Basketball Trial (Updated 2/18/11)
Update: Friday, February 18 at 9:12 PM ET
While I understand how easy it is to talk about our picks after two great days (and on a day without enough games to have strong opinions), with upcoming business and personal obligations - some foreseen, some not, this is the only time that I am confident that I can work in a recap of the second week of the basketball trial before the subscription period kicks in Monday (for college) and Tuesday (for NBA). To clarify, I have no concerns about my ability to publish and update picks, nor my/our ability to answer customer support contacts. It's just that, especially given the way that I blog (one user calls it "War and Peace"), I know that I cannot commit to that much content over the next few days. A quick (by my standards) blog update on our numbers since the last blog should suffice. I just added a lot of other content (NBA All-Star break Power Rankings and Projected Standings, Academy Award Projections, etc.) to feed any hunger for long-winded, almost-too-in-depth analysis for the next few days.
The best news is that our "normal" or better picks are 8-2 ATS in the last two days. That includes 3-0 ATS in the NBA on Wednesday and 2-0 ATS in college last night. Plays that were 60%+ to cover since we finalized the tweaks on Sunday are 3-0 against the lines (and averaged covering by 8.8 points). There will likely be far more opinions, including some 60%+ plays tomorrow in college basketball. Hopefully, we can keep up that strong performance.
It has not all been perfect - "weak" picks are right at .500, yet, with so much value elsewhere, there should be no need to play those much, and Tuesday's top plays, Indiana and Ohio State were both at least "pushing" with 15 seconds remaining the games, yet lost - but the few losses that we did take to better picks earlier in the week had at least as much to do with luck as any potential issue in the numbers. Just as we had hoped, we are very confident in the engine and the numbers as we complete the free trial period. We were so close. We thought we were even closer, but we learned, adapted and are happy with what we have seen - especially with our high confidence picks. The tweaks we made literally impact input values by just a few hundredths of a percent, but it all compounded to hurt performance. Homecourt advantage has shown great balance in both sports. (Some of you noted that we finally lost a road underdog pick and cited the tweak. Losses are going to happen and, in this case, had absolutely nothing to do with the tweak. It was most important to get homecourt right and we feel we have done that.)
So top picks have been doing very well; weak picks have been ok (not crushing, but not too strong either) and we feel we are good to go. Hopefully, that is backed up by our performance in the final two days of the trial period and continued into the subscription period. We wish the entire trial period would have been as successful as the last two days have been. We really do. That would put everyone in a great position. But I still think that we have done at least an adequate job of learning, interacting, explaining, adjusting and fixing in real-time. And for that, I think we will be better off for the long-run (which is how we and, hopefully, you view this and/or any investment strategy).
A few more customer support contact exchanges:
Question: I am assuming the regular season package does not include playoffs (college or pro). How much will the playoff's package cost and will there be a discount if you also purchased the regular season?
Answer: You are correct that the regular season packages do not include the playoffs. Our expected pricing for the NCAA Tournament this year will be $174.95, while the NBA Playoff Package will be $199.95. As we have done with our other sports, regular season purchasers will get some form of a discount on the postseason, but we have not yet finalized this.
Question: Do you know how much a better needs to wager per game to have a positive expectation? What I mean is, if you average 4 plays per night at 2-5% of your bankroll what does that need to be dollar wise to cover the cost of the package, vig (not sure if that applies here), etc assuming you win at something like 55-60%. I hope I am making sense.
Answer: While I understand the question, there is not a simple answer. There are many, many factors and assumptions that go into this and I also cannot guarantee anything with respect to net profitability. That being said, with the value that basketball provides with plays on a daily basis and multiple weeks remaining in the regular seasons of each sport, even at 2% play and 55% accuracy (timing matters, but we have to assume that it's 55% every night), the breakeven point for winnings minus losing minus cost of product would require a starting bankroll of around $750 for college. Due to the frequency of "normal" or better plays, NBA is not quite as good of a deal, but should still be profitable when factoring the cost with a bankroll starting around $850 under those same assumptions. Again, there are no guarantees and everything that you do with your bankroll is ultimately up to you, but this is the best way I can evaluate your questions. Now I am the one hoping I make some sense.
Question: I suggest you post recommendation on how much percentage to play on Normal play against the Weak Play (yellow marked) it helps us greatly to spread the bank roll more wisely. You already posted the recommendations for the Green and Normal play but did not stress out the amount for Weak Play 53-57. Will you do this?
Answer: Weak plays warrant playing less than a normal play. This could be anything from 1% to 99% of the normal play value. Please click on the "Calc" button next to the pick to determine our exact play value recommendation. As a suggestion, I would not be playing Weak plays unless you strongly agree with our pick. We should provide enough normal or better plays over time to be consistently profitable. Weak plays should also be profitable, but are not nearly as valuable as "normal" or better plays and can even take away from the value of those plays when played at the same time.
Question: Most of your NBA games were won in the Yellow 53%-57% where I was hesitant to go into the game. Was that a bad decision on my part?
Answer: I am very happy with the turnaround of the performance of our information over the last week. In general, I can assure you that we are continuously reviewing our numbers, looking for and addressing areas of weakness (and strength). While I will never be able to say that we are perfect and some perceived issues may actually be legitimate, most are products of short-term results that do not have merit in the long-term. We are very good at projecting the lines in the NBA. This ultimately means that we end up with a high quantity of "no picks" and "weak" picks. Weak picks are what they are. Our record in normal or better picks is and should be expected to be better than with weak picks. When we do have a strong opinion in the NBA, that's great news. We have found one of the few opportunities in the NBA to get significant value. With daily picks in the NBA, we have the opportunity to find such value and play just a couple games a day.
Question: Just out of curiosity, what were the site's %'s for the LA Lakers vs Cleveland game on Wednesday night? Not that I'm basing your results on one upset...just curious.
Answer: Fun question... We had the Lakers winning straight-up 78.5% of the time and by an average score of 105-94. The spread was LAL -11 so we didn't have an opinion against the number. I often like to think of upsets like baseball players' batting averages. Andy LaRoche hit around .210 last year, but is it a total shock when he gets a hit? Cleveland hosting the Lakers was like Andy LaRoche against a league-average pitcher. There were two other games on Wednesday where we had stronger straight-up favorites: Orlando over Washington 83% of the time and by a score of 107-93 and Boston over New Jersey 82% of the time and by a score of 99-86.
Question: (From Monday, February 14) It's a shame that u guys tweaked the nba systems b/c the Road dog system was doing great.I started playing them last night and they lost badly. I cant believe that u guys did that. I even read in Paul's blog that he was gonna tweak some things even if it meant sacrificing the road dog plays. Why in the world would u mess with something that has been working. I love your concept here and hope that things will go back to the way they were before all this tweaking.
Answer: should probably clarify what I meant by that note about homecourt advantage. We aren't trying to get picks wrong. We're trying to improve our accuracy in every game - to fairly account for our confidence in all picks. Looking at the data, we were clearly overvaluing homecourt in the NBA and we had to address. Under either system, the only road underdog we would have been picking last night was the Clippers. In this case, the Clippers actually came off stronger last night than they would have otherwise. The fact that the Clippers failed to cover is more fluke than anything. At some point, we were going to lose one of those picks. We would have been picking them - and no other road underdogs - no matter what.
Question: (From Thursday, February 17) I use the Play Calculator religiously, especially since my book has different lines that what your posted picks use at least 25% of the time. It is a great tool. Some guidance, if you would, about interpreting certain results. For each of the games I input into the Play Calculator, I'll typically replay the game (with the exact same numbers) a handful of times and use an average of the resulting, different %'s. For the most part, the different %'s range +/- 1% or so. Yesterday, on the Wyoming/Utah college game, my book had the over/under at 138.5 while your posted number was 138 with a confidence % of 58.4. My book's line was a little better so I obviously expected a bit of an increase in confidence but for that game, when run multiple times, the %'s ranged from as high as 61.1% down to 57.4% for the over/under. What, if anything should I take away from that variance? And does the Play Calculator re-run the game 50,000 times each time I click the Play button or does that re-run the game just once (I assumed the former but the latter would explain why such a variance)? Also, since your tweaks on Monday, my plays (60%+) have done well. I haven't tracked the other playable plays but are you seeing improved results from last week?
Answer: As you have figured, it is running the game 50,000 in real-time. I can tell you that I am very surprised to a 3.5% difference between extremes. That spread not only represents about the greatest extremes one could expect, but it seems like an anomaly in its own right. I would not expect to see values differing that much going forward. We will look into whether or not there is any specific reason for this outside of randomness (or if there is any unwanted, "artificial" randomness). Obviously, it's not changing the pick, but we want to be very clear with our recommendations. And as for our performance in the last three days, we are much happier with it. All playable picks in each category (NBA and College - ATS and O/U) are profitable. Last night, we went 17-9 on playable sides, 6-2 ATS on "normal" plays or better. It's a small sample size, yet great to see things turning around after our analysis/tweaks.
Monday, February 14 at 5:12 PM ET
This week's blog entry will recap the first week of basketball and tie-up the final numbers from football. With a free-trial, we received a great deal of questions and comments and will rely heavily on those to populate this blog. I apologize if some of this gets long-winded or redundant, but I think it is all important to cover... So I just spent the week getting stuffed on humble pie... Does anyone remember 11-0?... Save for a celebratory two-hour block on Twitter between midnight and 2:00 AM ET on Super Bowl Sunday, I feel like I kept my cool and kept our fortunate/lucky playoff performance in great context. And now we've moved on to something very different and results that have been unfortunately different, but I think/hope it's a good thing for the long-run. It's not a wholly bad thing to be brought back down to Earth and grounded - especially when everyone is watching now. While the first five days of the basketball trial were less than ideal, it has presented us with a much better opportunity to grow, improve and focus.
One of the greatest business lessons that I thought I had learned occurred at Cinergy Field when I took in a Reds' game the summer after my freshman year of college. Cincinnati is known for its chili. Skyline is the best and the most popular. There is a Skyline on campus that I had been to many times (I'm undefeated in coney eating contests and am accepting all challenges). The second biggest chili brand in the area is Goldstar Chili. Goldstar had secured the ability to sell its cheese coneys at Reds' games. Having never been to a Goldstar and curious about the differences between it and Skyline, I decided to get a coney. Looking at it, it wasn't much. The cheese was not melting, the amount of chili was minimal, yet I still went for it to try it. The coney hot dog was essentially frozen. It was the worst eating experience of my life. Especially as a business major, I was blown away. A company so eager to gain attention and establish itself in the industry had entered the largest venue for reaching the masses - and put its worst foot forward to the public. I've never eaten Goldstar since. I don't care if it means taking a loss on that one coney, Goldstar needed to provide me with its best possible product if it wanted me to ever think that it would be better than Skyline. Now, that coney cost twice as much as a normal coney at either restaurant, so the greed associated with ballpark concessions was an obvious factor, but, to an extent, I feel like we just unintentionally pulled a Goldstar. We have as much attention as ever and had an opportunity to introduce ourselves to the masses with a free product, hoping to sell them on our technology and information and timing dictated that it just wasn't our best product. We may lose some we could have kept around otherwise. Instead of the apathy and short-sightedness that afflicted Goldstar at the time though, we are the opposite. We care and we see how we have to learn and grow from this.
I want to get every pick right. I do. I genuinely like helping people make smart decisions to better themselves, so in no way was I anticipating starting out with a sub-.500 first week of the NBA trial. I thought we would win. Testing said we were winning. Our own bankroll had us winning in basketball. Ultimately, we probably had just as much good luck during testing as we did bad luck this Wednesday - Friday, but we were able to delve into and uncover a few legitimate concerns and we should emerge better for it. I just want to be clear that, despite the fact that we are chalking this up to a learning experience that will help, I personally feel bad and apologize for any and all losses. I'll never forget Mississippi Valley State -2 @ South Alabama (our first ATS loss in five weeks) and FAU -2 @ Denver (a huge pick for us that was a blowout the other way). Many have asked to see our record. While I could quote some of our wins and losses, there are two reasons why I'm not going to publish exactly what our record was last week in college basketball. One of the issues we found with our product is that our internal record-keeping and database of results for basketball was not working correctly. This may have thrown off results during testing, but had an equal chance of affecting numbers in good or bad ways. It definitely could have hidden some of the more obvious trends that many of you found quickly. This has been resolved going forward and I will publish another blog next Saturday that discusses this week's performance. The other reason is that, over the weekend, we worked very hard to incorporate our learnings and your feedback to optimize the engines. Multiple tweaks (explained below) have been put into place to aid performance (more for college than NBA). You can already see the results of those tweaks in the numbers tonight (despite very few games) and some of you even noticed that we accidentally overwrote the NBA picks for Sunday and college basketball picks for Thursday - Sunday. This is how the picks would have looked using the "new" Predictalator. We do not anticipate having to tweak the engines again and are very happy with retroactive testing (now that our internal tracking system application is working correctly), but we do not think it is fair to any party to focus on results from last week because of the notable differences in the current engines.
There are many reasons why we opened up basketball for a free trial. From a business standpoint, we generated so much interest from our NFL Playoff run that we wanted to make sure that everyone had something to come back to the site and do/see after the Super Bowl. We wanted to show off the Customizable Predictalator and Play Value Calculator and even our ability to quickly respond to questions and support our users. We also wanted to start generating interest in our other, new daily sports besides just the March Madness, NBA Playoffs and football that we published in the first year of the site. We were hoping to help people with their basketball picks. We were also hoping to be helped. That's where we eventually saw the greatest bonus from the trial.
This is the first time in my professional career that I/we have published daily picks for any sport, let alone basketball. There are many challenges that come along with that, some we were fully prepared for and others not. While testing for over a month was pretty strong, we knew that we may need some more time and fresh insight to make sure that our basketball was as strong and successful as possible. The free trial bought us a couple of weeks before the subscription period and thousands of extra eyeballs and minds. I really felt as though we were going to win. We took a risk by opening everything up. I thought we would win. I knew that if we didn't win, it could hurt our reputation and much of the goodwill we gained from the NFL Playoffs. And I am very sorry for those who did not win using our information in the first week of the trial (many probably did win - depending on when and what you did, yet our overall record would not suggest profit for the average user). But I also know that we are in this for the long-run, so if we can build a winning product, people will figure that out and come back to us. And I knew that if we were off in some way, our users would figure it out and we could adjust. That's exactly what has happened. We were close last Monday. This Monday, I really think we are there.
I am a firm believer that it is absolutely critical to small businesses to listen to their customer base. Feedback is how we grow; it's how we improve the user experience; it's the best way for us to help you, which then helps us. We have and will always take customer support and questions seriously. The record number of questions that we received last week gave us great insight into the thoughts and actions of our customer-base as well as many hints as to elements of college basketball and the NBA that we may not have exactly correct. Now, I think we have those items much more "correct." So, thank you for all of your help, insight and understanding.
Recent Findings and Tweaks (all in place for games 2/14/11)
NBA Homecourt: 11-0!! That was not just our NFL Playoff record; it was our record last week in games in which we were picking a road underdog. Our record in games where we were picking home favorites was a few games below .500. It actually was a weird week in the NBA where road teams performed much better than usual (all eight of them won straight-up on Saturday and Washington even won on the road - as predicted). Still, the optimist in me who believes in our approached looked at that trend (submitted by a user - thank you!) and determined the obvious. If we are really good when we like road underdogs and not good when we like home favorites, our homecourt advantage is just a little too strong. The road underdog percentages should have been a little higher, while the home favorites should have been a little lower. With accurate tracking and an optimality testing going back through our results, we believe we have more appropriately assigned value to homecourt in the NBA. This may mean that we actually get a game wrong where we like the road dog, but it's not likely going to be a game in which we have strong confidence.
College Non-DI Games: There was a flaw related to data from previous college games between DI and non-DI opponents. We had attempted to correct this during testing, yet over-corrected the flaw and created new issues. This was our first tweak and was resolved on Friday.
College Homecourt: Homecourt advantage in college basketball is the most difficult thing that I have ever had to analytically define. It's absolutely critical to team's performance, yet can vary widely. With our great success last March, we had not had to worry about homecourt. I knew this would be the biggest concern. Needless to say, with some more great research and insight from our users (thanks again), we were able to shed some light on the issue and realize that we were slightly undervaluing homecourt advantage in most cases for college games. Again, we believe we have more appropriately assigned this value through optimality testing and are very happy with the results.
College Talent/Consistency: Most people think that we should have great insight and find value in games between two teams that no one has heard of, but it's actually the opposite. I tweeted earlier this week that, "The more the public has an opinion on a game, the more likely it is to be wrong." That is equal parts overly simplistic and too deep. It's not really that the public is more likely to be wrong, but the line. And more importantly, it's more likely that we can accurately predict against a line that is strongly influenced by the public, but not just because that is so. It is true in general that I would much rather be picking against a line that is influenced by the average bettor as opposed to just the books and sharps/pros - books and sharps are tapped into similar analysis and information as us, yet this affect is amplified by considering the times in which the public influences the line more than the opening linesmakers and sharps. The general public tends to play what it thinks it knows. It likes the pros more than college, and, when it's college, it likes notable, usually nationally televised teams and games. It likes teams with great talent and where injury/suspension, depth chart and coaching information is most readily available. It also likes consistent teams.
Personally, assuming our confidence is similar, I'm more comfortable playing Illinois +1.5 @ Minnesota than I am playing FAU -2 @ Denver. All of you probably are as well. Yet when a "green" pick comes up, we want to jump on it no matter who it is, which is what we tell you to do. We discussed this in college football and a review of our college football results confirms that our performance is typically better with BCS conference games than non-AQ games. We were also better and more consistent in the NFL over the course of the season than with college football. And we have empirical evidence from our basketball trial and testing that illustrates that, the better the competition, the better our results. Good, consistent teams for which roster information is well-known (and for which there is motivation to succeed, though we'll table that for now) will provide our greatest opportunity for success. In doing so, they should also provide our greatest confidence. Up to this point, we had not worried much about what goes into the line or how consistent teams are. That's not realistic though. So, we have determined what I think are very logical and technically grounded ways to incorporate these consistency, talent and public factors into our straight-up, against-the-spread and over/under confidence percentages. At this time, our confidence truly is our confidence in a pick as modified by these factors while simulating games 50,000 times. You will note that games with similar scores against similar lines but involving teams from very different conferences/talent levels may be separated by a few percent in our confidence levels. This is by design in a way that we believe will benefit all.
Customer Support Q&A (apologize for redundancy here)
Question: What the heck happened with the FAU/Denver game? Any principled theory why the Predictalator got it so wrong?
Answer: Wow, good question. FAU was 10-1 in conference, playing very well recently and was even power rated as a favorite in this game by most other sources. I'm not sure what happened in this game or what we didn't know. Looking at the game, FAU was horrific in every category. They couldn't even shoot 50% in free throws. We're not going to predict that happening. That being said, I do have a theory...
We are typically better with better college teams and professional leagues than lesser college teams. Many think it should be the opposite. However, the lines on the non-marquee games are typically far more grounded in numbers to begin with because the public does not force the line much. So while we have a strong model, it's going head to head with similar models instead of the public. Furthermore, the less talented the player/team, the less consistent it is. That can create problems with our numbers because we are trying to predict the likelihoods of the most likely scenarios. Wide variances in expectations from game- to-game with weaker teams are going to throw those numbers off. And lastly, even with the ability to account for injuries, injury information is far less prevalent for lesser-known teams. We should still be able to come up with accurate confidence in all picks and I think the model is as strong as any that can exist.
Our best performance will always be at the end of the season, with as much data as we'll ever get, preferably on neutral courts/fields, when every team is pretty good, the public is steaming the line and when we assume every team is trying its hardest to win. In other words the NFL playoffs, March Madness and, in the scope of a series, the NBA and MLB playoffs.
Numbers in testing were much stronger than we have seen so far in basketball. It's been much more of unlucky week than I expected.
Question: Is there something I'm missing with regards to why I would purchase an NCAA or NBA regular season package? Were you that much better in football than basketball? Please enlighten me how this is an effective handicapping tool for regular season basketball or what I'm supposed to learn from this trial period.
Answer: Unfortunately, the picks projections have not performed up to our expectations over the last few days. While we are only four days into our first ever endeavor into daily, regular season basketball, as you note, there have been enough games to draw some conclusions.
In general, I can tell you that we are usually at least as good at predicting basketball as football. Our record in last year's March Madness tournament was over 62% against-the-spread and 75% straight-up. In season, I think there are some things that we are still working out. Homecourt advantage is the most critical piece to college basketball games that does not come in to play for the postseason. Teams playing games against non-DI teams and earning credit for those in our data is another challenge. Plus, with so many games and such varying levels of talent, motivation and consistency come into play (again, far more so than they do in the tournament).
We have put some tweaks in to attempt to address most of these issues and will do so throughout the free trial period. During our internal testing of these products, we achieved great results (over 60% ATS across the board and our ATS record equaled that of our SU record in college). Those numbers were probably a little inflated, while the performance of the last couple of days is clearly deflated with respect to our expectations. There were also some tweaks that we put in during testing that, in hindsight, were pretty clearly over-corrections. We try hard not to over-react or over-correct and believe that the product we will put forth by the paid/subscription period will be a very profitable one.
Also, during this trial period, we encourage everyone to utilize the Customizable Predictalator and Play Value Calculator features. Part of what we intend to do is show the capabilities of this technology and what is available to subscribers. We also want to expose our information to a much larger audience to generate as much feedback as we can. As we are learning, thousands of heads are better than a few. The feedback has been great, almost overwhelming, but in a good way that presents an opportunity for us to grow and improve.
Question: Quick question regarding the NBA and NCAAB predictalator. How do you take into account things like: Being on the back-end of a back-to- back; Most recent performance (teams get hot and cold, make coaching changes, etc.); Player injuries
Answer: We do factor back-to-back games to an extent, but it does not appear to be a significant as the books' believe. This is a reason that the Customizable Predictalator should not be used until the picks are published.
Most recent games are always weighted slightly more than other games, but we do take into account a player's entire career.
With simulation, we get to explicitly account for injuries by removing an injured player and plugging in his back-up in that role. The Last Updated time is always to the minute, so we use injury information that is updated through that point.
Question: Just thought I'd offer a possible addition to the site you may want to consider. I think a 'Who Should I Wager on Money-line (straight up)' could be a nice little extra feature on the site. It's rather quite simple to calculate yourself and most people probably do it themselves but there's no harm in including it.
Answer: This is definitely in the works. The point you bring up is an important distinction we are working harder to make. There can be value in money-lines and we are not yet giving you all the answers. I can tell you that we are building an application for baseball (where money-lines are about all that is played) that will better evaluate money-lines.
For now, please note that that Play Value Calculator and Key assume -110 odds, which is common ATS when the book takes 10% on losses. With any calculation though, you truly need to know if our confidence in a pick justifies the odds.
Here are the equations to determine that:
When looking at a pick that is favored (starts with a minus - sign), divide the absolute value of that by 100 plus itself. For instance, if the pick is -110, the calculation becomes 110/210, which is 52.38% or the value that it is needed to justify a pick ATS. If the odds are -200, the value is 200/300 or 66.7%. If we are picking a team to win 65% of the time straight-up, but the odds are -200, we can't justify that pick. If we are picking the same team to win 70% of the time, we can justify the pick because our confidence is greater than what is needed.
When looking at the underdog (starts with a plus + sign), divide 100 by 100 plus the odds to get the percentage needed. For instance, +150 becomes 100/250 or 40%. If we have the underdog winning straight-up greater than 40% of the time in that case, there is value there. It's not necessarily going to win more often than not, but it is worth the payout.
That's how you get to yes or no on playing the money-line. You will see that there is not always a play because the book adds juice that ultimately results in percentages that add to greater than 100%. Obviously, the greater the difference between our number and the percentage needed, the greater the value and recommended play. However, we will need to automate that process for our users to make it easier to handle. Watch for that soon.
Question: My question to you is should I be concerned with my bad returns so far on 53-57% and look at cutting them off or continue to place them?
Answer: Our confidence in our pick is always exactly how likely we think that pick is to cover and everything over about 53% (at -110) should be profitable. Ultimately, every decision you make is your call. What you are doing makes a lot of sense and is definitely not "wrong." In fact, it should be pretty profitable in the long-run.
That being said, basketball presents an opportunity, with its frequency of games, to target value. Many people have been asking about my preferences for strategy with basketball. We have even discussed at length internally. With the opportunity to have multiple daily picks above 57%, we really want to stress those - especially at 60%+ - while downplaying 53%-57%. Our biggest point of emphasis with marketing is and will likely always be our "normal" or better plays.
Also, our optimal strategies related to the Play Value Calculator suggest that no more than 16 plays should be put into place at once. Action is great when winning, but there can be such a thing as over-diversification. A happy medium of around 4-7 plays (if/when there are that many) is a smart way to go.
Question: I probably could use some reading on over-diversification. What I am struggling with is the math behind the Play Value Calculator. For example, (just using 60% for a round #),to me a 60% play is a strong as a 60% play on a day where I have only 1 play or 20 plays. Could you explain the logic behind how a 60% is basically a much stronger play on a day where I have less plays than a day where I have more plays?
Answer: It's not really that the pick itself becomes less valuable when more games are played, it is that you are pulling from a different pool for your bankroll when you do. If all 50 sides today were 60% and we recommended playing 5% of your bankroll on each, you would have 250% of your bankroll in play.
To minimize your overall exposure at one time - and to verify that future decisions are made off of the correct/updated bankroll, the recommended play values decrease when more games are played. The way that the factor scales, after 16 or so plays, there is not much difference in recommended plays regardless of the confidence, which means that the confidence will be lost.
Technically, 60% is 60% and you should expect to hit 60%. If every game is exactly 60%+, there is an extremely good chance that you will turn a profit when playing all of those games and you are not passing up any value from one game to the next because they are all the same.
But it doesn't work like that. Sure, by playing more than 16 playable games, you should still be in the black far more often than not, but you are hurting your margins by greatly diminishing potential reward. On nights like tonight with so many games with high confidence in play, it's ok to play more than on other nights. That just does not mean to play them all. Always be cognizant of the overall percentage of your bankroll in play - your exposure at that time. Also, continue to focus on high confidence picks, where the reward is the greatest.
Question: What time do you release your basketball bets?
Answer: We will release and finalize games no later than one hour before the first game in each sport. Usually, on weekdays, we will post picks around noon and then update (if necessary) in the early evening. To know exactly when the picks are posted, please sign up for the pick availability alerts by clicking on Account in the upper-right-hand corner of any page. Based on feedback, we will send out pick availability alert emails whenever picks are updated, including noting any added games or significant changes. This will likely only happen in the case of major injuries/suspensions and on Saturdays in college basketball when lines for later games are not published early enough for us to include in the original picks articles.
Question: I have a question about the push %'s you utilize when making your predictions with the game calculator. Using today's first play as an example, you give 537 fla atl +2 a 63% ATS prob. If i increase that number to +3, it becomes 65.2, to +4 67.2, to +5 69.7. Those %'s look quite low based on historical with the 2 pushing ~4%, the 3 ~5%,etc.
Answer: I am glad you are taking advantage of the Customizable Predictalator and asking questions. In this case, there's not a ton for me to respond with because this is just the way the simulations shake out. I know you are citing the historical push percentages at two-point, but that makes the invalid assumption that this two-point line is appropriate. The line may be 2, but our projected margin is closer to three points the other way, making a push actually 5 points away from our projection. For the most part, we are saying that FAU is most likely to win by about three points and has around a 3% chance of losing by two points. That seems fair to me.
Question: Would love to know your thoughts on betting the 2nd Half in a game. Many times, I see a team getting blown out in the first half. Given no in game injuries, is there value in chasing the 2nd half? If PM has an opinion on a game, and the team gets destroyed in the 1st half... should I hit it?
Answer: In the not-too-distant future, we will add in-running projections that will definitely help explicitly with halftime lines. In the interim, I think your contention is fair. If one team plays an exceptionally good or bad half relative to our expectations, it wouldn't really change our expectation for the second half (unless there is a major injury). To a good degree, you can assume that our projection for each half is about the same.
Question: Are you planning to do any NHL games in the near future?
Answer: We intend to provide daily NHL pick information for the duration of the 2011-12 season. The NHL "Predictalator" has been built and is being tested now.
And finally, after cleaning up and digging through our football database, I uncovered some NFL numbers I felt compelled to share. Including the playoffs, but excluding Week 17 (where we warned against playing many of the games), we finished 117-90 ATS (57%) with ALL playable games against-the-spread. Picks with 60%+ confidence went 46-30 ATS (61%). All playable totals went 103-83 O/U (55%), with 60%+ totals going 16 -8 O/U (67%). Please contact us if you would like to see other numbers.
As usual, if you have any of your own 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.