Baseball Trial (Update 4/11/11)
Update: Monday, April 11 at 2:06 PM ET
Much to the chagrin of our SEO advisers, I am going to update this blog entry with the final numbers from the MLB trial rather than start a new one. I prefer to keep entries together that are so closely related for each of reference. We'll also quickly cover some info about run lines that a stems from a customer support conversation and briefly mention recent NBA performance.
We feel as though baseball trial was a success and we are very appreciative of interest and feedback we have received with it. From a simulation standpoint, the engine is straight-forward and the numbers are easier to construct than with any other sport. No significant engine tweaks are planned nor deemed necessary. It will be interesting to see the performance curve of baseball. I expect it to the be the flattest of our products because the season is so long and day-to-day variation is just as significant as beginning season to end season. I'll keep you all posted as we learn more from the numbers (relative to lines and performance - though that info will come in a new blog entry).
From a wagering standpoint, with money-lines, the topic is much less straight-forward than with other sports. Yes, you are just trying to figure out who is going to win, but there are degrees of relativity to that that correlate with our confidence in the pick and lead to finding value. We hope that, with your help, we have done a good job of explaining this and will continue to make improvements when/where necessary. If you ever have any questions, contact us at any time.
Here are the final numbers through MLB trial period (all assumed to be based on $50 player and -110 with no other plays on that day - profit means more than the percentages):
- Money-line Plays: 59% +$201
- O/U Plays: 61% +$573
- Normal+ Plays: 65% (13-7) +$364
- All Playable Plays: $763
We only had two losing days out of ten. Interestingly, they were the two traveling days in baseball - Monday and Thursday - when there is not a full slate of games. That's probably more of a fluke than anything, but it stands to reason that, when there are less games, the impact of a couple losses is greater - the same would be true with a couple wins.
It should not be a surprise that money-line picks had more losing days (four) than our combined overall picks. Part of that is the same phenomenon as above (i.e. the more picks in sample, the more likely we are to be right - if indeed our methodology and numbers are right). But part of that is the general nature of money-lines as well. Many days, more than half of the wagers are on events that are unlikely (less than 50% likely) to happen. The value is still there, so the picks are good. It just takes a long-term view to get the expected return.
Unfortunately, we started the "subscription"/non-trial MLB period with a losing day yesterday - kind of, it was -$7. As long as the winning days are frequent and significant and the losing days are minimal and rare, we can live with that. That's going to happen. One really fascinating note from yesterday stands out, though. And apparently, it's been true for the entire trial thus far. Playable run-line picks went 7-1 yesterday. It's just one day and I haven't gone back through the lines to verify any records previous to today, but the anecdotal evidence from our subscribers is positive. Here is one piece of correspondence regarding run-lines that I recently had on support:
Question: I've been using your baseball picks with a lot of success. Not only to I bet every playable total or moneyline bet, but I also have been using your predictalator to simulate the games and bet the run-line in the same manner. This has been even more successful. I've been betting every single playable bet both on totals and on money lines, with great success. But I've had even more success playing all the playable run line games. Yesterday, the run line picks were 7-1. But herein lies my question. The predictalator almost always picks the underdog for the run line, for obvious reasons. Simulating 50,000 games will almost always produce an average margin of victory less than 1.5 runs. Sometimes the odds justify betting on the underdog, and sometimes they don't. My question is, I've noticed that sometimes the odds suggest that you should actually bet on the FAVORITE, even tho the predictalator's "pick" is the underdog. For example, yesterday the Chicago White Sox were favored by 1.5 runs at +135 according to my sports book. After simulating the game on your predictalator, it said that the TB Rays, their opponent, were the suggested pick bc they covered the line 54.6% of the time. Which means the White Sox are covering 45.4%. At +135, they need to cover 42.55% of the time. So am I right in concluding that the White Sox is a good bet? And if so, why don't you guys recommend this as a play? Why doesn't the predictalator suggest this as the play? And most importantly, allow me to use your calculator to figure out how much exactly I should bet?
Answer: It sounds like the concern here is that the run-line picks table in its normal state assumes -110 odds for both sides of the wager (like most spreads in other sports). The same can actually be said for O/U as well. We understand that these lines are often not -110, but we are trying to find a way to prevent confusion in the tool and those tables.
As it stands, you can change the odds once you load the pick into the Play Value Calculator, but that doesn't help when the table is giving you the wrong pick (all it would do in this situation is tell you that TB is not the playable pick at those odds). In an attempt to prevent confusion, we'd obviously also prefer not to prevent our subscribers from getting the info that they need.
This is a good issue to note. We'll try to come up with some way to improve how this is done. Thanks for the head's up...
I went on to answer a couple questions the user had about specific games today and then I emailed him a simple spreadsheet that allows inputs into the Play Value Calculator to determine optimal play value. Please contact me if you would like this emailed to you for now. We'll probably add a tool like that on the site as well as modify the Customizable Predictalator for the above issue in the near future.
And lastly, on to the NBA. We'll have more info in upcoming emails and articles, but I wanted to note that our performance has significantly improved over the last 3-4 weeks. Last week, the playable against-the-spread picks went 21-11 (66% ATS). Since we shifted to posting picks at 4:00 pm ET instead of 11:00 am ET to give us better roster and injury info, the NBA picks are 65-48 (56% ATS). And despite a couple brutal nights early, playable NBA ATS picks this season are 54.9% accurate. We worked very hard to get to that point. It may not quite be the accuracy that we expect from other sports, yet that level of performance can definitely be profitable over the course of a six-month schedule. Hopefully, the changes that we have made for this year pay-off and we can continue this positive momentum into and through the playoffs.
Update: Wednesday, April 6 at 11:55 PM ET
This blog entry will recap the first week of baseball. 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... Speaking of redundant, yes, I started the basketball trial blog in the same way. The difference here is that our performance has been significantly better for baseball than how we opened with basketball. I do not expect everyone to have used our information for every game in the trial (in fact, we intentionally tempered expectations after the basketball trial and advised a more conservative approach to start baseball than most normally may take), but the numbers have been very strong in just about any context or period of time. Our only losing day thus far was this Monday, when there were only six games and our picks netted -$22. I'll take that, especially given the long-term/value nature of baseball wagering.
Similar to the NBA, yet to a much greater degree, betting baseball requires a long-term approach that is focused on value and return. Quantity can often be as important (even on a daily basis) as quality. In the NBA, that is because we rarely have strong opinions and most plays are bunched together in confidence. In baseball, fortunately, it is much easier to apply the math and predict games (whew). However, value means something a little different. Instead of wagering against a spread, the majority of action relates to which team is going to win (straight-up). All wins and losses are not created equal. There are odds associated with the chances the books give for either team to win (plus the vig or "juice" that gives books the edge overall - if their lines are right). Payouts for correct plays are based on these odds. While we normally talk about winning greater than 52.4% of the time ATS or O/U as profitable, on the money-lines, someone could hit 60% and lose money or win 40% of the time and win (more on that below). Because of that, we are often recommending picks that are less than 50% likely, meaning that the chance of the pick (an underdog) being correct is not great, but it is worth the payout associated with the odds. A long-term approach (at least a week, though my preference would be looking at an entire month if not a season) is necessary.
To a degree, we are still figuring out what the best approach is to using our numbers, yet it is clear that money-line and total picks are both capable of being profitable. There is not much difference between the best and worst money-lines, so playing several games helps. Our more risk-averse users will probably play our top playable favorites. Our more risk-seeking users will probably play our top playable underdogs. And our patient, system players seeking optimality will probably play most of our playable money-lines (assuming we prove that we can maintain a level of value/success similar to what we have provided thus far). With totals, it's a little more obvious. Normal or better totals should be played more often and to a greater degree than weak plays (if weak plays are played at all). This would be similar to the money-management strategies that we have seen in football for ATS and/or O/U plays. We do assume, however, that our confidence (as noted by cover percentages) in O/U plays will decline over time as the market becomes more efficient. This is our first endeavor into baseball though, so we cannot confirm this trend.
These money-line odds are expressed as +/- numbers in the hundreds (or thousands, though not often in an MLB game). Favorites are expressed as minus numbers, like -150. Underdogs are positive numbers like +150. For favorites, one must risk that amount to win $100. For underdogs, if $100 is risked, that much is won (in addition to the risk). Calculating whether the confidence that we have in a pick winning straight-up (SU% proj) exceeds the percentage related to the payout from the odds (SU% needed) is pretty easy. For favorites, the confidence needed is the odds divided by the odds plus 100 (drop the negative sign, so technically: absolute value(odds)/(absolute value(odds)+100)). For instance, a -150 favorite must win 60% (150/250) to warrant action. For underdogs, he confidence needed is the 100 divided by the odds plus 100. For instance +150 yields 40% SU% needed (100/250).
We may think a game is exactly 50/50, but love the underdog as a very strong play if the odds payout for it is +150. That would be like someone telling you that you can bet a dollar on a coin flip. If you bet on heads and you win, you win $1.50. If you bet tails and it wins, you win 67 cents (assuming no vig and favorite odds are -150). If you lose, you lose your dollar. You would bet on heads every time (you'd better). That's the same concept in play to varying degrees in baseball.
I'll answer most of the other questions that we have received in the customer support Q&A below. For now, I want to talk about a couple quick trends and our results thus far.
We have already made one small tweak to the baseball engine and just made another tweak to the Play Value Calculator (PVC) tonight... I was one who generally held the belief that teams win at home far more because of ballpark fit than homefield advantage from fans, familiarity, umpiring bias, etc. That is true to a good degree - just not quite to the degree that I had originally assumed. On Monday night, we modified the homefield advantage to be more appropriate - especially relative to recent seasons. It's a small sample size, but I would expect that our total performance should stay the same from before the tweak, while our money-lines should improve.
The change we have made with our PVC is to better account for overall bankroll when determining optimal management. Previously, we had assumed that everyone's normal play was 5% of bankroll and that all "other plays" were at the normal amount. We also allowed "$0" to be displayed. Now, we will allow for more dynamic management by incorporating the actual/expected average play for the day based on one's normal play and the levels of confidence we have in our picks for that day. We also assume that 3% of one's bankroll is a normal play (our recommendation is 1%-3%, not exceeding 5%). And, we will not return "$0" plays. We will still, however, not allow our users to put more than half of one's bankroll into play at the same time based on these implications. This is for your protection and the whole reason that we use the "other plays today" dropdown. It can be a very viable and profitable strategy to ignore that dropdown. But, on a bad day, which will happen, it could also be catastrophic.
Here are the numbers through the first seven days of the MLB trial period (all assumed to be based on $50 player and -110 with no other plays on that day - profit means more than the percentages):
- Money-line Plays: 59% +$299
- O/U Plays: 57% +$417
- Normal+ Plays: 65% (13-7) +$364
- All Playable Plays: $716
- Cost of Full Season Package: $399.95
Obviously, we do not know what the future will hold for the MLB predictions product (and we understand that we cannot expect to sustain all of those winning percentages), but we are very happy with these results for the trial thus far.
Also, another way to look at this info is to say that we are +14.32 Units ($716/50) in five days. It's not terminology that we like to use because we want to be more precise than most in the industry who typically deal in whole number units for plays based on their perceived confidence (most of these picks were less than "one unit" plays, while none warranted two units or more). It's very important to note that this is not even apples-to-apples with the industry because of our fractions of units and conservative (aka "smart") approach. Technically, if every playable play was considered one unit or more, we'd be +$29.38 units. Of course, we also might have bled our bankrolls dry with that manner of thinking. However, we have received some questions about that. In typical terminology, units refer to the amount of your normal wager that you can put on a game. This often leads to tricky sales tactics regarding multiple unit (or star or dime) games and the concept of "chasing" to get back to positive units (where we would recommend playing less for the same confidence if bankroll is declining which is the opposite). Anyway, it's not a topic or rhetoric that I prefer, but it would fall into the bucket of "frequently asked questions" so I thought I would touch on it.
Customer Support Q&A (apologize for redundancy here)
Question: I'm new to the whole Baseball betting thing. My question is how do you pick the plays for the money line picks? What is SU% needed? What distinguishes the money line picks from the SU picks?
Answer: Money-line picks are ordered by value, which is based on the relationship between the Predictalator’s projected Straight-Up winning percentage and the Straight-Up winning percentage needed to justify a pick. Since when wagering on a favorite, it costs more to win less and vice versa with underdogs, determining the SU% needed for a favorite can be calculated by taking the absolute value of the odds divided by 100 + the absolute value of the odds. For underdogs, the necessary percentage is 100 divided by odds.
For instance, if the Texas Rangers are -156 favorites over the Seattle Mariners who are at +145, Texas must win more than 60.9% (or 156/256) to justify a play on the Rangers. If Seattle wins more than 40.8% (or 100/245), the Mariners have value. If neither wins at those rates or higher, it is a “no pick” because neither team has value. The gap between the projected SU% and the necessary SU% dictates our confidence in that pick and is used in the Play Value Calculator to determine the optimal wager amount for that pick. If the wager amount for a normal $50 player (the default setting) is $1-$49, this is a “weak” pick. If it is between $50-$99, it is considered a “normal” pick. Between $100-$149 is 2X normal pick and anything stronger than that is considered 3X normal pick.
Straight-up and money-line percentages are the same. One table caters to gamblers that need to know where the value play is relative to payout (the money-line). The straight-up table caters to fans who want to know their team's chances of winning and the projected score. They both use the same results, but they appeal to two different angles.
Question: For the first few days of MLB Trial, you noted that all playable picks are +$490 (encouraging, despite being a small sample size). I'm struggling with determining what an actual "playable" pick is, though. For example, there were 10 MLB money-line picks yesterday (4/3) that were Weak Plays and three O/U picks that were 56% or better (since those are the results you tracked in the blog, they're the easiest to use in this example). I think this makes 13 “playable” picks yesterday.
However, as the Other Plays Today number increases in the Play Value Calculator, most games are a suggested $0 play. As you’ve noted, bankroll management is the key to profitability and I’d like to follow the suggestions from the calculator as closely as possible, but I want to make sure I’m clear on when you think a game is “playable”. Despite 13 falling into the Weak or Normal play range yesterday, the high number of total plays resulted in the majority of them being a $0 play.
Do I simply go with the calculator and only wager on games where the Suggested Play is greater than $0? It seems counter-intuitive when games appear to fall into a playable range but the Play Value Calculator then essentially says not to wager. ?
Answer: Great points. Thanks for this feedback because it has helped us find some areas of bankroll management that we want to review for baseball.
In general, your last paragraph would be accurate based on how we have provided analysis for football and basketball. However, for baseball, where there is more value in quantity/long-run approach as opposed to just playing the few highest value plays, we are evaluating how to better account for "$0" plays. In today's numbers with "four plays" you would have three $0 plays. At $0, you would no longer have four playable games and would want to change the bet amount in the first game to just playing one game.
That's really confusing and probably not "right." We're working on making this more intuitive and appropriate for baseball.
With moneylines, the normal+ picks relate to the Play Value Calculator. The percentages in the table are the necessary percentages at the standard -110 odds that we typically see for ATS and O/U picks in other sports. I think I will remove the percentages in the tables because they do not apply to everything we do so well any more.
It's not the world's most straight-forward formula. Essentially, any time the difference between percentage needed to cover and projected percentage is significant to justify playing $50-$99 on the game (for a $50 bettor), it will qualify as "normal" (a one unit play). If the pick is $100-$149 it will be 2X; $150+ is 3X and $1-$49 is weak.
When it comes to a "$50 player" we are talking about using the play value recommendations from the calculator ($50 is default). So far, we have had some money-line games where up to $78 can be risked. There are no plays today (Sunday) greater than $25.
Also note that the aggregate numbers that I quote include all totals. "Playable picks" include everything that is not a "no pick." In other words, any O/U that is greater than 53% and any money-line where we find value in either team.
We have had 7 previous totals that covered greater than 57%. These are 6-1 O/U and +$361 on their own so far (as of Sunday). Two of those plays are available today. The O/U table still uses 57%+ and -110 as the defaults for normal+ even though O/U odds are often different than -110 (the table gets ridiculously complicated if this is added in and changing odds is easy to do in the calculator).
We have reviewed the Play Value Calculator recommendations/algorithms and will be making a couple of tweaks relative to "Other Plays Today." We are also going to update the "explanations" on the articles pages. Those tweaks should be available in the next couple of days. I will provide a blog entry afterwards... (This is that blog entry.)
Question: Just curious as to what data you are using, last years I assume, to predict the current mlb games. I used software but can't rely on plays until there have been at least two rotations in pitching. I.E, Garza for the Cubs today is switching teams/divisions so his stats last year are useless to me currently.
Answer: We go back throughout every player's career, looking at minor league stats and even Japanese leagues, etc. Players are fit to an expected performance curve based on age and experience. We also weight the most recent 486 games (three full seasons) heavier and then even more so for the last 162 games (one season). Probably the most important thing that we do in addition to that data analysis is that we ballpark and strength-of-schedule adjust the numbers so that we have a good representation of what the player should be expected to do against different caliber players and in different ballparks.
Question: When the play value calculator says to play for $50, is this meant to be the risk amount or the win amount? Makes a pretty huge difference in the results.
Answer: That is a good question and rarely asked, but, as you said it can make a big difference, especially when looking at money-line implied odds for value. The play value calculator, with its input of other daily plays, is designed to optimally manage how much of your bankroll you are risking at any given time relative to confidence that the pick will win and other areas you are risking bankroll at the same time.
In other words, you should be entering the value you wish to risk. What you win may be significantly greater or less than this value based on the odds.
This is a good distinction to make as well because this is how we will be tracking our performance as well. For instance, we are +$40 on money-lines so far today (though Seattle is not looking good). Among the wins are St. Louis for a risk of $15 that won $9 at -165 and Baltimore at $14 with +135 odds (when we published it) that won $19.
Question: How come you don't show the full lines for the no picks in the Money Line section? I would still like to see the predictalator's math for those games as well. Also, how do I check the previous days' lines?
Answer: Money line picks are related to value. When there is no pick, then choosing whether to show the favorite or the underdog would be arbitrary and potentially misleading. This is why we include the straight-up table below that with the same information. Changing the URL to any previous date will display that day's picks.
Question: What is the best way to play the MLB plays? Is it more like college football where you are probably ok playing the Normal and better plays? Or is it better to play it more like the NBA bc there are more weak plays and not as many Normal or better plays? Or maybe I am off on both of my examples and you have better approaches altogether.
Answer: To a good degree, we're still learning about this. It seems as though totals are more successful, though that could be a bit of an aberration. It's not that we would expect to do significantly worse than we currently are with over/unders, but that value for betting money-lines has to be figured out in the long-term. We could hit 60% of our money-line picks and lose money (if we are only hitting -200 favorites for instance) and we could only hit 40% of our picks and make money. Totals are far more straight-forward.
Thus far, in five days, playing all playable picks would have returned a profit on four of five days and would be up $451 for a $50 player. It's not realistic nor even recommended to play all playable plays, though (as we have discussed in the past - though baseball is different with the amount of games per day and the reliance on value rather than plays that are expected to win 50%+). I would recommend playing normal+ totals and continuing to monitor money-line plays.
We actually just made an engine update today that should improve money-line plays (I will blog about this today or tomorrow). We will be making another update (likely tonight) that will improve the Play Value Calculator recommendations for playing multiple plays. This should help quite a bit with baseball wagering.
Eventually, risk-averse subscribers will probably play more of our "safest" favorites in MLB, while the more risk-seeking subscribers will play our more likely upsets.
To more directly answer your question, I think you can treat totals like you did in college football - playing top rated plays; whereas, money-line plays should be evaluated more like the NBA where it will take several plays a day (or at least per week) to create value.
Question: I've been following your MLB picks pretty carefully, looks like there's a lot of potential here. One quick question, do you always default to -110 for all over/unders? Most of the time I see anywhere from +105 to -125 and sometimes this changes the play between weak/normal/no pick. Or is this something to always use the PVC for?
Answer: Thanks for the note. We assume -110 for O/U picks when it comes to assigning the category and tracking, despite the fact that we know that this is not always the case. The table and concept can become significantly complicated by adding more info. I recommend using the PVC and altering the odds there (this is fairly new). This will provide the "correct" play values.
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.