Data Dishin' with DJ Dimes: Gamblers Anonymous (Week of 1/11/16)

Cheers Degenerates!

3MW is officially back to bring you the 4th chapter of the Data Dishin' series.  This edition of Gamblers Anonymous includes a brand spanking new section, which covers the best and worst teams against-the-spread (ATS) since the beginning of conference play.   The latter sections contain the weekly updates for the two primary components of the "Gamblers Anonymous" series:  General Betting Trends  & Green Magic results. 


Best & Worst ATS Records (1/1/16 - 1/19/16)

3MW diehards may find this section eerily familiar to the previously written "Vegas Darlings & Duds" , which was posted back in early December.  I have finally set up the Green Magic system to automatically track ATS records for all teams, both at home and on the road.  As a self-acclaimed DJ specializing in all things gambling, I feel it is my born duty to share these ATS results with you the people:


Best ATS Teams

Worst ATS Teams


Some observations & comments on the conference ATS standings thus far:

  • Oklahoma: Despite clinging to the number 1 spot in the current AP poll,  the Sooners may actually be getting TOO much respect from the Vegas oddsmakers. OU has posted a shoddy 1-5 ATS record in their first 6 conference matchups, with the lone cover coming in the epic 3OT showdown at Allen Fieldhouse.  This 1-5 record is ming-boggling, considering they are 4-2 outright over this span, with wins against Iowa St., West Virginia, Kansas St. & Oklahoma St. 
  • Clemson:  The Tigers are starting to make waves in the ACC, after rattling off 5 straight conference wins (Florida St., Duke, @Syracuse, Louisville & Miami), before losing a close one at Virginia last night.  Though 4 of these victories technically came "at home", they were actually played about an hour away from campus in Greeneville, SC (Little John Coliseum is undergoing renovations).  Clemson currently sits at 7-0 ATS in conference, and will host Pitt a week from today at "home".
  • Air Force: I'll admit this is a selfish call-out, which is rooted in complete & utter frustration from my personal wagering on the not-so-mighty Air Force Falcons.  Given I projected Air Force to be a competent, middle-of-the-road Mountain West team, I recently convinced myself that now is an all-time buy low opportunity for the Colorado Spring based military academy.  Per the results above, you can see just how idiotic this rationale was. Not only are the Falcons 0-5 ATS to begin conference play, but they haven't even sniffed a cover in over a month.  The 12.9 average cover margin against them implies that over this 5-game stretch, Air Force has NOT covered by an averag of 13 points a game (!!!!).   Yikes ... 
  • San Jose State:  Even though San Jose continues to be the punching bag of the Mountain West Conference, they are beginning to rake in some nice coin for their gambling backers (if there are any...) thus far in 2016.  In contrast to Oklahoma, San Jose State is still just 1-5 outright in conference action, but picked up a HUGEEEEE win at home against Wyoming. At this point, any and all wins for this program is a step in the right direction, especially after a year where they finished 349th out of 351 teams in Kenpoms final 2015-2016 rankings.
  • Missouri Valley Conference: The point I want to make here is the noticeable gap that is emerging between the top dogs and cellar-dwellers of the MVC.  Over the past decade, this is a league that has been consistently deep & competitive from top-to-bottom and I suspect anyone who has followed the conference, or regularly attends the MVC conference tournament in St. Louis ("Arch Madness"), would wholeheartedly agree.  While the Shockers are back and rolling with a healthy Fred van Vleet (5-1 ATS in their last 6), and Evansville begins to pick up steam, the polar opposite is true for Drake & Bradley.  Bradley is truly in a league of their own, currently sitting at 341st in Kenpoms overall rankings.  If the season ended today, the Braves would officially go down as the worst MVC team in the past 15+ years.  At this point, It would be a major reach to call 1st-year head coach Brian Wardle's move from Green Bay to Peoria, IL a "lateral" move.


General Betting Trends (1/11/16 - 1/19/16)

It's about damn time the "dogs" started barkin...

After home favorites opened 2016 with a strong start,  the "homeless pups" (road dogs) charged back with a 88-70 ATS record last week, generating an 11 unit return.  Hopefully many of you stayed away from sexy home favorites, given their 70-88 record would've put you in a 27 unit hole.

The YTD records (2nd table below) still indicate the oddsmakers in Vegas are absurdly well-trained, given the impressive balance between home, away, public & non-public records.  All results are as follows:


Green Magic Results (1/11/16 - 1/19/16)

I considered making the tagline of this section, "the Demise of Green Magic"... But that would simply be poor marketing for our product here at 3MW...

Make no mistake about it, this was truly a week of regression for Green Magic.  Just ask my buddy and fellow gambler @mbinson888 , who began to serve as a 3rd party case study for the system, and hopped on the Green Magic train at its absolute peak of YTD returns.

Though the overall record was just under .500 for the week, it was the large unit plays that really did the damage.  Now that I have observed day-to-day results for four large Saturday slates, I realize assigning many plays with 3, 4 and even 5 units is foolish and contradicts the true value of a long-term system.  While I enjoyed the gravy train for the first few Saturdays, the most recent Saturday showed the true downside of an overly volatile system, without a proper underlying Value at Risk (VaR) calculation.  For those of you not on Wall Street, the concept of VaR simply refers to the downside risk potential for a portfolio of investments during a given time period (in this case, downside losses for a large Saturday slate).  Using this VaR principle, I adjusted the current GM system to limit the total money at risk on any given day. Going forward, the primary and most obvious change will be much fewer large unit plays.  While the previous Green Magic system included many plays in the 4 & 5 unit range, the current system has a max unit of 4 for any one game.  With these unit adjustments, I backtested the YTD returns for the past ~ 3 week. Shown below is the cumulative return over this timeframe:

Though the spikes and dips are still noticeable for the Saturday slates (1/2, 1/9, 1/16), the daily ups and downs throughout the week appear to be much more steady.  

In addition to reducing the unit size per play, I am also considering reducing the overall number of plays in general.  Shown below is the YTD GM records by unit, with the aforementioned unit conversion tweaks incorporated.  The lowest confidence 1-unit plays are hovering barely above .500, which corresponds to a 12 unit loss for those 286 games (144 + 142 = 286):


In addition to the blah record for the lower confidence plays, larger unit plays, particularly the 4-unit bombs, are also underperforming, relative to the 2 & 3 unit returns.  One common theme I have identified with some of the big unit plays is that they tend to have one historical matchup that is significantly more similar than the other 150 games  considered in the weighted-average similarity, or distance, calculation.  The implication is that this one matchup is heavily influencing the Green Magic projected final score for the current matchup today.  Thus, if the results of that historical matchup were an anomaly (i.e a home team that should've been favored by 10 ends up losing by 15), it will skew the Green Magic projected score in the favor of that historical result.  I am still thinking about the proper way to handle these cases where an eerily similar historical matchup pops up.  I will continue to investigate these large confidence plays, and will call out a few examples in more detail in the next Gamblers Anonymous update.

In the near future, I may begin providing the actual historical game results for all GM plays.  So in addition to just posting a Green Magic plays with corresponding units, the raw historical game results would allow you to look at the specific team characteristics of the top 5 or 10 (or even all 150 for those of you with way too much time on your hands) historical matchups that are similar to today's matchup.  Shown below is an example of what this output looks like for the upcoming Saturday showdown when Baylor hosts Oklahoma:


In addition to the team characteristic data shown above, I could also provide the detailed box score data for all of these specific games.  However, at some point information overload becomes very real, and I believe most of our 3MW readers have alternative lives outside of gambling  (though I could be wrong...).

Please DM me on Twitter at @_3MW_ if you need any further clarification, or if would like to discuss the Green Magic approach in more detail.  Thank you to those who have already reached out directly and expressed interest.