Thursday, August 16 at 4:00 PM ET
With podcasts, previews, further preparation for the oncoming football season and, admittedly, with the fortune of a well-timed schedule and a growing staff on hand to help maintain rosters and analyze all of our information on a day-to-day basis, a little chance to breathe around the MLB All-Star break (thank you Bar Harbor, Maine), the blog has been a little quiet recently.
As most who have been following along for the last couple of years know (or as you can all see by looking at the archive on the right-hand side of this page), that changes around this time on the sports calendar as a voice for our content and information gains significant value from August through January.
While the ResultsFinder was designed to take constant reporting out of my hands and allow the first-of-its-kind, completely transparent performance recording and publishing application to provide any and everyone who comes to the site with the answer to just about any performance-related question that could be asked, I do still intend to check in often with quick thoughts on what we have seen, what we can expect and what questions you may have in this space throughout football season. But, based on overwhelming feedback, this is not the best place to discuss pick performance (good or bad for any sport – poor performance is certainly not the reason MLB picks have not received much attention here; there are a lot of great things going on with MLB that I want to tell the world about and there are things that notably have not been performing up to expectations that I freely acknowledge and have worked diligently to research and improve if/when warranted, but this is and will not be the place for those discussions).
This specific blog entry, though, as we catch-up on all that we have prepared for the football season is probably the opposite of “quick” and probably resembles something closer to a “get-caught-up mega-bag.” (For what it’s worth, I had intended to discuss wagering on O/U team win totals – the current focus of my sports investing - and provide links to previous blogs – the search tool is great for this right now – that cover handicapping and investing in football, but I have tabled that until next week, the last full week before football starts.)
Remember, that you can also always catch me, Jimmy Shapiro and others talking (a lot) about sports and the numbers on our PredictionMachine.com Podcast – also available on iTunes. In the month of August alone, we will record and publish 23 podcasts, ranging in topics from Fantasy Football strategy and who to draft where, to conference-by-conference College Football breakdowns, to Over/Under Team Win Total picks, to division-by-division NFL Analysis and more. During the football season, we will continue to record a weekly podcast hitting on fantasy football, wagering, sports by the numbers and all other topics football related (and some non-football related).
New Football Content
As alluded to above, it remains our goal to make PredictionMachine.com as close as possible to a one-stop destination for sports analysis, particularly with statistical information, fantasy sports and wagering on sports. For the 2012 football season, that will entail more than quadrupling the amount of information and original content that will be available.
Some of the new content that we did not have from last season has already been put in place. We have already mentioned the podcasts and the added depth in our college football and NFL previews. In addition to those items, we will also incorporate weekly commentary from previous contributors including: Dave Tuley’s Vegas Beat, Matt Richner’s NFL Draft thoughts and John Ewing’s 3 Up 3 Down team evaluations.
Other new content includes:
Live ScoreCaster™: We have teased the availability of the Live ScoreCaster in recent emails and for the Super Bowl and look forward to launching the state-of-the-art technology on the site and as an app in the iTunes store at the end of the month. Live ScoreCaster will allow anyone to view current scores and play-by-play of all NFL (and select college football) games as well as live, continuous projections from 50,000 simulations of the remaining time. After each play in a game, the Predictalator simulates the rest of the game 50,000 times to determine the projected score and the likelihood of either team winning. We hope the Live ScoreCaster, which will include a free game per sport, per day (and will cost, at most $1/game otherwise), will be of value to the average fan who wants to know his/her team’s chances of winning a game after every play. We also expect it to be valuable with live, in-running wagering as that market grows in Vegas and we lead the way with innovative technology designed to exploit live market inefficiencies… It’s pretty sweet (to me/us at least).
Team and Player Stats: Last season, we introduced a “Data” tab to the site that included the ResultsFinder, League Schedules and Team Schedules. While that information will still be available (including archived information from previous seasons), we are adding several new items to the Data section. Most notably among the new Data items will be comprehensive Team and Player Stats (and I’m not sure that “comprehensive” is significant enough of a word to apply to the amount of information we will provide in these totally free sections). I often reference “sack rate,” “interception rate,” “points-per-play,” “yards-per-play” and other statistics that I am constantly looking at yet are difficult to find elsewhere. Now there is no need to look elsewhere as we will provide all of that information and more for teams (including league averages – i.e. context) and players and broken down even further for situational splits. Team stats are presented in 36 different categories for NFL and FBS college football, while player stats are segmented into 18 different options. All “normal,” conventional, boxscore stats will be available in addition to the more in-depth information that is of more importance to our projections.
ATS and O/U Team Stats: In addition to what a team does in the boxscore and play-by-play, it can be at least as valuable to know how teams are performing relative to public/Vegas’ opinion. For NFL and FBS college football (and launching 8/28 with 2011 data available), we present team performance against-the-spread and with over/under total lines under ten unique circumstances. There will be some great information here about which teams are under/overvalued each season.
Strength-of-Schedule Rankings: “It does not just matter what a player has done, but against whom he has done it.” I have said that many, many times. Now, not only will you get to see the Team and Player Stats that frame the way that I look at what a player has done, you will also see the general way that I put that into context with our Strength-of-Schedule Rankings. Strength of schedule rankings will be based on the strength of the opponents that the team has played in the season to-date. For this reason, they will not be published for the first time in a football season until at least three games have been played. Factors considered in these rankings include: margin of victory and wins and losses of opponents and the opponents of a team's opponents.
Injuries: Traditionally, for NFL games it has been easy to know which offensive, skill position players are considered healthy for our simulations. However, for non-skill position player and for any player in any other sport, this has not been as obvious. Beginning with the NFL and college football seasons, we will highlight the notable/relevant players with injuries that have been removed from the simulations for all game predictions that appear on our site. If a player does not appear as injured on a team, he was either utilized in the simulation or does not provide a significant impact to his team’s chances either way. In football, with several days before the game predictions that we post at 8 pm ET on Wednesdays, we have the ability to update game outcomes when absolutely necessary (and let you know that we have done so), but, as with our daily sports, we do not make modifications to game information within two hours of the first game of the day. The injury report will keep track of who is out of the games we project.
Along with preparing new content for the football season, I/we have spent considerable time updating our college football and NFL simulation engines (even though this borders on performance reporting, we receive enough questions about this to warrant discussion going into the season). While we had great seasons last year providing consistent value/profit with college football over/under plays and NFL against-the-spread picks, it was essentially the opposite (though still consistent) for college football ATS and NFL O/U. I have researched and implemented everything that I could find/come up with to improve the performance of the areas in which we are weak, while not negatively impacting the areas where we were strong (for what it’s worth, universally with our sports, when we have a strong opinion ATS and O/U in the same game, our performance in the typically weaker of the category is better than when that is not the case, which aids in the argument that improving performance in one area should not hurt the performance in the other – though there is obviously some correlation between total points expectations and the spread).
2010: A few things are important to note about 2011 college football performance. First of all, remember that our most important goal with our predictions is to produce information that, over the long-run, will win at the confidence we project (i.e. 60%+ picks win 60%+ of the time). Also, as bad as 2011 was for college ATS, and as difficult (impossible) as it would be for me to claim that poor performance was entirely due to luck (assuming that 57%+ college ATS picks last year should actually have hit at 57%+, the chance of achieving our eventual record over the sample size we saw is well below 1%), there is some fluky nature to the performance (and not just because four of the five regular season college football picks on this list were “normal” picks that we lost – the other was a “no pick”). The season before, 2010, was a banner year for us in football ATS picks, which helped to prove that we were at least capable of providing value for a full season at set expectations high for 2011 (prior to that, we had just published picks on the previous Super Bowl and NCAA Tournament).
My own professional experience suggests that the weaknesses of 2011 are not expected to return (at the same level – unless, I guess, everyone blew right by the capabilities of our technology at the time in just those areas). As I noted in this blog (from September 2011), “As many previous seasons of data and personal performance supports this technology is not going to continue being this rough in college football, nor will it likely sustain 64%+ picking every NFL game against-the-spread. Over seven seasons predicting the NFL ATS, four previous seasons predicting college football ATS, four NCAA Tournaments, two MLB Playoffs, two NBA playoffs, one college basketball regular season, one NBA regular season and one MLB regular season (all documented professionally), our only events that have not turned a profit have been in the NBA (every season and postseason in all other sports has been successful/profitable). Last year, through two weeks, our NFL accuracy picking every game ATS was below 40%. We ended up at 56.9% ATS for the year (through the playoffs - and not including Week 17). In college, in the first two weeks, we were hitting 65% picking every game ATS and ended the season accurately picking 54.6% of all games correctly ATS. Do we expect to be a little better in the NFL than college? Yes, our confidence and performance suggests this is accurate. It also makes sense given the massive public response to the NFL that is exploitable. Do we expect to do worse going forward than our confidence suggests in the NFL because we have performed so well thus far and things need to "even out?" No. Similarly, do we expect to do better than our confidence suggests in college because of the last three weeks of poor performance? No. Our confidence is our confidence and we fully expect that will be reflected in our performance numbers. Great and terrible weeks will happen and they may even happen in bunches, but, unless there are obvious reasons for them (which I would not contend there are in either sport right now - but we will delve into a couple possible areas for improvement), when it is all said and done, performance should match expectations.”
Finally, let’s take a look at those 2010 football numbers:
ATS Locks of the Week: 28-12-1 (70%)
Daily Top ATS Plays: 73-36-1 (67%)
Paul's Picks ATS: 98-63 (61%)
ATS - All Playable Games: 56%
O/U - All Playable Games: 55%
In 2011, NFL ATS picks were better and college football picks were much, much worse. Locks of the Week and Daily Top ATS Plays were not as strong either (all items that the ResultsFinder clearly indicates). As is usually the case in situations like these, the answer and expectation for 2012 is somewhere in between – and we’ve done all we can to keep the good, as good as possible, while improving the weak.
Home Field Advantage: In 2011, all college football venues were assumed to provide the same home field advantage. Research suggests that is definitely not the case. The average home field advantage is 3.8 points, but actual advantages seem to range from zero (at Navy) to nine points (at Oklahoma). That means A LOT to these predictions and has been fully incorporated into our college football simulation engine (as it was for college basketball as well) for this upcoming season. Here is more information from a previous blog outlining our analysis and results. A great follow-up in an independent study conducted by BeyondtheBets.com provides strong support for our approach.
Strength-of-Schedule: As the sports and media worlds shrink, so does the relative gap between most college football programs. I covered this topic at length in an “SOS” blog on college basketball, but it applies just as much to the now 124 teams in FBS. The same exact concepts that we put into place for college basketball (and seemed to result in some improvement throughout the regular season) has been implemented into college football (with even a large amount of additional research completed to discern optimal SOS modifications at each step in the simulation process).
Here is a valuable excerpt from that blog that provides a good analogy about these adjustments: “I fully believe that adjusting for opponents in college sports in our analysis was excessive (in most areas - double accounting in some). Basketball provides a great example. Jeremy Lin played in a sport with 345 different teams, on a team that would rank somewhere around the middle of that enormous group and against opponents that would rank anywhere from the top ten to the bottom ten. Yet choosing any sample of 25 consecutive games in his career (25 is around the threshold for statistical significance in this example) - including college and NBA - unveils a player who shoots around 55% from two, 33% from three, 73% from the free throw line (though FTs were never impacted by opponent in our numbers), had about 1.5 assists for every turnover, rebounded a high percentage of the game's misses relative to his position and was typically the best on the floor with respect to steals per defensive possession. It's the same guy on any court. We trust that is the case for professional athletes, but that concept is true for most DI college athletes as well. Far more so than ever, DI college athletes are talented and consistent. Furthermore, not nearly as much separates the good from the bad as their used to.”
Statistics Weighting: That blog entry also includes this note: “The most difficult, yet important things to get right when manipulating data are: removing bias from previous opponents (strength-of-schedule adjusting), deciphering what portion of performance in a new season is "real" (as opposed to what career numbers indicate), role changes (mostly having to do with how a coach will use a play) and evaluating the impact of injuries on a player who is going to play.”
We just covered strength-of-schedule adjustment modifications and we discussed during basketball season that we have brought in additional resources to aid us with roster maintenance and pick generation. That leaves the questions regarding what is real.
At one point last year (after Week 5 in college football), I wrote this: “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 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.”
Most of you still reading at this point know that college ATS picks were very good in the first two weeks and weaker in weeks 3-4, then did not improve incrementally (more like dropped off dramatically), but the point of this section is that this season, after a long college football offseason, presents our first opportunity to improve and illustrate that improvement with how we handle this situation.
Unfortunately in college football, when teams incur losses, they can be quickly eliminated from the national championship picture, which immediately puts our most basic/important assumption – that every player on the field is 100% and trying 100% to win each game – in doubt. There is nothing we can do about that. However, in talking with others who succeed in this field and researching strategies as well as our own strengths and weaknesses relative to predicting game outcomes, the most successful approaches seem to be those that consistently stick closer to what was expected of teams and players leading into the season than what has been seen on the field thus far.
Major injuries obviously change that discussion, but we are better set up to handle those situations than those who do not simulate the game with the players. We have built what we believe is a much smarter, automated system to handling player inputs based on previous data than we have ever had. The way we were able to build it allowed for more back testing from previous seasons than we are normally able to do and I am pleased with the way it performs. None of this may mean anything to anyone else, but it’s something I am very much looking forward to seeing in action for football (particularly in college where too much work, number crunching, analysis and roster evaluation goes into preparing inputs for teams going into the season to be discounted once teams start playing).
Other Football Upgrades
Play Analyzer: Many have already noticed the +/- 0.5 point option with every over/under game in MLB. This will also be available to those utilizing the Play Analyzer, still the best way to get the most out of our information, to evaluate NFL and college football plays for both sides (ATS) and totals (O/U). The +/- 0.5 point line indicator allows users to quickly make decisions as lines move or vary in different books, especially across key numbers (for example, easily see what happens to an against-the-spread pick on a -3 point line if it shifts to -2.5 or -3.5).
ResultsFinder: The ResultsFinder is already the ultimate tool for pick performance tracking and analysis. Soon, it will become even more transparent and allow for all visitors to dig even deeper into our numbers than ever before.
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.