Monday, September 26 at 8:00 PM ET
I will always recap each football weekend (usually on Mondays - though there will be no blog on Monday, October 3 and may not be any blog next week due to travel). It is very important to us to be transparent and honest about our picks. At this point, through Monday, Paul's Picks, which include the top three ATS plays Saturday and Sunday as well as the top weekday college against-the-spread (ATS) plays and the Monday Night play, are 16-11 ATS (59%) to start the season. The ATS Top Plays of the Day, the strongest opinion ATS each day overall in football are 9-8 ATS (53%) and the "Locks of the Week" are 3-3 ATS (50%). Including last season, this brings our all- time record during football to 114-75 ATS (60%) for Paul's Picks, 82-44 ATS (65%) in ATS Top Plays of the Day and 31-15 ATS (67%) for Locks of the week.

As a reminder, at midnight ET each day, we make all of our previous day's subscriber content available for free for registered users. We're never going to hide anything. So even though we have to swap out articles in the archive to focus on new ones, articles never go away. Just make sure to use the correct week and date in the URL - or ask us for the link...

We're 2-5 with Paul's Picks (and we're looking for a Dallas win by 4-6 points tonight to clear up that situation). We get it. No one understands nor feels the weight greater than I do. We can't promote the 10-1 start of Paul's Picks and then hide from the fact that we had a 2-5 week with them in the third full week. We had one week like this last year. Given tough losses from the Play Analyzer in college and the overall poor nature of the picks, this year's instance of the 2-5 Paul's Picks weekend may have been the most frustrating weekend since we launched. While we obviously never want, nor believe this is going to happen for any specific week - and it is very important to us to make it up to our subscribers who have placed continued trust in our information, it is definitely possible to have such a week. In fact, as noted in last year's blog on this topic, it is actually likely to occur one to two times a season.

Here is the recap of that explanation, "Within isolated events with determined (by the Predictalator) probabilities, expectations are not guarantees. If it were an absolute guarantee that we would win (or at least never lose) every week, every rational person in the world should put all of his/her money on our picks every week. As it stands, we expect we will have a winning Paul's Picks record every week. We always publish eight Paul's Picks, but at this point in the week, we have only seen the results of seven games, so we'll focus on that for now. Each pick comes with confidence. The average confidence of a Paul's Pick has been about 60.5%. To-date, Paul's Picks are 64-38 ATS on the season, or 63% ATS, which is actually slightly better than expected based on our average confidence. At 60.5% to cover, the 'expectation' is that we get 4.235 of the first seven picks right. The likelihood of getting all seven right is about 3%. The likelihood of going 4-3 or better is about 70%, which means that we should expect to be at 4-3 or better ATS in 70% of the weeks (as of Monday morning). The likelihood of going 2-5 or worse is about 10%. That means that, over the course of the 13 weeks in which we will have a full slate of both NFL and college games, we should expect 1.3 weeks with a 2-5 or worse record. In other words, while we hope it never happens, the fact that it has happened just once all year so far is not out of line with full season "expectations." Our cover percentages and confidence ratings are what they are - and they are proving to be in accordance with our long-term success."

That being said, there are some takeaways from this specific weekend, including early-season projections and expectations, line movements as they relate to injuries and the Predictalator, unexpected/unlikely turnovers and trusting inconsistent/lesser talented teams and flat-out "bad luck." After "The Football Numbers," we will get into those "takeaways" as well as MLB picks and playoff packages and a few new applications on the horizon. Since the overall nature of this week's pick performance was negative, we will not get into Best Wins and Toughest Losses this week. We don't want to give the impression that we think this is was a successful weekend. Similarly, though it can be cathartic to discuss, we do not want the analysis of each loss to sound like an excuse.

The Football Numbers:
Below are our win/loss stats from Week 4 in College Football and NFL Week 3 using playable picks (53%+ to cover) from our published articles. Play Analyzer/Customizable Predictalator plays appear to have provided better results and stronger plays in the NFL, where Philadelphia/New York UNDER, for which the line was as high as 48.5, was the top overall play for the week, covering 67%+ of the time at that line; 65% of the time at Sunday Morning's 46.5-point total line. Atlanta +1, which had been the Lock of the Week at +1.5, was no longer even a "normal+" play on Sunday morning.

• All Playable Games ATS: 35% (19-26)
• All Playable Games O/U: 46% (18-21)
• Paul's Picks ATS: 29% (2-5)
• Lock of the Week: 0% (0-2)
• ATS Top Plays of the Day: 0% (0-4)
• Picks that Cover 60%+: 43% (3-4)
• Normal+ Picks (57%+): 35% (9-17)

And here are the combined numbers for the season thus far:

• All Playable Games ATS: 51% (78-75)
• All Playable Games O/U: 53% (79-71)
• Paul's Picks ATS: 59% (16-11)
• Lock of the Week: 50% (3-3)
• ATS Top Plays of the Day: 53% (9-8)
• Picks that Cover 60%+: 59% (17-12)
• Normal+ Picks (57%+): 48% (52-57)

Early Season Expectations:
Deciphering between what is real, to be expected, and what is anomaly is the most important and the most difficult task in projecting what is going to happen going forward. Early in the season, this is of the utmost importance. We do not want the numbers to over-react, but we also need to give results thus far proper weight. It's the trickiest thing to get "right," especially because there is no perfect answer. In general, everything a player has done in his quantifiable career could impact the numbers, with more recent data having greater weight than out-dated information. I have been working in the sports prediction and simulation business doing almost exactly what I am doing right now for eight years. Few things give me more headaches than weeks 3-4 or so of any sport. Look at our track record. We have consistently performed well in the first 1-2 weeks of the college football, NFL and MLB seasons. Then, it can be a roller coaster for a couple weeks (sidebar, I'm not sure how many people were paying attention back then, but our overall NFL performance through three weeks is actually significantly better this year than last) before performance stabilizes and generally improves incrementally. With our strong performances in early weeks, it may make sense to minimize the impact of early-season on-field performance, especially relative to competition, but there are always going to be a few teams that come out playing completely different than expected, to the point where significant adjustments above and beyond the norm are clearly needed. It's a delicate balance and one that I believe we do better than anyone else, yet it's also our opportunity for greatest improvement.

This is the "good news/bad news" situation that surfaces just about every season around this time, I'm focusing on making tweaks to our automated data manipulation that will have far more to do with what happens in the early weeks of the following season. It's bad news because we never want to have to make tweaks and that these tweaks do not really "fix" anything going on right now. It's good (even great) news to consider that we should not have to worry about this going forward because, through three-plus weeks of football, data from this season can be trusted more and more each week.

I, likely more so than anyone, took this weekend's frustrating performance very hard. When Atlanta jumped offsides on Tampa Bay's fourth-and-1 and Kevin Kolb almost simultaneously threw an interception that essentially sealed the Cardinals' fate, I shut the TVs off and just sat on the couch for a few minutes considering what I had just witnessed from the week. After some time, I got back on the computer, uploaded college team data, reviewed notable injuries from Saturday and ran the college football power rankings. It was then that I became oddly content. I saw the phenomenon that I just described happen in front of me. The results looked great. I realized that so many things went against us all week, that weeks like that can and do happen about as often as we would expect them to and that, going forward, we will put our subscribers in the best position to succeed. I'm not happy about last weekend; I am just looking forward to the weeks to come.

Inconsistent/Untalented Teams
12 fumbles lost, seven interceptions thrown, -10 turnover margin... Those are the combined stats of the five Paul's Picks teams that failed to cover. That is a tough way to lose (especially when one of those fumbles occurred when ECU thought it had scored a TD to go up 34-17, at which time I, happily, left the room only to return to see a 28-23 score). We know that is an unlucky way to lose. We incorporate likelihood of turnovers into the numbers; however, these five teams were a combined -1 in turnovers in their 12 previous games this season. Special teams or defensive scoring plays, a high percentage of loose, "50/50" balls going to one team, bad officiating calls that are not corrected, in-game injuries to previously healthy players and weather beyond our expectations (though we have some great weather resources that have benefited us greatly when looking from Wednesdays several days into the future) can all lead to unexpected outcomes. When they happen in ways that change the outcomes of the side and total, there is little that can be done other than to assume it will balance out over the course of the season.

In the NFL, where players, teams, officials and injury reports are more consistent (plus, there are more domes, so less weather issues), these unexpected occurrences are relatively rare. However, in college, with wide variances in levels of talent, inconsistent players, teams, officials and injury reports are relatively commonplace - and they all played a huge role in this weekend's biggest losses. I believe that something can be done about this. We did research in the off- season that indicated that our Locks of the Week that involved two BCS-AQ teams went 6-0 ATS last year. If we had determined our Locks of the Week based on our top play between two BCS-AQ schools last year, we would have been 12-4 ATS (75%) with those picks, which is even better than we actually finished. Based on that research and a similar approach that we took to college basketball, we implemented a tweak to account for that lack of consistency, talent and confidence in weaker teams (especially if playing other weak teams) in our projected confidence in the play. Had our top plays Saturday simply come from that subset of games, we would have gone 2-1 with our top two plays - Oregon (-15) and UCLA (+4) hitting. In the meantime, all three Paul's Pick losses - ECU, Ohio U. and UCF - came from BCS non-Automatic Qualifying schools. But these picks lost for the exact reasons why it is more difficult to trust those teams. Needless to say, I have reevaluated that tweak to the Predictalator and will be implementing it going forward.

Line Movements/Play Analyzer
We are very proud of the Play Analyzer as we feel it is the embodiment of efforts to provide our consumers with the best possible information at all times. We hope that is a trustworthy and beneficial tool for all. Thus far, it seems to have excelled in the NFL, identifying plays like the Giants-Eagles Under this week, which was a 67%+ play when the line was at 48.5 on Friday, and both the Saints over the Bears and Jets over the Jaguars in Week 2 of the NFL, when those lines moved from -6.5 to -5 and from -10 to -9 respectively to become two of the top three plays of the day last Sunday. For college picks, the Play Analyzer has not been as beneficial, especially in the last two weeks. Bad luck plays a role, but there are a couple things that we are going to do to better handle college line:

Injury information is far less consistent and far more difficult to track and find publicly in college football than in the NFL. No matter how deeply I/we researched Nevada @ Texas Tech, it was a little-known injury that ultimately kept the Red Raiders' middle linebacker and top defensive player out of the game. We have vastly increased the man-power, and information resources devoted to tracking and accounting for college injuries. Not only will we be more proactive as it comes to injuries, we intend to be more active responding to our findings in keeping the picks and Play Analyzer and Customizable Predictalator data as up-to-date as possible. Pick Availability update alerts will still be sent for major injuries, while there may also be some minor injuries or other changes that affect the Play Analyzer and Customizable Predictalator numbers slightly, which will be updated and noted on those pages.

We have long maintained that, in order to appropriately credit the knowledge of the market, the line plays a role in our actual projections. While this is true, its role has previously been minor and only truly applied to the published picks. Our fear has generally been that adjusting data simply due to line movements is not consistent with the context of our simulation approach. Whether that is true or not probably pales in comparison to our overall goal of providing the most profitable information, which may be partially compromised (especially in college football as noted with injuries and the consistency concerns in the above section) by large line shifts. Line movement will have more of an impact - and in our mind the appropriate impact - on Play Analyzer and Customizable Predictalator picks going forward. Picks will still gain in confidence as the line moves in our favor; however, the further the line moves, the less incremental confidence will be added.

MLB Performance and the Playoffs:
Last week's blog was mostly positive with football and not positive with baseball. This week, it is the opposite. Since that was published, playable MLB picks are 110-86, + $514 for a normal $50 player, run-line picks are 37-26 +$92 O/U picks are 27-26 -$20. That's a solid finish to an otherwise tough month and a great way to enter the playoffs. For the season, money-line and o/u normal or better plays are 89-60 (60%) +$1,005 for a normal $50 player. And since May, even including September, all playable MLB plays are +$2330 for a normal +$50 player.

Speaking of playoffs, MLB Playoff Picks packages are now available. As we noted last week in reference to a tough September with motivation, roster and lineup concerns, fortunately, we don't have to worry about this in the Playoffs! In the MLB Playoffs, every team has a 25-man roster, sets its rotation well in advance and is motivated to win every game. As we have seen in other postseasons, I expect our performance to replicate the strong levels of June, July and August in October. The Playoff Odds article will be available to all on Thursday morning, while daily picks articles will be published as far in advance as possible.

New Applications/Product Enhancements:
And lastly, we have been working very hard on utilizing (your) feedback to make the site as user-friendly as possible. Watch for the following enhancements (and more) in the near future:

• TrendFinder Database
• Exporting Play Analyzer Data to CSV file (for use in Excel, etc.)
• Advanced Play Value Calculator with Ability to Calculate Multiple Play Recommendations
• A Tutorial Video on the "Data" Pages
• FAQ Page

As usual, if you have any of your own comments about this article or suggestions about how to improve the site, please do not hesitate to contact us at any time. We respond to every support contact as quickly as we can (usually within a few hours) and are very amenable to suggestions. I firmly believe that open communication with our customers and user feedback is the best way for us to grow and provide the types of products that will maximize the experience for all. Thank you in advance for your suggestions, comments and questions.