Ganch, HELP
I lost my data and I'm looking for value of 1/2 pt at each different line.
I lost my data and I'm looking for value of 1/2 pt at each different line.
-
-
<script> function OpenCalc() { var winTool = window.open('http://bettingtools.sbrforum.com/newcalc/half-point.html', '_blank', 'width=215, height=275, resizable=yes, status=no'); winTool.focus(); } </script>
<input style="font-size: 13px;" type=button value="click to open half-point calculator beta" onClick="OpenCalc();">Comment -
Ganch,
Do you have anything for adjustments in regular season total win props? For example, these 4:
Where would you place the vig (with a 30c line) if we switched the total to 9 regular season wins?9/1/2007 7:00 PM California regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
211 Over 9½ reg season wins +220
212 Under 9½ reg season wins -280
How about if this were to read 9 regular season wins (with 30c vig)?9/1/2007 5:00 PM Florida regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
215 Over 10 reg season wins +320
216 Under 10 reg season wins -400
At 9.5 with 30c vig?8/30/2007 6:00 PM Rutgers regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
251 Over 10 reg season wins +220
252 Under 10 reg season wins -280
At 10.5 with 30c vig?9/1/2007 11:00 AM West Virginia regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
269 Over 10 reg season wins -180
270 Under 10 reg season wins +150Comment -
Ganch- That is truly unbelievable. I've never seen a chart so comprehensive and user friendly. It's the most useful thing that I've ever seen on any forum.
What data did you use and from what years?Last edited by raiders72002; 08-14-07, 08:14 PM.Comment -
Thanks. I appreciate that.
The data used is from Covers.com, and stretches back as far as the data was available for each sport with the exception of NFL. For NFL the data only goes back as far as the 1990-1991 season.
The data is only minimally transformed. I'm not currently adjusting for such changes in regime such as the advent of the NFL 2-point conversion in 1994 or the change in OT rules for NCAAF. I'm basically just following the Stanford Wong methodology, except I equal weight all games that lie the same distance from the target point spread or total.
I plan to improve on the sampling methodology in the very near future. Users retain the option to enter their own probability estimates.Comment -
Although it may seem otherwise, the methodology for evaluating half-points in season win totals is actually rather different from evaluating half-points in spreads and game totals.
If you really wanted to come up with exact figures the first step would be to come up with win probability estimates for each game of the season, and then (either from first principles, or probably more easily using a Monte Carlo simulation) come up with estimates of winning any given number of games. Of course even that methodology would be flawed as there almost certainly exists some correlation between single-season game results.
Without considering individual match-ups, using the binomial distribution we can still come up with a first-order approximation of win total likelihood based solely on a single "fair" estimate of the likelihood of any given win total.
Using this methodology, I'll go to go through each of your examples one-by-one:- A market of o9½ +220, u9½ -280 corresponds to a zero-vig over probability of 29.781%. Using Excel and Excel Solver, we see that a single-game win probability of 71.779% would correspond to a 29.781% probability of winning more than 9 games out of 12 (=combin(12,12)*71.779%^12 + combin(12,11)*71.779%^11*(1-71.779%)^1 + combin(12,10)*71.779%^10*(1-71.779%)^2 ≈ 29.781%).9/1/2007 7:00 PM California regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
211 Over 9½ reg season wins +220
212 Under 9½ reg season wins -280
Where would you place the vig (with a 30c line) if we switched the total to 9 regular season wins?
This then implies a 25.010% probability of winning exactly 9 games (=combin(12,9)*71.779%^9*(1-71.779%)^3 ≈ 25.010%), so a probability of going over 9 conditioned on not exactly 9 of 29.781%/(1-25.010%) ≈ 39.713%.
A 39.713% corresponds to US odds of about +151.807 on the over.
Given vig of 30c this corresponds to a market of approximately:
o9 +140.73
u9 -170.73 - A market of o10 +320, u10 -400 corresponds to a zero-vig (conditional) over probability of 22.936%. Using Excel and Excel Solver, we see that a single-game win probability of 75.802% would correspond to a 22.936% probability of winning more than 10 games out of 12, conditioned on not winning exactly 10.9/1/2007 5:00 PM Florida regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
215 Over 10 reg season wins +320
216 Under 10 reg season wins -400
How about if this were to read 9 regular season wins (with 30c vig)?
A 75.802% single-game win probability implies a probability of winning more than games of 41.589%, and of winning exactly 9 games of 25.757%. This corresponds to a conditional probability on the over 9 of 41.589%/(1-25.757%) ≈ 56.017%, corresponding to US odds of about -127.361.
Given vig of 30c this corresponds to a market of approximately:
o9 -145.12
u9 +115.12 - Over 10 conditional prob: 29.781%8/30/2007 6:00 PM Rutgers regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
251 Over 10 reg season wins +220
252 Under 10 reg season wins -280
At 9.5 with 30c vig?
Single game conditional win prob = 77.857%
Prob of > 9 games = 48.378%
48.378% => US odds of ~ +106.705
Given vig of 30c this corresponds to a market of approximately:
o9½ -107.00
u9½ -123.00 - Over 10 conditional prob: 61.644%9/1/2007 11:00 AM West Virginia regular season wins Wager is for regular season games only. Bowl and Conference Championship games not used in wager. Team must play all regular season games for action. Max wager $500
269 Over 10 reg season wins -180
270 Under 10 reg season wins +150
At 10.5 with 30c vig?
Single game conditional win prob = 84.800%
Prob of > 10 games = 43.569%
48.378% => US odds of ~ +129.523
Given vig of 30c this corresponds to a market of approximately:
o10½ +117.40
u10½ -147.40
So anyway, that's it. I don't think I've made any silly computational errors, but it's always a possibility. I've sort of raced through, so if anything isn't clear, just ask.Comment -
All of this ciphering is way over my head.
When I bet football I just make sure that the + is over the next combination of field goal / touch down. So, if the + is 3 I buy the 1/2 point to get 3 1/2, if the + is 5 I buy 2 1/2 points to bet 7 1/2...
If the points are -, i buy to get below the next combination. so if it was -8, i would buy 1 1/2...etc.
Can't tell you if this is the "best" thing to do mathematically. I just know that psychologically I would be a basket case if I didn't buy that extra 1/2 point and then lost a game because of it.Comment -
Buying points on a consistent basis will put you in the poor house.Comment -
Take the NFL week 1 Pinnacle line, Tampa Bay +6 -106 at Seattle -6 -104. That's vig of 1.914%.
Take Tampa at 7½ and you'll be laying -151 at Pinnacle.
According to the half-point calculator, buying Seattle up to 7½ (paying -151) will cost you an extra 1.889% of vig for a total of 3.803%.
It is of course up to you to determine for yourself whether paying twice as much in vig is worth the possible basket case risk.Comment -
According to another chart I saw you publish, which listed the probabilities of scores ending in factors of 7, 10, and 3, I assumed that making sure the point spreads are above these factors for a + and below these factors for a - was the point of the chart.
I think there is more to this than the vig, given the probabilities of certain totals being reached chart.
Or maybe I misunderstood.Comment -
I don't know if 1.8% vig matters or not. I don't have a way to calculate it. I do know that in the NFLX, I am 8-2 because of buying 1/2 point in one game and buying 2 points on the OVER from 35 to 33 on another game. Without that I would be 6-4.Comment -
I'm not sure I'm 100% following your train of thought, but it doesn't sound as if you had been properly utilizing the push probability chart.
I'm definitely not not following you here. Vig is a measure of expected loss. When determining whether or not to buy or sell a half-point the only other issue are the odds received (lower being better, ceteris paribus, for an advantage bettor attempting to maximize bankroll growth).Comment -
You might want to check out An introduction to expectations and theoretical hold, which is the follow-up to An introduction to betting lines and percentages.Comment -
Yes, I understand advantage play, I am an advantage blackjack player. But, advantage is very different in blackjack. If I know the composition of the remaining cards I know my advantage. Advantage in handicapping sports is altogether different. To place a bet I have to pay vig, the only way I turn that into an advantage is if I can handicap the game so that the actual probability of outcome is different than the probability of outcome implicit in the odds I am getting. I would then only bet in cases where I thought the true probability was greater than the quoted odds by a large enough margin that the vig could be overcome.
And there's the rub. How does anyone know what true probability is for a sporting contest? You can guess, but you can't know for sure. In blackjack I can tell you with mathematical certitude what the expectation is for any composition of the remaining cards to be dealt.
When I read the push probability chart, which had larger probability at totals like 3, 7, 10, 14, 17, 20....etc., than for any other numbers, I took that to mean that I didn't want to have spreads of those numbers or else I end up with a push instead of a win. I don't like betting my hard earned money only to get a push.
So, I figured, if I am betting a football team using the spread instead of the moneyline, I would buy enough points to stay off those numbers. For 3 that is easy, buy 1/2. For the others it is more expensive.
I have only been betting sports a short time, my results so far are:
MLB turned original bankroll into bankroll times 5
CFL 12-1 including one $2000 win
NFLX 8-2
For the CFL and NFLX some of those wins are losses if I didn't buy the points.
Since I don't expect to get anywhere near the law of large numbers where 1 or 2% of vig is going to make a difference, I prefer to use buying points to smooth out volatility by turning potential pushes and losses into wins. I don't go crazy and buy 10 points. I usually buy 1/2 or 1, sometimes 2.Comment -
Actually, I'd argue the two are almost exactly the same. Unless you're using computer software when playing BJ or have access to Rain Man, you're most likely using one of several card counting systems. These systems only provide estimates of the expected true deck probabilities, which in certain cases may be off substantially.
The same dynamic exists with sports betting. We don't know the exact probabilities of a given outcome occurring, but nevertheless, through a combination of quantitative and qualitative analysis we can still determine (hopefully) unbiased estimates of event win/push probabilities.
Compare the probabilities implied by whatever counting system you use with the true prior probabilities over a few hundred million deck runs, and what you'll probably see are that your system's predictive figures possess a similar degree of accuracy as that which might be expected of an advantage sports bettor.
I *suspect* (and I may well be wrong) that you don't quite understand how vig factors into a bettor's decision calculus (Again, you might want to check out An introduction to expectations and theoretical hold, which is the follow-up to An introduction to betting lines and percentages.)
The majority of what you wrote (up until you mention vig), "To place a bet I have to pay vig, the only way I turn that into an advantage is if I can handicap the game so that the actual probability of outcome is different than the probability of outcome implicit in the odds I am getting. I would then only bet in cases where I thought the true probability was greater than the quoted odds." is entirely correct on its own. However, the final phrase of you statement. where you continue with "that the vig could be overcome" belies the statement's accuracy.
Lets look at two ways to consider vig/expectation in the context of a market of, say, +2½ -110/-2½ -110, where your handicapping leads you to believe that the underdog has a 53% probability of covering the spread:- Because 53% > 110/(100+110) ≈ 52.381%, a positive expectancy bet exists. This implies edge of 53% * 100/110 - 47% ≈ 1.182%.
You can just then stop there ... vig has already been calculated by looking at the -110 money line of the dog spread.
- The vig on the -110/-110 market would 50% * 100/110 - 50% ≈ -4.545%.
If we consider the market midpoint of +100 (implying zero vig), the edge would be 53% * 100/100 - 47% = 6.000%.
This means that the vigged line will have an expectation of = 6% + (1 + 6%)*-4.545% ≈ 1.182%, which is the same as we found above.
So anyway, there you have it, method 1 allows you to calculate expectation without explicitly mentioning vig, while method 2 allows you to calculate expectation by specifically applying the vig to the midpoint expectation. Either method will always produce identical results. Just don't make the mistake of trying to somehow do both, which is what you seem to be implying when you continue with, "I would then only bet in cases where I thought the true probability was greater than the quoted odds by a large enough margin that the vig could be overcome."
That's simply not how an advantage player ought to be thinking.
When ever making a purchase of, well, anything really, there is always one important consideration that must always be taken into account (even after shopping around and finding the lowest possible price), specifically, is the price being paid less than the value being received?
So right off the bat this suggests a potential difficulty with your methodology. Sure, key NFL numbers do tend to push fairly frequently, but what an advantage bettor needs to ask himself, prior to making his determination as to whether or not to purchase ½ or more points, is whether or not the value he's receiving from picking up the extra numbers is greater than the the price he needs to pay to buy those points.
Let's take the previous example of ±2½ -110, where the underdog has a 53% probability of covering.
If you can buy up to the +3½ -170 would that be worth it?
Using the half-point calculator, we see that the ±2½ -110 market corresponds to +3½ -167.66. Entering a line of -170 on the dog side would raise vig by 0.490% to 5.036%.
Since the initial bet wins with probability 53%, we enter 53% into the Dog odds box, and 1-53%=46% into the Fave Odds box. (This implies a fair line of ± 100 * 53% / (1-53%) ≈ ±112.766, which the calculator automatically converts the percentages into.)
So we see that a line of +3½ -170 corresponds to a total expected loss of -0.270%. So in other words, by purchasing the two ½ points we've turned a money-making bet into a money-losing bet.
It's just this sort of analysis that an advantage player needs to perform each and every time he considers buying or selling one or more ½ points.
2% vig is huge. Gigantic.
With its -104 lines, Pinnacle charges "only" 1.92% vig. The vast majority of bettors still can't beat that.
2% is the approximate difference in vig between a Matchbook market of -113/+107 and a CRIS market of -120/+100. All else equal, an advantage bettor should never transact at the CRIS market, even though it represents only 2% additional vig.
An advantage bettor needs to always be considering the impact on his expectations of any change in the terms of a bet. A Kelly bettor would certainly be willing to accept slightly lower expected value in exchange for shorter odds, but only after precisely calculating the exact loss in expected value, and then noting whether the lower EV taken in conjunction with the shorter odds will actually increase expected bankroll growth.Comment - Because 53% > 110/(100+110) ≈ 52.381%, a positive expectancy bet exists. This implies edge of 53% * 100/110 - 47% ≈ 1.182%.
-
You should write a book Ganchrow or teach a class.
Comment
