2007 -- The Florida Gators were getting ready to take on the Ohio State Buckeyes for the BCS National Championship Game. Neither team had faced each other and the oddsmakers penciled the Buckeyes in as an 8pt favorite.
Fans from both teams were fighting for position on message boards, cheering in local sports bars and pubs, while commentators talked about the Gators and their dismal chances to beat a team led by veteran QB, Troy Smith.
In fact, the dominant face staring down at everyone that night was one filled with laughter. How could the BCS committee put the Gators in this position? It was downright absurd, right? Afterall, there was a strong perception that they simply didn't belong there.
Ohio State had won 4 national titles, including one in 2002, and Troy Smith was a heisman candidate. The Buckeyes were undefeated and coming off a 3 point win over #2 rated Michigan. People were clammoring for a rematch in the BCS title game. Florida, on the other hand, had one loss at Auburn mid-season and were not considered an offensive powerhouse of any kind.
Meanwhile, I was sitting on my computer chair looking over the TSRS ratings and trends for the game. I knew the analysts were wrong. I wanted to shake some sense into them.
The ratings and trends didn't lie. They painted a far different picture of these two teams. Florida was playing below their offensive potential and their defense was trending high. Statistically, they had played far better competition and exceeded their marks in those contests. They were not nearly as dispersed as Ohio State when Standard Deviation tracked consistency between the two teams. If Florida played even with their current offensive strength trends, this was going to be a strong win for the Gators. The Gators' defense would be creating a lot of confusion and OSU totals would be low. And, I still wanted to shake the announcers out of their slumber.
At halftime the score was 34-14. In the second half, the Buckeyes didn't score. Troy Smith was harrassed the entire game. The dual-QB tandem of Leak and Tebow chomped on the Buckeyes and Florida looked like a different team. By the time it was over, Florida led everywhere. They led in first downs 21 to 8, time of possession 40:48 to 19:12, and total yards (370 to 82). But, I wasn't shocked at all. The probabilities didn't lie. Straight facts can't be changed into opinions.
Florida was the #1 TSRS Rated team in the nation. Ohio State? OSU was ranked #9. The pollsters had it wrong. The analysts had it wrong. Many of the fans had it wrong. TSRS had it right.
TSRS systematically takes each team apart and finds their relative strengths and weaknesses. The strength algorithm digests each team's raw data, munching on each stat category like a bag full of multi-colored M&Ms, and after comparing the competitive values of each opponent, spits out only the data that matters. No bloated stats remain. When facing an opponent, each stat category is compared against its opposite. If the opponent is stronger in the comparison, and the team facing them does not deviate from its measured potential, the team will increase slightly in its strength index.
TSRS measures more than 40 comparisons during a game. It refactors the overall index by comparing all teams at the end of the week. It also houses a reverie complex because at the end of each week, it goes all the way back to week one and recalculates everything again to measure trending. TSRS is a number crunching robot that utilizes many calculations, including standard deviation, variance, and interquartile range (IQR).
TSRS calculates strength of offensive, defensive, and special teams groupings, rates and assigns strength of schedule, and uses covariance between these groups to measure how much these areas change over time.
Calculating TSRS difference between teams has proven to be very successfull. But, it still wasn't enough. What if two teams were similar or their respective strength ratings were almost the same? I needed to examine two-team scenarios more closely.
I wanted to create an extremely unique probability system. Head -2- Head takes into account all possible scoring opportunites in a game, and then mitigates scoring through complex defensive scenarios, turnovers, and trends. If a team has a Head -2- Head probability value of 62.5%, it would mean that it would account for 62.5% of all scoring in the game.
And, for bragging rights, wouldn't it be nice if you could play any two teams against one another? Not only does it give you a visual snapshot of the probable outcome, it can also be used for future trending. How would each team fare in week 3? In week 8?
Head -2- Head Virtual match-ups allow you to create unlimited game scenarios for any two teams. These match-ups include cross-comparison of all team statistics, strength ratings, historical regression data, risk factors, and trends indicators for each simulation. The forecasting model determines the probability of scoring potential, and provides the results for each match-up.
Many calculations are incorporated into each Head-2-Head Match-up, including TSRS (True Statistical Rating of Strength). The H2H system utilizes strength ratings from red zone calculations to determine what type of score is most likely to occur, if any. It also calculates ball control offense, ball control defense, and probability of turnovers.
Head -2- Head has been used successfully in Division I College Football since 2008 and the NFL was added in 2010. The win probability during the regular season is close to 73% for college and slightly over 69% for NFL games.
By the end of the 2012 season, I wanted to dig through every bit of regression data I had collected and find specific probability value ranges for each one. If I took the top 5 or top 6 trends that were relevant towards game outcomes I could build key differentiators that would expose the risk in every game. This allowed anyone reading the data the ability to understand and define their own risk appetite. I included it in weekly reports and head -2- head match-ups. I saw great success across both sports once the risk mark went below 20%.
In 2014, I read the Lord of the Rings trilogy for the 7th time and a light bulb came on in my head, "One Ring to bring them all and in the darkness bind them."
I asked an interesting question -- What if I could add and combine every possible probability component or tool used in my system together to forecast games? TSRS, H2H, Risk, Trends, Charts, Tables, Indicators, and Scoring Profiles; all combined to create one rating. One rating to rule them all.
Enter Loss Profile Ratings. LPR is calculated using "many" components, all of them proven to have high probability indicators. The higher the LPR rating, the more likely the favored team will win. Once you reach a certain LPR level, the probability jumps incredibly. For the NCAA, any team that had an LPR value of 28 or higher in a division 1 vs division 1 game had the following win%: (456 wins - 9 losses) (98.06%). In the NFL, any team that had an LPR value of 21 or higher in a game had the following win%: (170 wins - 19 losses) (89.95%).
In order to make things simpler visually, I created a chart that highlighted value on one side and risk on the other. So, now I could cross rate and line up any game that met my risk level and my value level at the same time. But, I wanted to do more than just see a visual game line. I added a table (see below) that provided me the number indicators for everything I wanted to see. And, by the way, all of this is explained in detail in the help file of the rated games section. In the example below, I aim for a low risk / medium value profile (LR/MV).
So, where am I today? I'm releasing my tried and true battle tested services. I'm not an invisible person either. I've gone ahead and built my own picks into the site and will showcase them in the rated games section every single week. These are the games "I" like, but for the average person it may take time to understand it all. I've made them available for you to read and review. Also, the number one site visitor here? Me. I'm here developing, creating, reviewing, and making things as clean and crisp as possible. If I'm not here I'm working on the new beta product being released next year - DraftStats. You will be able to use DraftStats to measure player strength and also player schedules to determine whether to play, sit, or trade the player. If you like DraftKings or FanDuel, you will love DraftStats because it offers the ability to create perfect drafts.
Last, you probably want to know who I am? I'm Joel Dezenzio, the owner of JD Stats, LLC and the developer of every system on the site. I'm a die hard programmer (I program in 7 languages), I have a photographic memory, I absorb information like a sponge, and I love people. I really care about the people I meet each day. I have one thought for you as I end this piece and it's from Jimmy V.
I look forward to hearing from you. I read everything! I also respond to everything. :)