135.8425 90.5665 -8.80493
    2005 179 66 -1.48238 136.1713 90.7282 -10.4519
    2006 141 71 60.6297 135.253 90.152 -4.0892
    2007 114 92 90.04877 136.8066 91.1948 -4.26833
    2008 142 82 92.60003 138.6729 92.4264 -4.79366
    2009 161 89 68.06341 137.0826 91.4157 -7.13625
    2010 166 64 93.46999 132.6637 88.3972 -6.52997
    2011 119 83 88.79669 136.7699 91.2062 -3.41189
    2012 174 52 106.6793 137.153 91.4163 -6.0613
    Last edited by matthew919; 01-07-13 at 12:07 AM.
    Reply With Quote Reply With Quote
  9. 01-07-13 07:48 AM #9
    dlunc3
    dlunc3's Avatar Become A Pro!
    Join Date: 10-31-09
    Posts: 9,129
    Betpoints: 1076

    just so i am reading this correctly..

    You are not releasing the aspects of the system with the hopes of it doing well so that you can eventually go tout-- is that correct?
    Reply With Quote Reply With Quote
  10. 01-07-13 07:59 AM #10
    EXhoosier10
    EXhoosier10's Avatar Become A Pro!
    Join Date: 07-06-09
    Posts: 3,122
    Betpoints: 4390

    Who cares if he goes tout? If you have a system that finishes a year +50 units 5 out of 7 years, you'd be dumb not to go tout.
    Reply With Quote Reply With Quote
  11. 01-07-13 08:08 AM #11
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    That's a possibility. Obviously I'm planning to bet these picks myself, but I'm currently a broke grad student and my starting bankroll is very small. My aim is to build a BR that I can work effectively with, and to do it as quickly as possible. So if people like the results from 2013 and are willing to pay for picks in the future, I'd be willing to sell. I started this thread just for validation purposes; nothing wrong with having options.
    Reply With Quote Reply With Quote
  12. 01-07-13 08:22 AM #12
    sweep
    USA! USA! USA!
    sweep's Avatar Become A Pro!
    Join Date: 10-09-10
    Posts: 16,753
    Betpoints: 2457

    Reply With Quote Reply With Quote
  13. 01-07-13 08:31 AM #13
    VegasInsider
    Update your status
    VegasInsider's Avatar Become A Pro!
    Join Date: 12-12-10
    Posts: 14,593
    Betpoints: 18

    Quote Originally Posted by matthew919 View Post
    That's a possibility. Obviously I'm planning to bet these picks myself, but I'm currently a broke grad student and my starting bankroll is very small. My aim is to build a BR that I can work effectively with, and to do it as quickly as possible. So if people like the results from 2013 and are willing to pay for picks in the future, I'd be willing to sell. I started this thread just for validation purposes; nothing wrong with having options.
    Goodluck to you, sir. I'm looking forward to backing these plays in the spring. Thanks
    Reply With Quote Reply With Quote
  14. 01-07-13 08:49 AM #14
    runnershane14
    runnershane14's Avatar Become A Pro!
    Join Date: 07-23-07
    Posts: 803
    Betpoints: 84

    does your model account for bullpen pitching?
    Reply With Quote Reply With Quote
  15. 01-07-13 08:59 AM #15
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Shane - yes, although not directly. My opinion is that it's best to sweep as much under the rug as possible when letting the data drive the model. So there are a lot of small-effect variables that are smoothed over in the statistics that I use, and that's one of them. Overall, most teams' bullpens are pretty even (with some exceptions - the 2010 ARI team comes to mind), and since there's no way to know how relief pitching will pan out, trying to model variables like this directly will generally lead to overfitting.
    Reply With Quote Reply With Quote
  16. 01-07-13 09:55 AM #16
    tbird509
    tbird509's Avatar Become A Pro!
    Join Date: 11-05-09
    Posts: 761
    Betpoints: 3256

    ill be tracking may1st gl
    Reply With Quote Reply With Quote
  17. 01-07-13 10:57 AM #17
    runnershane14
    runnershane14's Avatar Become A Pro!
    Join Date: 07-23-07
    Posts: 803
    Betpoints: 84

    Quote Originally Posted by matthew919 View Post
    Shane - yes, although not directly. My opinion is that it's best to sweep as much under the rug as possible when letting the data drive the model. So there are a lot of small-effect variables that are smoothed over in the statistics that I use, and that's one of them. Overall, most teams' bullpens are pretty even (with some exceptions - the 2010 ARI team comes to mind), and since there's no way to know how relief pitching will pan out, trying to model variables like this directly will generally lead to overfitting.
    thats basically what I have been thinking..basically even across teams and hard to model/overfit
    Reply With Quote Reply With Quote
  18. 01-07-13 11:00 AM #18
    runnershane14
    runnershane14's Avatar Become A Pro!
    Join Date: 07-23-07
    Posts: 803
    Betpoints: 84

    have you considered results over 1st 5 inning totals?
    Reply With Quote Reply With Quote
  19. 01-07-13 11:02 AM #19
    runnershane14
    runnershane14's Avatar Become A Pro!
    Join Date: 07-23-07
    Posts: 803
    Betpoints: 84

    Quote Originally Posted by matthew919 View Post
    Tate - April is purely burn in; I only look at games from ~5/1 onward. That's because the model needs about 10 home/away games played to get a good read on each team. But no adjustment is done from that point on. There's an average of 1-2 plays a day from May - October, pretty consistently. Very rarely are there 4 plays on a single day. Some days there are none, though those are pretty rare too. Overall a pretty low variance in that distribution over the length of the season, which I was happy to see.

    The prediction can be calculated as soon as the starting pitchers for both teams are known. So you can have it in hand by the time the line opens. I actually plan to post my plays for the day sometime in the AM, since I'm anticipating those lines will be a little softer. But we'll see. Thanks for the comments.
    what does no adjustment from that point on mean?
    Reply With Quote Reply With Quote
  20. 01-07-13 11:13 AM #20
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Just that the parameters that are used remain the same, with no change in how they are computed or weighted. And no, I'm only interested in looking at full game totals- my understanding is that finding historical odds for 1st 5 inning for benchmarking purposes would be impossible. Why do you ask? Have you tried modeling that? If so, do you have a good resource for historical odds?

    Matt
    Reply With Quote Reply With Quote
  21. 01-07-13 12:58 PM #21
    runnershane14
    runnershane14's Avatar Become A Pro!
    Join Date: 07-23-07
    Posts: 803
    Betpoints: 84

    i haven't tried modeling that..it just seemed the way you described your model that it potentially could be profitable in the 1st 5. www.sbrodds.com has 1st 5 inning totals back to 2009 (4 years worth)
    Reply With Quote Reply With Quote
  22. 01-07-13 01:11 PM #22
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Thanks man, wasn't even aware of that. This will make a nice little side project, assuming I can find enough free time before April.

    Cheers.
    Reply With Quote Reply With Quote
  23. 01-07-13 01:12 PM #23
    gregm
    gregm's Avatar Become A Pro!
    Join Date: 03-14-11
    Posts: 3,535
    Betpoints: 121

    I am not really sure what you are looking for in terms of data or historical odds on 1st five innings but I would think oddsportal would probably be the place to look.
    Reply With Quote Reply With Quote
  24. 01-07-13 01:13 PM #24
    gregm
    gregm's Avatar Become A Pro!
    Join Date: 03-14-11
    Posts: 3,535
    Betpoints: 121

    Quote Originally Posted by runnershane14 View Post
    i haven't tried modeling that..it just seemed the way you described your model that it potentially could be profitable in the 1st 5. www.sbrodds.com has 1st 5 inning totals back to 2009 (4 years worth)
    good stuff. I didnt know that either. thanks
    Reply With Quote Reply With Quote
  25. 01-08-13 03:28 PM #25
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Going back to the "units" discussion and recommended bankroll management for these plays, I decided to make some new plots which reflect each season's running bankroll, assuming 5K to start and wagering 5% of BR at a given time (with the 5% wager size split evenly for days with multiple bets).

    Here they are. Like I said, 2004 and 2005 were not exactly spectacular years (in comparison to the others), but of all 8 seasons tested, 2005 was the only negative ROI. I'm hoping that this has something to do with the residual effects of the steroid era, where the predictions just aren't as good (2005 was also the year the Expos moved to WAS...) But I'll admit it's more likely that it was just a particularly bad year. 2012 was the best season overall; the bankroll plot for that year just blows my mind.



    RunningBalancePlotWithSemiKelly2004.jpg
    RunningBalancePlotWithSemiKelly2005.jpg
    RunningBalancePlotWithSemiKelly2006.jpg
    RunningBalancePlotWithSemiKelly2007.jpg
    RunningBalancePlotWithSemiKelly2008.jpg
    RunningBalancePlotWithSemiKelly2009.jpg
    RunningBalancePlotWithSemiKelly2010.jpg
    RunningBalancePlotWithSemiKelly2011.jpg
    RunningBalancePlotWithSemiKelly2012.jpg
    Last edited by matthew919; 01-08-13 at 03:33 PM.
    Reply With Quote Reply With Quote
  26. 02-19-13 04:41 PM #26
    dlunc3
    dlunc3's Avatar Become A Pro!
    Join Date: 10-31-09
    Posts: 9,129
    Betpoints: 1076

    Getting all geared up and ready?
    Reply With Quote Reply With Quote
  27. 04-16-13 09:48 PM #27
    dlunc3
    dlunc3's Avatar Become A Pro!
    Join Date: 10-31-09
    Posts: 9,129
    Betpoints: 1076

    what happened to this?
    Reply With Quote Reply With Quote
  28. 04-16-13 10:04 PM #28
    daneault23
    daneault23's Avatar SBR PRO
    Join Date: 09-08-09
    Posts: 3,796
    Betpoints: 4955

    Was wondering what happened to this thread myself
    Reply With Quote Reply With Quote
  29. 04-17-13 08:29 AM #29
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Sorry guys, I decided not to post these after all. The other day I posted a link to my blog (which was then speedily quarantined to SBR's subforum equivalent of an underground Siberian bunker- I guess the mods don't approve of that sort of thing).

    In any case, that's where I'll be doing all of my writing/summary analysis/random stuff. Not day by day plays, but mostly insights on statistical modelling and qualitative observations (with some funny mixed in, hopefully). Which is useless if you're looking to tail. But like they say: Feed a man a fish, feed him for a day. Teach a man to fish...
    Reply With Quote Reply With Quote
  30. 04-17-13 08:51 AM #30
    oZgoodZ33
    oZgoodZ33's Avatar Become A Pro!
    Join Date: 02-28-12
    Posts: 525
    Betpoints: 0

    Sooooo why exactly aren't you going to post plays? Because the mods moved your blog link, really? Really?
    Reply With Quote Reply With Quote
  31. 04-17-13 09:08 AM #31
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Oh, not butthurt at all- that's got nothing to do with it. Mods are just doing their jobs. Sorry if that wasn't clear in the last post. I just want to focus on writing more qualitative stuff (I enjoy it more), and there are only so many hours in the day.

    Anyway, the lines for BOS/CLE and OAK/HOU moved down slightly from the open, but they are still too high IMO.
    There- plays. I'm exhausted now. Time for a nap.
    Reply With Quote Reply With Quote
  32. 04-17-13 09:22 AM #32
    sweep
    USA! USA! USA!
    sweep's Avatar Become A Pro!
    Join Date: 10-09-10
    Posts: 16,753
    Betpoints: 2457

    Quote Originally Posted by matthew919 View Post
    Oh, not butthurt at all- that's got nothing to do with it. Mods are just doing their jobs. Sorry if that wasn't clear in the last post. I just want to focus on writing more qualitative stuff (I enjoy it more), and there are only so many hours in the day.

    Anyway, the lines for BOS/CLE and OAK/HOU moved down slightly from the open, but they are still too high IMO.
    There- plays. I'm exhausted now. Time for a nap.
    Reply With Quote Reply With Quote
  33. 04-17-13 09:33 AM #33
    BiffTFinancial
    BiffTFinancial's Avatar Become A Pro!
    Join Date: 01-29-09
    Posts: 22,670
    Betpoints: 5368

    Quote Originally Posted by oZgoodZ33 View Post
    Sooooo why exactly aren't you going to post plays? Because the mods moved your blog link, really? Really?
    Not looking to get into it with you personally (since many do it), but people absolutely must stop using "really?" as a punchline. I weep for the future when that is what passes as argument and intellectual rigor. I generally do not support the death penalty, but if it is to be used, it should be used on Seth Myers for the proliferation of this trite platitude (also, Oprah for general crimes against good taste and decency).
    Last edited by BiffTFinancial; 04-17-13 at 09:48 AM.
    Points Awarded:

    EVPlus gave BiffTFinancial 2 SBR Point(s) for this post.

    Reply With Quote Reply With Quote
  34. 04-17-13 09:35 AM #34
    YankThunder06
    YankThunder06's Avatar Become A Pro!
    Join Date: 04-01-13
    Posts: 75
    Betpoints: 0

    REALLY? LOL
    Reply With Quote Reply With Quote
  35. 04-17-13 09:43 AM #35
    matthew919
    Update your status
    matthew919's Avatar Become A Pro!
    Join Date: 11-21-12
    Posts: 421
    Betpoints: 5869

    Just for the record- I'd lay off BOS/CLE under at this point. Value is gone after that 20 cent move. I'd still take OAK/HOU at under 8 -105 though. But no lower than u8 -110.

    Ok, nap time for real now.
    Reply With Quote Reply With Quote
1234 ... Last
Top
« Previous Thread | Players Talk | Next Thread »