INVESTING 27 NEWS
May 22, 2018
​
Investing 27, formerly known as Sharpshooter Investments, competed in the Skills 21 Capstone Expofest and placed first in Entrepreneurship contest and won overall outstanding Capstone at the Expo. Several judges from the Expofest were given presentations on the business as well as had an extensive view of the website before the contest. To view the award, check skills21.org .
April 7, 2017
​
After testing the 2017 schedule with our new projection formula, we decided to test what the overall profit would be based off of a various number of bets are made. Considering the standard wager is $100, we conducted over 15,000 simulations to calculate the average return on investment based on a specific number of bets. After the mean's were calculated, they were placed on a regression plot to calculate the R-Squared value. The R-Squared value, as indicated on the chart below, is 0.991, with a highest value of 1.0. In other terms, the number of games bet is a strong predictor of return on investment. Not only is Sharpshooter Investment's an investment that produces high rate on returns, but with predictability and consistency.
After completing all of the simulations, we additionally created normal (also called bell) curves to test probability of negative return on investment. If only 1 game is bet in the entire season, there is a 45.63% chance of losing money. However, betting on 256 games in the entire season then that probability significantly decreases to .06%.
A $100 bet is expensive, especially for 256 games. However, even a smaller amount of money can be wagered, but still have the same reliability and potential for high volume.
December 5, 2017
​
After developing Sharpshooter for over several months, I have come to an important realization. A lot of other betting systems claim they include almost every variable including the altitude in weather and which referees are officiating the game. At a distance, it appears that these algorithms are closing the gap between human performance and mathematical projections. However, does a single referee in a lineup cause the New England Patriots to score three more touchdowns than projected? Of course not. After doing many studies with multiple types of regression, we have found that factors such as wind, officials, and even the day of the game have very little to no effect on the score. Including these factors would mean including extraneous variables that do not show a true representation of the team's performance and quality.
As such, we now have changed to a more basic yet logical and efficient approach. In order to truly see which team is the best in an upcoming game, we now look at ELO ratings, home field advantage, season record, past occurrences with both teams, the strength of schedule, and point differentials. These factors matter the most, as they are the most relevant and accurate details to measure the success of a team.
From now until May, we plan to implement these new variables to provide more accurate and relevant data to our algorithms. By the end of testing and making adjustments, the new, redirected algorithm should be functional for the 2018 NFL season.
November 10, 2017
After working for several months on the task, Sharpshooter has created its first prediction variable: modified ELO ratings. ELO ratings are an ideal strategy that base predictions off of different competitors each team faces. In addition, ELO ratings account for the score for each team as well as if they played at home or away. These games do not take account of weather, injuries, referees, or even players on the team. We first developed a formula that would create a win probability based on the team they are playing and if they are playing at home or away. However, this method only works roughly 68% of the time. This is the standard winning percentage for most projection systems on the market and clearly exemplifies how low the floor can be for company's projection systems. We decided this was not good enough, not even close.
We organized how often the ELO rating system is correct for each individual probability output, using the past three NFL seasons as data. For example, when a team has a 90% chance of winning, they actually win 83.33% of the time. In other terms, in the past three seasons, 12 teams have had a winning probability of 90%, but only 10 of those teams have actually won. We did this type of study for each ELO win probability from 0% to 100%. This factor is called the Overall ELO Variable.
Next, we organized how often the ELO rating system is correct for each individual team on a certain winning percentage. Like the Overall ELO Variable, we used the past three seasons of data as well as using a similar analytical approach. Looking at an example of this multi-staged process makes it much clearer. This week, the Seattle Seahawks faced the Arizona Cardinals. The algorithm predicted they had a 58 percent chance of winning based off of previous game history. When the Seattle Seahawks have had a 58 percent chance of winning, they have won 80% of their games. Thus, the Team ELO Variable is found.
We then averaged the two percentages of the Overall and Team ELO Variables to give us the Adjusted Average Score. This provides us with more in depth and accurate prediction.
During Week 10, we first tested the Adjusted Average Score to see how accurate it is. With great success, the new adjusted ELO system predicted 92.86% of the NFL games, having only two incorrect predictions.
While this is a great number, it does not mean this is a full-proof variable. The Adjusted Average Score will be updated as well as tested on past games to determine more validity of the variable.