Basketball Review (3/7/11)
Monday, March 7 at 6:12 PM ET
This week's blog entry is more to the point than most and directly related to last week's, so I am just tacking it on. I will recap the second week of the basketball subscription and briefly set up the next few weeks to come. I accidentally overwrote the spreads we have in our database for tracking accuracy, so I had to go back manually and track our performance. This did not affect the posted picks at all - just how we track them after the fact. I only mention this because, if our numbers off, that is totally not intended. Please just let me know if you see something off and I can update the numbers if necessary.
We're deep into preparation for the NCAA Tournament (watch for the Bracket Odds by next Monday morning - likely no blog next week - the March Madness package will be available for purchase this Wednesday) and baseball (the free 2011 MLB preview will be up by the end of this week - while I thought we'd be all over Boston and Philadelphia, the expectations there are too high; early on, I love the value in the Chicago White Sox). We also have not received many new topics through support over the last week or so (that being said, with every day in basketball being a good day or a bad day, we definitely hear from people on the latter, especially as opposed to the former; in other words, we admittedly received a decent amount of feedback late Wednesday night and Thursday morning).
Here are some general observations from the last week (I've already broken a few grammatical rules above and will now commence a bit of a stream-of-consciousness, run-on paragraph of "quick" thoughts). We are definitely happy with the tweaks that we have made over the last month - all announced - as we have seen overall performance improvement, most notably with "normal" or better plays in college... As I just alluded to though, there can be bad days and we did have one difficult day on Wednesday night. Without that day, our accuracy would look tremendous, but I know it doesn't work that way. With the strategies outlined below, which focus on all playable picks in the NBA and the "normal" or better plays each day in college, looking to find more winning days than losing days is a major objective. Fortunately, we have had more winning days than losing days under that pretense (Monday through Sunday, we had four winning days and three break-even or losing days in the NBA and five winning college days to just two break-even or losing college days - though that one losing day was obviously a rough one in college)... As a reminder, while this blog gives "big picture" detail, all picks are published for all to see as soon as the day is complete. Even if you are not a subscriber, you can check in after the fact to see how we did. We never remove articles, so just update the date in the URL to see what you are looking for... College totals got better, but are still not where we want them to be... I will never again pick Drexel covering over Towson as our top play of the day (for those who didn't keep track, we went 0-2 in the last two weeks in college top plays of the day that involved those two teams)... Conference tournaments are starting now, which makes me happy because it means that we can, for the most part, ignore homecourt advantage (regular college basketball homecourt advantage is the greatest challenge I've faced in doing this professionally) and we can more rightfully assume that everyone is playing his hardest to win the games. Theoretically, as it was with the NFL Playoffs and last year's NCAA Tournament, the next four weeks of college basketball should be our best... We are still struggling with a couple of areas in the NBA: 1) We rarely have many significant opinions and 2) Injury information is almost impossible to track... With #1, the low sample size is clouding our performance a little bit because it makes it hard to tell what our "true" accuracy is (or more appropriately, if our accuracy matches our confidence). As discussed last week though, playing all playable sides (and/or totals) appears to be a winning strategy, it's just at a lower per-game margin than most of our subscribers expect because of the low opinion. There is definitely still value there, the payoff is not as immediate (like it is in football for instance)... And with #2, Tony Parker just about ruined the weekend by showing up on the court Friday night (Kevin Durant playing at the same time had a similar affect). Parker was listed as "out" and expected to miss multiple weeks, yet he showed up to play. Did he mean 28 points to that team against the Heat? Not directly, but maybe indirectly. While we do our best to project which players will be in each game, it's just about impossible to get everything right. We are looking forward to next season when we can introduce starting lineups to the Customizable Predictalator. NBA teams over the last week were tremendously inconsistent. Injuries and player movement are not the whole picture, but they have played a big role and the biggest piece of the puzzle that seems to be off when our picks are off... There is definitely value in NBA and we hope that its subscribers can ultimately have the patience (and bankroll) to eek it out to turn a good net profit. That being said, it is important that we do everything we can to see that our subscribers get the most value from their investment with us. Over the next several weeks, we will be contacting those subscribers to do what we can to add value to this service... Run-on paragraph concluded.
Let's dive into the numbers for the subscription period to-date (Monday, February 21 through Sunday, March 6). ATS Top Plays of the Day went 5-2 ATS (71.4%) last week and are now 9-5 ATS (64.3%) in the subscription period (this should also be a profitable strategy if you have the patience for it and prefer high per-play margin over more consistent profit with lower per-play margins). Interestingly, as player movement from trade deadline and major injuries is settling in the NBA and conference tournaments have already taken over a great deal of the college basketball action, all playable sides over the weekend went 40-29 ATS (58%) and normal or better college sides went 12-6 ATS (67%).
All Playable Games: 54.0% ATS (20 games over .500)
"Normal" Or Better: 55.0% ATS (8 games over .500)
All Playable Games: 45.9% ATS
"Normal" Or Better: 37.5% ATS (just 16 games - only one last week with tweak)
All Playable Games: 55.0% ATS
"Normal" Or Better: 42.1% ATS (just 19 games)
All Playable Games: 54.5% ATS
"Normal" Or Better: 50% ATS (just 2 games total)
Tuesday, March 1 at 7:42 PM ET
This week's blog entry will recap the first week (and a day) of the basketball subscription and discuss a couple of minor tweaks as well as feature a couple recent questions from subscribers. The last ten days or so for me have been a little chaotic, so this blog may not be as long as most (famous last words), but I hope to cover any underlying questions out there. something that we touched on during football is being magnified now with a daily product, generating quite a bit of feedback and great discussion. Unless we get every game right or every game wrong each night, we will have some subscribers who win and some subscribers who lose EVERY night. There is really no way around it. Our goal is to provide the most thorough and accurate information that is consistently accurate (very difficult to do, yet very frustrating to the consumer if it is not) and ultimately lives up to the confidence that we give it. We obviously want to provide the best possible user experience to all who visit the site. While we are happy with the improvements to the site and to our performance over a short period of time, nothing is ever going to be perfect, so I hope that it is understandable that we continue to take your feedback and make improvements to the site. Let's dive into the numbers to discuss where we excel and what we are doing when we do not. Overall, ATS Top Plays of the Day are 5-3 ATS (62.5%) during the subscription period.
All Playable Games: 53.5% ATS
"Normal" Or Better: 56.3% ATS
"Upset Watch" Picks: 14-10 SU
Notes: Our "normal" or better plays (57%+ to cover) started very strong and only had one sub-.500 day (Saturday, they went 7-9-1 ATS). Overall, the picks are profitable playing every game (need 52.38% to turn profit) and pretty strong (six games over .500) with normal or better picks. That's the optimistic way to look at it. The realistic view is that we are slightly below our thresholds for the confidence we project. We're close, but not quite there and we've noticed a couple of ways in which we can improve (both of which have already gone into action).
One way in which we can improve is to better account for the compounding nature of injuries in college basketball. Taking one player out of a lineup and plugging another in is very straight-forward with our system. And through all of my research and working in this field (almost eight years now), I got very good at setting up the algorithms to understand how certain players make others better (or worse) while on the court and how the team is impacted then when lineups change - in the NBA. While I've also worked considerably with college basketball, it's actually pretty rare to see major injuries period, especially those that can completely change teams. Kyrie Irving goes down and he is replaced by Nolan Smith. Not a big deal for us to figure out. However, when Chris Wright at Georgetown (for whom this entire paragraph is essentially written) goes down and he is replaced by no one (seriously, they don't have another point guard), it means a little bit more than we anticipated. Anyway, we have done some extensive analysis into this phenomenon and feel like we are better accounting for this. For the most part, you'll probably only notice changes with Georgetown unless a major player with no obvious back-up (who has a significant statistical base) gets hurt. However, we will also better account for reported injuries that do not cost players time, but hurt their performance, such as what we have seen with Derrick Williams and Arizona recently.
The most significant change/tweak we have recently implemented is similar to what we did with our confidence in inconsistent teams in weaker conferences (a change that has very positive returns and will likely help us greatly in college football as well). Previous to today, we were simply ignoring all games with lines over 16.5 points because this seemed to be the most appropriate cutoff when performance started to wane when motivation to win a game by X points dwindled. The - 16.5-point line cutoff for college basketball picks was probably an overly simplistic/borderline lazy approach. There is no magic cutoff where motivation to win by X points is drastically different than otherwise. So we have implemented a strategy to temper our confidence in the higher lines, rather than just have a cutoff (we will now be publishing all games, including those with very high lines). There is no reason for us to LOVE Kansas -14.5 @ Oklahoma. Is Kansas at least 15 points better than the Sooners in Norman? Definitely. Do the Jayhawks care if they win by 19 (which is what they were winning by with two minutes remaining) or by 12 (which is what they ultimately won by)? Not really. The higher the line, the less confident we will be in the favorite to cover it. This doesn't change the pick, but it does (appropriately, in our opinion) change the confidence we have in it. Simply applying this tweak to last week's picks nets 62.2% accuracy on normal or better plays and kicks several of the low-end double-digit favorites out of the "playable" range. We can't guarantee that we will replicate those results, but we do feel that we are better handling the issue with motivation. This is another tweak that will likely translate to football, yet in a slightly different way due to the way scoring is handled there.
I am very happy to see the success of the upsets because there is value there. Until we have our money-line calculator in place (launching with baseball picks), I recommend sticking with projected upsets when looking for value in college basketball money-lines. If you do want to venture into mining for upsets even if we are not predicting them to be more likely than not (such as with Kansas State last night), I would stick to major-conference teams. The logic on that gets tricky, so I'll spare you here, but you can read the question and answer below for more info.
With daily picks ranging from five total games where only three are playable to 100+ games, the learning process and debate continues as to how to optimally assess bankroll. Again, all final decisions are left up to the individual and his/her risk-reward preference. That being said, it appears fair to play all (or most - on a day like Saturday) "normal" or better plays, especially with the tweaks discussed above. Also, with conference tournaments on the horizon (pun, somewhat intended since that conference's tournament starts today), there will be opportunities for the rest of the season to make decisions on bankroll management throughout the day, rather than having to decide how many plays to make each day. And lastly as it relates to bankroll management, please use the Predictalator to assess plays at your lines.
All Playable Games: 43.5% ATS
"Normal" Or Better: 40% ATS (just 15 games)
Notes: Ouch. I wouldn't fade just yet. One brutal day (Thursday, 2/24) accounts for most of that and we have worked in some minor tweaks to better account for games in which two extreme teams play each other. The results we have seen in testing these tweaks produce far fewer totals with strong opinions, but at least we are getting them right far more often when we do have opinions. I understand if you want to stay away right now. It takes a lot for me to publish those percentages, but I do so in the name of transparency and with confidence that we can flip them over the next two weeks of the season and into the tournament.
All Playable Games: 56.8% ATS
"Normal" Or Better: 40% ATS (just 10 games)
Notes: The best way that I can think of to look at this is that we have had five winning days, one break-even day (though we lost our top two picks that day, so it's a losing day by our play value calculator) and one losing day. It's important to view that way because I think that is the way the picks should be played. We so rarely have "normal" or better opinions that they are subject to sample size as an issue. While they should ultimately win more than 57%, that takes quite a bit of time when there are only a few of those plays a week. The better recommendation for those looking to get the most value out of the NBA picks is to look at all playable picks (the same can go for totals too - we'll get there in a second).
One of them is doing a better job of staying on top of who is playing. Oddly, this has been much easier to do in college than in the NBA, even though there are 315 more teams in college basketball. Injuries, especially meaningful ones, are actually rare in college, where as they are more prevalent in the NBA, notably between back-to-backs. Last week in the NBA was frustrating from the standpoint of what I have to do. Not that we were unable to keep up with player movement; it's that the players moved didn't always play (and some of the healthy ones still haven't). I would assume that will even out as we focus and devote even more resources to staying on top of teams. It's a fine line to walk between getting the picks up and sticking by them and updating frequently within the scope of a day. We can improve. This impacted pick performance last week, particularly with the normal or better picks on Tuesday and Friday. We addressed everything that we could in reasonable time, yet most information broke too late to make changes. In the last three days, when player movement and issues have died down considerably, we are 10-4 ATS (71.4%). Obviously, we don't expect to always be 70%+ ATS, but it's a good sign.
Along these lines, we are targeting a significant update to the Customizable Predictalator for next NBA season that will allow users to set the starters for games as well as the lines to get the most up-to-date pick information. Hopefully, this can take some pressure and uncertainty off of us, while giving you more control and the best possible information. If all goes well, we will attempt to roll this out for all professional sports.
All Playable Games: 61.3% ATS
"Normal" Or Better: 50% ATS (just 2 games)
Notes: For the life of me, I can't figure out how our NBA totals can perform so strongly, while our college totals do not. If I am going to blame part of the poor performance in college on one bad (horrific) day though, I need to acknowledge that we had one tremendous (6-0 O/U) day in the NBA that has buoyed this great performance.
Customer Support Q&A:
Question: On straight-up NCAA hoops picks, is the "flip side" of the Predictalator's pick equal to 1 minus your pick's win %? So, for example, Texas was your pick in last night's game vs. Kansas State with a win % of 67.5%. I assume that means Kansas State's win % would have been 32.5%? My book had K-State at a +350 to win outright, so a 22.2% win rate was needed to break even. Hopefully that example makes my question make sense. I'm trying to better understand the SU plays, primarily as MLB comes up, but also to have other avenues of finding value.
Answer: The answer to your question given your example is, "yes." Kansas State on the money line was a strong play. That being said, I would be careful with non-upset money-line picks, especially in weaker conferences. Weaker conferences typically have inconsistent teams, which makes the likelihood of a favorite winning a little less likely than a similar situation for a more consistent team. It's hard to make your flip-side argument make perfect sense in that case.
Let's use Cleveland State hosting UIC as an example. Cleveland State is a relatively inconsistent team, which is reflected in its SU% and ATS%. So, even though we had Cleveland State and another, stronger team, San Diego State, win tonight by around the same points, we have more confidence that SDSU will win at all because it is more consistent than CSU. But if you are trying to make the opposite point, it gets tricky. If you have to choose between wanting Wyoming to pull the upset or UIC, it's not as clear. The answer is still that we have more confidence in UIC to win because CSU is less consistent, but it is not as direct a relationship in these situations.
Long-story short, there is definitely value in the money lines, but I would probably stay away from non-major conference money-lines unless we are predicting an upset - at least until we have the money-line calculator to explain better. There was great value in Kansas State, just as there is good value in Boston College tonight (a similar situation). But I would stay away from UIC and Gardner-Webb, etc. for now, even with the big pay days associated with those.
Question: And separate question but what are your plans for NIT and NCAA Tourney picks? I assume they were/are not part of the regular season package?
Answer: Yes, the NCAA tournament will be sold separately, though current college and NBA subscribers will receive significant discounts on that package. Full tournament packages will become available mid-next week, with daily picks packages for the first weekend of the tournament soon to follow. We did not sell NIT last year and will not do so again this year. Motivation is just too much of a concern with any other tournament besides the NCAA Tournament. A reasonable analogy can and has been made between the bowls and the NIT, though the NIT isn't the only thing going on at the time, so it represents all of the difficult parts of bowl season.
Question: I have a question about using the Predictalator. Here's the scenario from 2/27 North Carolina vs. Maryland. The line on the site was at -6 and in the "normal" play range. My book had it at 6.5 and when i use the Predictalator to run the simulation, it falls out of that range. I did not play it. On the other hand, the posted total on the site was at 147 and was to play the under. This however was a "weak" play at 147. My book had this total at 148.5 and when i plug the numbers in this time, it's in the 58 percent range and definitely a play. Unfortunately, the game went over but Carolina would've covered the 6.5 anyway. I know that's just 1 game but my question is: Am i using the Predictalator correctly? I would say that yes and i want to keep a regimented system with what falls into the proper range for a play. However, it seems like public money moving these lines one way or the other ends up having a large effect on what to play and what not to play according to UTP.
Answer: While my standard answer to this topic is that randomness is involved and that, while I am surprised by that much of a swing, it's an anomaly, your empirical evidence and my follow-up testing suggests that there is some element of excess randomness related to the totals. Playing a game 50,000 should not result in such significant swings. For now, I would stick to looking at the o/u picks published and I will research this situation in the next day or two to see what we can do to address it. It makes sense for the picks to move in the direction that you mentioned above in this support contact. However, it does not make sense for picks to shift as much without changing the line. We do not want that ambiguity and forced confusion for our users.
Question: It just hit me when I was reading one of your earlier blogs in which you listed your results from the past football season - specifically, can you back test your results to see what your records in just the Major Conferences (Big 10, Big 12, PAC 10 etc.) were ATS, LOCKS OF THE WEEK, PP's ). I bet you even had a higher overall success rate in these conferences is why I ask.
Answer: Interesting question and alludes what we have seen with basketball (as you read in the blog). Our Locks of the Week that involved two BCS-AQ teams went 6-0 ATS this year. If our Locks of the Week had just been our top play between two BCS-AQ schools last year, we would have been 12-4 ATS through the bowls, which is even better than we were. And, 60%+ picks between two BCS-AQ teams, were over 60%+ ATS, while all other games were slightly under 60%+ ATS. There is something to this and we'll work it into our football picks for next year.
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.