Tuesday, September 30, 2014

Tuesday Sept 29: Picks for NCAA Football Week 6

Picks for 10/02/2014

TeamSpreadProb WinnerKelly
UCF+3.061.90%20%
Houston-3.0  


Picks for 10/03/2014

TeamSpreadProb WinnerKelly
Louisville-3.0  
Syracuse+3.075.00%48%
TeamSpreadProb WinnerKelly
San Diego State-1.556.00%8%
Fresno State+1.5  


Picks for 10/04/2014

TeamSpreadProb WinnerKelly
Stanford-2.083.33%65%
Notre Dame+2.0  
TeamSpreadProb WinnerKelly
Florida+2.083.33%65%
Tennessee-2.0  
TeamSpreadProb WinnerKelly
Southern Mississippi+17.5  
Middle Tennessee-17.571.43%40%
TeamSpreadProb WinnerKelly
Nebraska+9.0  
Michigan State-9.066.67%30%
TeamSpreadProb WinnerKelly
California+4.065.31%27%
Washington State-4.0  
TeamSpreadProb WinnerKelly
South Carolina-5.5  
Kentucky+5.561.11%18%
TeamSpreadProb WinnerKelly
Toledo-6.0  
Western Michigan+6.060.00%16%
TeamSpreadProb WinnerKelly
Louisiana-Monroe+9.5  
Arkansas State-9.560.00%16%
TeamSpreadProb WinnerKelly
Oklahoma-5.0  
TCU+5.057.14%10%
TeamSpreadProb WinnerKelly
Oregon State-7.5  
Colorado+7.557.14%10%
TeamSpreadProb WinnerKelly
Michigan+3.0  
Rutgers-3.057.14%10%
TeamSpreadProb WinnerKelly
Miami (Florida)-1.556.00%8%
Georgia Tech+1.5  
TeamSpreadProb WinnerKelly
UAB+9.5  
Western Kentucky-9.554.55%5%
TeamSpreadProb WinnerKelly
Massachusetts+4.0  
Miami (Ohio)-4.054.10%4%
TeamSpreadProb WinnerKelly
North Carolina State+14.0  
Clemson-14.053.57%2%
TeamSpreadProb WinnerKelly
Boise State-2.553.33%2%
Nevada+2.5  
TeamSpreadProb WinnerKelly
Utah+13.052.94%1%
UCLA-13.0  
TeamSpreadProb WinnerKelly
Baylor-16.0  
Texas+16.052.63%1%

Monday, September 29, 2014

Monday Sept 29: Picks for NCAA Football Week 6

Picks for 10/02/2014

TeamSpreadProb WinnerKelly
UCF+2.561.90%20%
Houston-2.5


Picks for 10/03/2014

TeamSpreadProb WinnerKelly
Louisville-3.568.18%33%
Syracuse+3.5
TeamSpreadProb WinnerKelly
San Diego State-1.556.00%8%
Fresno State+1.5


Picks for 10/04/2014

TeamSpreadProb WinnerKelly
Stanford-1.583.33%65%
Notre Dame+1.5
TeamSpreadProb WinnerKelly
Florida-1.0
Tennessee+1.083.33%65%
TeamSpreadProb WinnerKelly
California+3.565.31%27%
Washington State-3.5
TeamSpreadProb WinnerKelly
Michigan+3.565.31%27%
Rutgers-3.5
TeamSpreadProb WinnerKelly
Louisiana-Monroe+10.0
Arkansas State-10.060.00%16%
TeamSpreadProb WinnerKelly
Southern Mississippi+16.5
Middle Tennessee-16.559.26%14%
TeamSpreadProb WinnerKelly
South Carolina-7.5
Kentucky+7.557.14%10%
TeamSpreadProb WinnerKelly
Oklahoma-4.5
TCU+4.557.14%10%
TeamSpreadProb WinnerKelly
Oregon State-7.5
Colorado+7.557.14%10%
TeamSpreadProb WinnerKelly
Nebraska+10.0
Michigan State-10.054.55%5%
TeamSpreadProb WinnerKelly
Massachusetts+4.0
Miami (Ohio)-4.054.10%4%
TeamSpreadProb WinnerKelly
Baylor-14.553.85%3%
Texas+14.5
TeamSpreadProb WinnerKelly
North Carolina State+13.5
Clemson-13.553.57%2%
TeamSpreadProb WinnerKelly
Miami (Florida)-0.0
Georgia Tech+0.053.33%2%
TeamSpreadProb WinnerKelly
Boise State-3.053.33%2%
Nevada+3.0
TeamSpreadProb WinnerKelly
Utah+13.052.94%1%
UCLA-13.0
TeamSpreadProb WinnerKelly
Idaho+14.052.53%0%
Texas State-14.0
TeamSpreadProb WinnerKelly
LSU+8.552.38%0%
Auburn-8.5

Sunday, September 28, 2014

College Football Week 5 Stats Available

The week 5 stats are now available on the stats page.

Sept 27th 1-3 on picks: A Bad Day but Not That Unlikely


My picks for Sept. 27 were pretty bad but will happen from time to time as as I explain below.

First, an overview of my pick outcomes:

$510 (30% of bankroll) on Indiana -3.5 against Maryland
LOSS Maryland beatsIndiana outright 37 – 15
$510 (30% of bankroll) on Purdue +8.5 against Iowa
LOSS Iowa beatsPurdue 24 – 10
$407 (24% of bankroll) on South Carolina -5.5 against Missouri
LOSS Missouri beats South Carolina 21 – 20 outright
$271 (16% of bankroll, rest of bankroll) on Tennessee +17.5 at Georgia
WIN $517.36 While Georgia beat Tennessee 35 – 32 Tennessee covered the generous +17.5 spread

Now my bankroll is $517.36 from $1678.98 or down 71% and I am 2 – 3 for the week.

While the chance of this exact outcome (losing the first three of my picks, and getting the last one) is very unlikely, the probability of only getting one pick right has a low probability, but as unlikely as you may think. Let me explain.

Calculating the probability of events that can only have two outcomes in this case is my pick right or wrong is called binomial. To calculate, the probability that a particular event will happen, in this case of me getting one pick right, you have to add together all the scenarios that causes an event to happen.

So, one of the scenarios of getting one pick right was what happened yesterday, the first three picks were wrong and the last pick was right which is represented by the set, {lllw}.

Where the “l” represents a lose on my part for that particular pick and “w” is a win. So you can read {lllw} as picks 1, 2, and 3 are losses and pick 4 is win. For the event, 3 loses and 1 win, there are four scenarios that can cause that event the one given above and three more: {llwl},{lwll}, and {wlll}. The probability of a 1 – 3 pick day for the probabilities from yesterday's picks would be as follows.

The probability of {lllw} written in shorthand p({lllw}) given my probability estimates: game 1 UI -3.5 vs Maryland 71.43%, game 2 Purdue +8.5 vs Iowa 71.43%, game 3 South Carolina -5.5 vs Missouri 64.00%, and Tennessee +17.5 at Georgia 60.47% would be

p({lllw}) = (1-0.7143) X (1-0.7143) X (1-0.64) X 0.6047 = 0.0177 = 1.77%

So, the scenario {lllw} occurring is very rare, but the probability of only picks one game correct includes the sum of all scenarios so let calculate the other three

p({llwl}) = (1-0.7143) X (1-0.7143) X 0.64 X (1-0.6047) = 0.0206 = 2.06%
p({lwll}) = (1-0.7143) X 0.7143 X (1-0.64) X (1-0.6047) = 0.0290 = 2.90%
p({wlll}) = 0.7143 X (1-0.7143) X (1-0.64) X (1-0.6047) = 0.0290 = 2.90%

Finally, the probability of one win out of three picks this week represnted as p(three losses) is

p(three losses) = p({lllw}) + p({llwl}) + p({lwll}) + p({wlll}) = 0.0963 = 9.63%

While these are probabilities for week 5 of the College Football 2014-2015 season, my pick probabilities for the 4 games are about the same every week (between 60% – 70%), so you can expect any given Saturday that there will be a 10% chance of only getting one pick right (in other words it is most likely to happen a few times a season) and for the dire probability of picking all the games wrong at 1%, but on the flip side I have a 99% of not loses all my money in one week.

So, I will, literally, probably make the money back, the probability of picking 3 or more games correctly is about 60% (or the sum of getting 4 picks correct, ~20%, and getting 3 picks correct, ~40%, on a any given Saturday). In the long run, I will get my money back and it will keep growing.

Saturday, September 27, 2014

Free College Football Statistics Now Available on Dr. Wag Picks


This week I added a statistics page which are the statistics I use for my picks.

Initially, this season I used stats from the Sunshine Forcast, but the page owner Walter Repole has not updated his projections or stats in a while. Last Sunday, I decided to write an program to collect my own stats and spreads. To collect the data, I wrote a computer program called a web crawler, a web crawler traverses a web site and retrieves information, I wrote one that gets publicly available stats from espn.com and the spreads are from www.bettorsworld.com.

In the past seasons, I used stats from CFBStats which used to be free to access current week stats and archived stats, but the site began charging for them this season under the domain http://coachesbythenumbers.com. CFBStats had tons of stats from not only team game performance but individual drive and play-by-play stats as well as stuff like game attendance, and it looks like the pay site has the same data. The packages range from $500/yr to $2000/yr based on how in depth you want the statistics. They had tons of stats so I do not blame them for charging. They did the work, and they should be compensated. My statistics are not as in-depth and only include basic statistics you can get from a regular box score.

I will add stats for additional sports throughout the year.

If you have any other data you would like to see list them in the comments.

Picks for Record Sept. 27, 2014


Of my picks for record for the 2014-2015 NCAA college football season that with week 4, I have gone 4-2 over the spread and with my starting bankroll of $1000 increasing 68% percent to $1678.98.

 Dr. Wag's Lock of the Week


Indiana -3.5 vs Maryland with a 71.43% change of being correct and calculated 40% Kelly criterion

Indiana's strong offense has a mismatch against a relatively weak Maryland offense and this game gives the best value. The rest of the top picks are all have strong offensive mismatches and have a high probability of beating the spread.


Other Picks for Record

  • Purdue +8.5 vs Iowa also with a 71.43% probability of being correct and 40% Kelly bet
  • South Carolina -5.5 vs Missouri with a 64.00% probability of being correct and a 24% Kelly bet
  • Tennessee +17.5 at Georgia with a 60.47% probability of being correct and a 17% Kelly bet

Based on my own betting rules, I will not take put more than 30% of the bankroll on a given game so my bets are as follows with a $1678.98 starting bankroll:

$510 (30% of bankroll) on Indiana -3.5 against Maryland
$510 (30% of bankroll) on Purdue +8.5 against Iowa
$407 (24% of bankroll) on South Carolina -5.5 against Missouri
$271 (16% of bankroll, rest of bankroll) on Tennessee +17.5 at Georgia

Again, I'm betting my whole bankroll but the probability of getting all the picks wrong is low, about 1% (I got the probability from the equation (1-0.7143)X(1-0.7143)X(1-.64)X(1-.6047) or multiplying all the one minus all the probabilities together).

Saturday Sept 27: Picks for NCAA Football Week 5

Picks for 09/27/2014


My algorithm adjusted for the spread increase of the UNLV vs San Diego State game from UNLV -17.5 to -18.5 and thought that was too high, so the Maryland Indiana game moved to the top spot. On the top pretty much every other game stayed the same. 

TeamSpreadProb WinnerKelly
Maryland+3.5
Indiana-3.571.43%40%
TeamSpreadProb WinnerKelly
Iowa-8.5
Purdue+8.571.43%40%
TeamSpreadProb WinnerKelly
Missouri+5.5
South Carolina-5.564.00%24%
TeamSpreadProb WinnerKelly
Tennessee+17.560.47%17%
Georgia-17.5
TeamSpreadProb WinnerKelly
Boise State-12.559.09%14%
Air Force+12.5
TeamSpreadProb WinnerKelly
Texas State+2.5
Tulsa-2.557.14%10%
TeamSpreadProb WinnerKelly
Oregon State+8.0
USC-8.057.14%10%
TeamSpreadProb WinnerKelly
Akron+20.5
Pittsburgh-20.556.25%8%
TeamSpreadProb WinnerKelly
UNLV+18.555.56%7%
San Diego State-18.5
TeamSpreadProb WinnerKelly
Tulane+12.552.94%1%
Rutgers-12.5
TeamSpreadProb WinnerKelly
Bowling Green-5.552.83%1%
Massachusetts+5.5
TeamSpreadProb WinnerKelly
South Alabama-4.052.78%1%
Idaho+4.0
TeamSpreadProb WinnerKelly
Central Michigan+13.552.53%0%
Toledo-13.5
TeamSpreadProb WinnerKelly
Washington State+13.552.53%0%
Utah-13.5

Friday, September 26, 2014

Friday Sept 26: Picks for NCAA Football Week 5

Picks for 09/27/2014


With UCLA crushing ASU by 35 points, my UCLA -4.0 vs ASU was an easy win for me.
Maryland vs. Indiana are trending up from a number six position from last week. But no major changes from last week.

TeamSpreadProb WinnerKelly
UNLV+17.5
San Diego State-17.572.73%43%
TeamSpreadProb WinnerKelly
Maryland+3.5
Indiana-3.571.43%40%
TeamSpreadProb WinnerKelly
Iowa-9.0
Purdue+9.071.43%40%
TeamSpreadProb WinnerKelly
Tulane+11.5
Rutgers-11.563.64%24%
TeamSpreadProb WinnerKelly
South Alabama-4.5
Idaho+4.562.50%21%
TeamSpreadProb WinnerKelly
Missouri+5.060.00%16%
South Carolina-5.0
TeamSpreadProb WinnerKelly
Oregon State+9.5
USC-9.560.00%16%
TeamSpreadProb WinnerKelly
Boise State-12.559.09%14%
Air Force+12.5
TeamSpreadProb WinnerKelly
Central Michigan+14.5
Toledo-14.557.78%11%
TeamSpreadProb WinnerKelly
Texas State+3.0
Tulsa-3.057.14%10%
TeamSpreadProb WinnerKelly
Washington State+13.056.41%8%
Utah-13.0
TeamSpreadProb WinnerKelly
Akron+19.5
Pittsburgh-19.554.05%4%

Thursday, September 25, 2014

Sept 25, 2014: Pick For Record Overview



My bet for tonight's UCLA vs ASU game was -4.0 for UCLA, but the UCLA crushed ASU by 35 points and easily covered the spread.

$320 (Kelly 30%) on UCLA -4.0 against ASU Win UPDATED $290.90

I started with $1069.08 today and I will go into Saturday's picks with $1389.98.

Update 10/17/2014: put my bankroll $1689.98 on a previous post. It is a miscalculation caught by jiceman on Reddit.com, but I don't want to go back and 3 weeks worth of values, so let's just say I won an extra 320 units playing craps and I'm carrying $1689.98 into the next period.

Pick for Record Sept. 25, 2014

My bet for September 25th starting with $1069.08 bankroll based on what I had on Saturday.

$320 (Kelly 30%) on UCLA -4.0 against ASU

UCLA has had an 68% probability of beating the spread for the last two days based on my algorithm, and per my empirical testing 30% max bet of bankroll produced the best results. If you want to read to read more about Kelly betting you should read my post of the topic.

Thursday Sept 24: Picks for NCAA Football Week 5

Picks for 09/25/2014


TeamSpreadProb WinnerKelly
UCLA-4.068.18%33%
Arizona State+4.0

Picks for 09/27/2014


Major movement with UNLV v San Diego State as the top pick. With Iowa and Purdue in the 2nd spot.

TeamSpreadProb WinnerKelly
UNLV+17.5
San Diego State-17.572.73%43%
TeamSpreadProb WinnerKelly
Iowa-9.0
Purdue+9.071.43%40%
TeamSpreadProb WinnerKelly
Missouri+5.5
South Carolina-5.564.00%24%
TeamSpreadProb WinnerKelly
Tulane+12.0
Rutgers-12.063.64%24%
TeamSpreadProb WinnerKelly
South Alabama-4.5
Idaho+4.562.50%21%
TeamSpreadProb WinnerKelly
Maryland+4.560.00%16%
Indiana-4.5
TeamSpreadProb WinnerKelly
Oregon State+9.5
USC-9.560.00%16%
TeamSpreadProb WinnerKelly
Boise State-12.559.09%14%
Air Force+12.5
TeamSpreadProb WinnerKelly
Central Michigan+15.0
Toledo-15.057.78%11%
TeamSpreadProb WinnerKelly
Texas State+3.0
Tulsa-3.057.14%10%
TeamSpreadProb WinnerKelly
Washington State+13.056.41%8%
Utah-13.0
TeamSpreadProb WinnerKelly
Akron+19.5
Pittsburgh-19.554.05%4%
TeamSpreadProb WinnerKelly
Illinois+20.0
Nebraska-20.052.94%1%

Wednesday, September 24, 2014

Wednesday Sept 24: Picks for NCAA Football Week 5

Picks for 09/25/2014


Today is the first time the UCLA vs Arizona State Game has come up on the algorithm picks as a good bet.  

TeamSpreadProb WinnerKelly
UCLA-4.068.18%33%
Arizona State+4.0


Picks for 09/27/2014


Baylor vs Iowa State shots to the top of today's picks. According to my software this is the top pick. Baylor has been on fire this season, but I would be a little hesitant about a bet with such a large spread, but that is why I use computer models for betting because my gut is often wrong.

TeamSpreadProb WinnerKelly
Baylor-20.5
Iowa State+20.570.37%38%
TeamSpreadProb WinnerKelly
Iowa-10.0
Purdue+10.066.67%30%
TeamSpreadProb WinnerKelly
Texas State+4.065.31%27%
Tulsa-4.0
TeamSpreadProb WinnerKelly
Missouri+5.5
South Carolina-5.564.00%24%
TeamSpreadProb WinnerKelly
Tulane+12.0
Rutgers-12.063.64%24%
TeamSpreadProb WinnerKelly
Maryland+4.560.00%16%
Indiana-4.5
TeamSpreadProb WinnerKelly
Oregon State+9.5
USC-9.560.00%16%
TeamSpreadProb WinnerKelly
Boise State-12.559.09%14%
Air Force+12.5
TeamSpreadProb WinnerKelly
Central Michigan+14.5
Toledo-14.557.78%11%
TeamSpreadProb WinnerKelly
Miami (Ohio)+5.0
Buffalo-5.056.67%9%
TeamSpreadProb WinnerKelly
Washington State+12.556.41%8%
Utah-12.5
TeamSpreadProb WinnerKelly
UNLV+17.0
San Diego State-17.056.25%8%
TeamSpreadProb WinnerKelly
Akron+19.5
Pittsburgh-19.554.05%4%
TeamSpreadProb WinnerKelly
South Alabama-5.5
Idaho+5.553.85%3%
TeamSpreadProb WinnerKelly
Illinois+19.5
Nebraska-19.552.94%1%

Tuesday, September 23, 2014

Tuesday Sept 23: Picks for NCAA Football Week 5

Picks for 09/27/2014

Big changes for the top picks for today with the Iowa vs Purdue as the new top pick and Boise State vs Air Force as the number 2 pick.
TeamSpreadProb WinnerKelly
Iowa-10.583.33%65%
Purdue+10.5
TeamSpreadProb WinnerKelly
Boise State-12.080.00%58%
Air Force+12.0
TeamSpreadProb WinnerKelly
UNLV+16.0
San Diego State-16.080.00%58%
TeamSpreadProb WinnerKelly
Illinois+18.0
Nebraska-18.071.43%40%
TeamSpreadProb WinnerKelly
Missouri+5.5
South Carolina-5.564.00%24%
TeamSpreadProb WinnerKelly
Tulane+12.0
Rutgers-12.063.64%24%
TeamSpreadProb WinnerKelly
Maryland+4.560.00%16%
Indiana-4.5
TeamSpreadProb WinnerKelly
Duke+5.060.00%16%
Miami (Florida)-5.0
TeamSpreadProb WinnerKelly
Washington State+10.0
Utah-10.060.00%16%
TeamSpreadProb WinnerKelly
Central Michigan+14.5
Toledo-14.557.78%11%
TeamSpreadProb WinnerKelly
Bowling Green-7.0
Massachusetts+7.055.26%6%
TeamSpreadProb WinnerKelly
Akron+19.5
Pittsburgh-19.554.05%4%
TeamSpreadProb WinnerKelly
South Alabama-6.0
Idaho+6.053.85%3%

Monday, September 22, 2014

Monday Sept 22: Picks for NCAA Football Week 5

Picks for 09/27/2014

Notes: South Alabama v Idaho still top pick but with Duke and Miami with a 0.5 move moves to #2 pick from #8

TeamSpreadProb WinnerKelly
South Alabama-6.5

Idaho+6.572.34%42%
TeamSpreadProb WinnerKelly
Duke+4.0

Miami (Florida)-4.071.43%40%
TeamSpreadProb WinnerKelly
Illinois+18.0

Nebraska-18.071.43%40%
TeamSpreadProb WinnerKelly
Missouri+6.0

South Carolina-6.064.00%24%
TeamSpreadProb WinnerKelly
Tulane+12.0

Rutgers-12.063.64%24%
TeamSpreadProb WinnerKelly
Akron+17.560.47%17%
Pittsburgh-17.5

TeamSpreadProb WinnerKelly
Tennessee+17.560.47%17%
Georgia-17.5

TeamSpreadProb WinnerKelly
Washington State+10.0

Utah-10.060.00%16%
TeamSpreadProb WinnerKelly
Vanderbilt+15.0

Kentucky-15.057.78%11%
TeamSpreadProb WinnerKelly
UNLV+17.0

San Diego State-17.056.25%8%
TeamSpreadProb WinnerKelly
Texas State+5.5

Tulsa-5.555.88%7%
TeamSpreadProb WinnerKelly
Maryland+3.054.55%5%
Indiana-3.0

TeamSpreadProb WinnerKelly
Central Michigan+14.052.53%0%
Toledo-14.0

Sunday, September 21, 2014

Sunday Sept. 21: Picks for NCAA Football Week 5

I'm going to start posting my algorithm picks everyday, and update them as the spread changes. 

Picks for 09/27/2014

TeamSpreadProb WinnerKelly
South Alabama-6.5
Idaho+6.572.34%42%
TeamSpreadProb WinnerKelly
Illinois+18.0
Nebraska-18.071.43%40%
TeamSpreadProb WinnerKelly
Missouri+6.0
South Carolina-6.064.00%24%
TeamSpreadProb WinnerKelly
Tulane+12.0
Rutgers-12.063.64%24%
TeamSpreadProb WinnerKelly
Akron+17.560.47%17%
Pittsburgh-17.5
TeamSpreadProb WinnerKelly
Tennessee+17.560.47%17%
Georgia-17.5
TeamSpreadProb WinnerKelly
Duke+4.560.00%16%
Miami (Florida)-4.5
TeamSpreadProb WinnerKelly
Washington State+10.0
Utah-10.060.00%16%
TeamSpreadProb WinnerKelly
Vanderbilt+15.0
Kentucky-15.057.78%11%
TeamSpreadProb WinnerKelly
UNLV+17.0
San Diego State-17.056.25%8%
TeamSpreadProb WinnerKelly
Texas State+5.5
Tulsa-5.555.88%7%
TeamSpreadProb WinnerKelly
Maryland+3.054.55%5%
Indiana-3.0

Saturday, September 20, 2014

Sept 20 Picks Overview: 3-1


I went 3-1 today and 3-2 for the week.

$240 (Kelly 30%) on Utah State +2.0 against Arkansas State
Loss Arkansas St Pulls the win in OT

$240 (Kelly 30%) on Mississippi State +9.0 against LSU
Mississippi St wins outright 34 – 29 Win $458.18
$216 (Kelly 27%) on Utah +3.5 against Michigan
Utah
beats Michigan outright 26 to 10 Win $412.36
$104 (rest of the bankroll) on Maryland + 2.5 against Syracuse
Maryland crushes Syracuse 34 to 20 Win
$198.54
Bankroll going into next week $1069.08

If I had done even bets on the all the game ($200), I would have a bankroll of $1145.45. While I am lower than if I was even betting, in the long run I will be ahead of even betting with Kelly betting.

Pick Ground Rules and Sept 20th Picks

On this blog, I am also going to record a hypothetical betting for this college football season. The rules for the betting are:
  • I will start the season with a bankroll of $1000 on Sept 19th, 2014, I know this is a day late
  • There will be no maximum to the amount I can bet
  • The minimum bet amount is $5
  • If the bankroll goes below $100, it plus up back to $100
  • The following day bankroll will be the based on the amount of the bet from winnings and loses from the previous day, so even though I could have a compounded bets multiple times on Saturdays due to the stagger of game times, I will consider all games starting at the same time for simplicity
  • All bets will be in whole dollar amount; where $100 would be allowed but not $100.49
  • I will cap the Kelly bet to 30% of the bank roll, I do this for a few of reasons
1.   I’m a conservative better and the idea of betting 60% or 70% on one game makes me squeamish
2.       This allows me to have at least four bets a day therefore hedging my bets a little more
3.       My testing on previous season shows this was a good number for maximum winning
  • I will do no parlays, as all explain on later blogs parlays are not good bets individual game betting has an overall better pay back
Based on this criteria my betting for Sept 19th would have been:
$200 on UConn or 20% of my bankroll based on my picks. UConn lost 14 to 17 to South Florida the spread was UConn +2.5, UConn could not cover the spread and I lost my bet. I will have $800 to bet on Sept 20th
My bets for Sept 20th are based on my algorithmic picks will $800 in the bankroll:
$240 (Kelly 30%) on Utah State +2.0 against Arkansas State
$240 (Kelly 30%) on Mississippi State +9.0 against LSU
$216 (Kelly 27%) on Utah +3.5 against Michigan
$104 (rest of the bankroll) on Maryland + 2.5 against Syracuse
It might be considered risky to bet the whole bankroll, but statistically if my algorithms estimates are correct there is a very small chance I will lose all the bets. The probability of losing all the bets are
(1-0.8333) X (1-0.6667)X(1-0.6531)X(1-0.6190) = 0.0073
So, I only have a 0.7% of losing all the bets. This formula is similar to a binomial distribution where I would have 4 outcomes where none of them are true, but the value of p is different for each pick so the above equation work. 
While I believe that algorithm works based on testing with real data and Monte Carlo testing on simulated outcomes, even if my predictions were totally wrong and the picks are totally by chance the probability of all four being wrong is 6.25% or 0.50 X 0.50 X 0.50 X 0.50 where random picks have a 50% of being correct.  

I get my data from the Sunshine Forecast, and this weeks stats were not updated so my confidence in my predictions is not as high as I would like.

Friday, September 19, 2014

Picks for Sept. 20, 2014

TeamSpreadProb WinnerKelly
Utah State+2.083.33%65.00%
Arkansas State-2.0
TeamSpreadProb WinnerKelly
Mississippi State+9.066.67%30.00%
LSU-9.0
TeamSpreadProb WinnerKelly
Utah+3.565.31%27.14%
Michigan-3.5
TeamSpreadProb WinnerKelly
Maryland+2.561.90%20.00%
Syracuse-2.5
TeamSpreadProb WinnerKelly
Florida+14.0
Alabama-14.061.54%19.23%
TeamSpreadProb WinnerKelly
Northern Illinois+13.5
Arkansas-13.561.54%19.23%
TeamSpreadProb WinnerKelly
San Jose State+10.0
Minnesota-10.060.00%16.00%
TeamSpreadProb WinnerKelly
Hawai'i+8.0
Colorado-8.059.46%14.86%
TeamSpreadProb WinnerKelly
Miami (Florida)+8.0
Nebraska-8.059.46%14.86%
TeamSpreadProb WinnerKelly
Oklahoma-7.5
West Virginia+7.557.50%10.75%
TeamSpreadProb WinnerKelly
Georgia Tech+7.5
Virginia Tech-7.557.14%10.00%
TeamSpreadProb WinnerKelly
North Carolina+3.0
East Carolina-3.055.00%5.50%
TeamSpreadProb WinnerKelly
Georgia Southern-2.554.84%5.16%
South Alabama+2.5

Pick for Sept. 19, 2014

TeamSpreadProb WinnerKelly
Connecticut+2.561.90%20.00%
South Florida-2.5

Kelly Criterion for Fun and Profit

The Kelly criterion a betting optimization formula that tells you what percentage of your bankroll you should bet on a particular wager based on the odds and what probability you estimate you have of winning. The simple formula is

f* = (bp - q)/b

where f* is percent you should wager, b is the odds, p is the probability that the wager is correct and q is the probability the wager is wrong or 1-p.

For most American football and basketball bets for the spread or over/under the odds are 100/110.

For example you want to bet on college football game where you estimate that the probability of beating the spread is 60% (or .6), based on the Kelly criterion you should bet 16% of bankroll.

If the value of f* is less than 0.0 you should not take that bet. For example, you estimate you have a 52% percent change of getting the bet right the value of f* is -0.008. For football and basketball you need to have a probability of .525 or better in either direction to have a positive Kelly bet.

Despite my PhD (or because of it), I did most of my research for this topic on Wikipedia, but there is a good technical paper about by Jane Hung at Washington U. 

There are lots of articles for and against using the Kelly criterion I have found it works quite well with my prediction algorithm where I modeled 9 college football seasons using Kelly bets with a few modifications (basically, I capped the percentage of bankroll at 30%) and my model compounding the winnings, in 8 seasons my model would have won money with average increase over starting bankroll of 40,054.23% (sic). This increase is really high a small sample due to an exceptional year for 2013 model where the model had me winning 22,932.89% of my payroll. If you remove that year there was still a 2140.175% increase over the initial bankroll.

I will include my raw Kelly criterion values on the page.

About This Blog



My name is Greg Wagner, the purpose of this blog is record my betting predictions and to discuss how I made the picks. Initially this blog will only include college football, but I hope to add NFL, college and NBA basketball, baseball, and soccer in that order.

I have always loved sports and I was actively involved with football as a player and later as referee from when I was in middle school until I went to grad school. In grad school I studied computer science eventually graduating with a MS and PhD in computer science from Texas Tech University. My research was in computer vision. You can read more about my thoughts on artificial intelligence and computer science on my sister blog, drwag.blogspot.com.

Currently I teach computer science classes at Pima Community College and the University of Arizona, as well as work as a researcher in industry.

Over the last 2 years I have been working on a computer program that I could use to predict college football games. Based on using real data and based on Monte Carlo testing, I believe the prediction algorithm will work over the long run. This blog will document my picks.