​

About Me: Analytics Geek Since Strat-O-Matic Football

I'm not one of those NFL talking head ‘experts’, but I've been participating in the same confidence pool for the last 15+ years. As an Engineer and analytical ‘numbers’ person, over the years I developed and have continuously optimized computer models based on multiple variables that when combined, predict the picks and the confidence points for each game each week.

 

For the price of a cup of Starbucks coffee, you can have access to all 5 of my models and get great picks each week for the entire season, with significantly better performance than some of the top confidence pool sites at a fraction of what they charge.

​​​​​​

2024 Inaugural Season

​

Past Results: Coulda Been a Contender

Over the last 15+ years, I have won my fair share of weekly, and end of season rewards out of a ~70 person pool. But If I'm being brutally honest, I haven’t won more than I have spent on league fees over that duration for one simple reason, I let my personal opinion override a few of my computer model’s picks and points each week!
 

I Decided To Investigate: What If I Trusted My Model?
That said, in 2023 I still overrode my computer model with my personal opinions, but I finally started tracking my performance (computer picks with my stupid opinions) vs. just what my computer model predicted with no emotion. Guess what I found?
 

2023 Confidence Pool Results: 11.25X My League Fee! 
Despite having a pretty good year with my overrides picks (I was on track to finish 9th overall at 1515), I finished in 20th overall at 1486. As I'm sure you've all done, since I was too far behind to place in the overall standings (top 5), I spent the last 4 weeks picking a few aggressive upset picks to try to pickup some weekly rewards. That strategy paid off by finishing 2nd in week 17 (half the league fee in rewards).

​

My untouched computer model picks, however, would have finished 2nd overall at 1551, and would have also finished 1st in week 9. Combined, my rewards would have totaled 11.25 TIMES my league fee!

​

Further Analysis: 99.95% Percentile in 2023!

Not only did my computer model perform well in my Confidence Pool, but as I studied it in more detail, I found out that my model's Overall win percentage was 68.8%. The best of the 287 'experts' on Pickwatch in 2023 finished at 68.0% and only 4 'fans' out of 7,519 on Pickwatch picked better than my model. As for Favorites, my model was at 69.8% while the NFL Average is 66.2%. But my model's best performance was reserved for the Underdogs coming in at 56.3%, while the NFL Average is 33.8%. ​​​​​​​​​​​​​

​

Unexpected Benefit: Underdog Moneyline Picks

This observation leads me to a secondary benefit that can be derived from my model: Applying the model's Underdog picks to the Moneyline since I picked anywhere from 8-14% better that the breakeven point for underdogs (depending on the odds).  I separately plan to track that performance over 2024, and personally apply it on every Underdog my model picks this season to see how much additional value is possible.​​

​

2025 Season Two 

​

Last Year’s Performance 
Despite improving my overall win percentage from 68.8% to 71.7% and accruing 127 more points in 2024 than 2023'a computer model, I ended up in 6th place (out of 71) by 3 points, just outside of the money in my league this year. In addition, I didn't win any of my pools’ 18 weekly top 2 slots either, for a relatively disappointing season with $0 winnings despite being in the season’s top 11 the last 10 weeks down the stretch.

After 271 Games…

The heartbreaking thing for me at the end of the season was my week 18 1-point pick of Seattle over the Rams. I ended up winning the point but had the Rams won the game (which was close but didn’t matter by then), I would have only lost 1 point but 2 of the guys ahead of me had picked them with 6 & 13 points, which would have moved me into a tie for 4th place (3.5X the league fee), 13 points away from finishing 1st!

 

Additional Computer Models Created

Since my computer model is based on several variables, last season I tracked 4 other computer model versions that I created based on different combinations of the same variables. From an overall season perspective, they all performed in a relatively tight range between 1658 (11th place) to 1692 (3rd place), with 2 of the 5 models also winning 1 week each in the pool. When I compared my 5 models’ performance (in yellow below) against some of the other ‘experts’ on the internet, I found most of the other models don’t seem to compete for the season but just try to win a few weeks, resulting in terrible overall totals.

​

​

​

​

​

​

​

​

​

​

​

​​

​

​

​

​

​

​

​

 

​​​

Buzzworthy Expanded Offerings

This season, I am pleased to offer 5 different Confidence Pool Models to choose from, adding 4 new models along with original Classic Model: Favorites, Low Risk, Medium Risk, High Risk. You can choose which model you would like to use depending on your goals for your pool, which might change as the season progresses. The nice thing is you can experiment with your risk level while knowing all of the models aren't so aggressive that you'll be out of season contention before the 1st bye week, like others.​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​

​

Variations Season to Season

While an initial perspective of my 5 models would imply the Favorites model is the best model, it needs to be stated that 2024 was a significant outlier when it comes to the favorites winning games in the NFL. On average in the NFL, favorites win 66.2% of the games, while they won 73.2% in 2024. That explains why the Favorites model did better last year, and it also explains why the Classic model didn't place in the top 5 like it solidly did in 2023.

 

Variations Within the Season

When I further analyzed all 5 models over last year, I found that my different models performed better or worse at different times of the season. When I break the season into 4 quadrants (first 4 games, next 5, next 5 and last 4) I found the following tendencies.

 

- For the first 4 games, the Favorites & Low-Risk models led the Classic model by 11 points each, while the Medium-Risk model trailed the Classic by 2 points and the High-Risk model trailed the Classic by 3 points.

 

- For the next 5 games, the Favorites model led the Classic model by 1 point, while the Medium-Risk trailed the Classic by 5 points, the Low-Risk trailed the Classic by 6 points and the High-Risk trailed the Classic by 11 points.

 

- For the next 5 games, the Medium-Risk & High-Risk models led the Classic model by 3 points each, while the Favorites trailed the Classic by 2 points and the Low-Risk trailed the classic by 6 points.

 

- For the last 4 games, the Favorites model led the Classic model by 4 points, while the High-Risk model trailed the Classic by 9 points, the Medium-Risk trailed the Classic by 10 points and the Low-Risk trailed the classic by 15 points.​

​

Underdog Moneyline Pick Disappointment

My analysis heading into 2024 (based on 2023's results) led me to believe that my Classic model was fairly good at predicting underdogs and it would make sense to bet from a moneyline perspective. Unfortunately, after putting my money where my mouth was, I ended up losing 22% of what I bet overall. I'm not nearly as confident the model is great at picking underdogs for the moneyline, but given the abnormally high favorites winning last season, I will continue to monitor it again this season. For reference, I also tracked my other models created last year and the High-Risk model lost 17% of total bet, while the Medium-Risk model won 3% of total bet, and the Low-Risk model won 9% of total bet.​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​