It's been quite a few years since the OLG started using even numbers for spreads in the NFL.
The biggest effect of this was to allow the OLG to set more accurate lines around key numbers. In most cases the difference is negligible - if the correct line is 8.0 then the being forced to use a spread of 7.5 or 8.5 is not in the end going to change anything. However if the correct line is 3.0 then being forced to 2.5 or 3.5 will be enough to leave edge players with an +ev spot on those games.
But there is also in effect a more subtle benefit for the OLG. Reducing the number of games in a parlay also changes the ev slightly in favour of the OLG. For a long time this was largely ignored but lately it seems a lot of players are concerned by this. With football and basketball I personally believe that the effect is too small to concern yourself with. With hockey and baseball because of the high percentage of games that end on 1 the effect may be more significant.
I'm going to try to explain the reasons why push lines are bad for the players and analyze the effects of them in this thread.
PLP
Pointspread - Push Lines
-
ProlinePlayer
- Site Admin
- Posts: 2076
- Joined: Sun Mar 31, 2013 6:33 am
Re: Pointspread - Push Lines
Why does a push favour the OLG?
Despite all the complaining about push lines on the forums, I don't think most players really understand why they hurt your net win chances. To understand this you first need to study the pointspread payout chart http://www.prolineplayer.com/Reports/Po ... hp#Ontario.
The key numbers you need to note here are the 'Per Game Odds'. Notice that in general they are higher as you add more games. for example a 4 team parlay has a per game payout of 1.78 while a 5 team parlay has a per game payout of 1.82. So when you bet a team parlay and one game pushes the damage isn't winning a 4-teamer instead of a 5-teamer. It's that the payout for those 4 teams has dropped from 1.82 per game to 1.78 per game.
Another way to look at is that you expected a payout of 20-1. With the push you receive a payout of 10-1. It's not the drop form 20 to 10 that is the problem. The thing is that had you received the payout you expected (1.82) then the 4-teamer should have paid 11-1. It is the drop from the 11 to 10 that is causing the damage.
But how bad is this? Probably not a big deal. Again using the 5-teamer as an example, in an NFL game the push probability probably averages around 5%. So that's if the other 4 teams win, then 1 time in 20 you take a 10% payout hit. Hurts, but hardly a big event.
Also some players have expressed concern that the ParlayMaker does not factor these pushes into the calculations. First with the addition of baseball and hockey pointspreads I am in the process of adding this to the program. I have not done this in the past (football/basketball) because I do not believe the effect is important enough. There is a downside to actually doing this - speed. For NFL pointspread lists with 4-5-7-8-9 team parlays the number of parlays returned can run into the thousands. And don't forget, that only represents a fraction of the total parlays analyzed. The program may actually have calculated the probabilities and expectancies of tens of thousands of parlays in creating that list. Now if every parlay is also recalculated for various possible outcomes the analysis time can increase to an unacceptable level.
But instead of just guessing at the effects of pushes we can instead try breaking down the parlay using various methods to see the actual effect of the push and to see what degree of inaccuracy this causes in the ParlayMaker program.
Despite all the complaining about push lines on the forums, I don't think most players really understand why they hurt your net win chances. To understand this you first need to study the pointspread payout chart http://www.prolineplayer.com/Reports/Po ... hp#Ontario.
The key numbers you need to note here are the 'Per Game Odds'. Notice that in general they are higher as you add more games. for example a 4 team parlay has a per game payout of 1.78 while a 5 team parlay has a per game payout of 1.82. So when you bet a team parlay and one game pushes the damage isn't winning a 4-teamer instead of a 5-teamer. It's that the payout for those 4 teams has dropped from 1.82 per game to 1.78 per game.
Another way to look at is that you expected a payout of 20-1. With the push you receive a payout of 10-1. It's not the drop form 20 to 10 that is the problem. The thing is that had you received the payout you expected (1.82) then the 4-teamer should have paid 11-1. It is the drop from the 11 to 10 that is causing the damage.
But how bad is this? Probably not a big deal. Again using the 5-teamer as an example, in an NFL game the push probability probably averages around 5%. So that's if the other 4 teams win, then 1 time in 20 you take a 10% payout hit. Hurts, but hardly a big event.
Also some players have expressed concern that the ParlayMaker does not factor these pushes into the calculations. First with the addition of baseball and hockey pointspreads I am in the process of adding this to the program. I have not done this in the past (football/basketball) because I do not believe the effect is important enough. There is a downside to actually doing this - speed. For NFL pointspread lists with 4-5-7-8-9 team parlays the number of parlays returned can run into the thousands. And don't forget, that only represents a fraction of the total parlays analyzed. The program may actually have calculated the probabilities and expectancies of tens of thousands of parlays in creating that list. Now if every parlay is also recalculated for various possible outcomes the analysis time can increase to an unacceptable level.
But instead of just guessing at the effects of pushes we can instead try breaking down the parlay using various methods to see the actual effect of the push and to see what degree of inaccuracy this causes in the ParlayMaker program.
-
ProlinePlayer
- Site Admin
- Posts: 2076
- Joined: Sun Mar 31, 2013 6:33 am
Re: Pointspread - Push Lines
So let's use some actual games and do an exact analysis of the probabilities and expected values.
I pulled the following game data off the ParlayMaker program yesterday afternoon. The probabilities are of course the same as those on the Ontario web page.
I'm going to use the following 5 games - Win% Edge 1H Baltimore +3.0 54.89 -2.30% 2H Carolina +2.5 53.04 -5.59% 7H Cincinnati +3.5 55.61 -1.01% 10H Miami +0.5 56.75 +1.01% 11H NyJets -0.5 56.67 +0.87% running the parlay list Parlaymaker gives us the following 5-teamer - Win% Edge 1H - 2H 7H -10H - 11H 5.21 +4%
first let's just verify the above
the win probability is calculated as 100 * .5489 * .5304 * .5561 * .5675 * .5667 = 5.2068
the edge is calculated as (20 * 5.2068) -100 = +4.1355
But we know that isn't going to be completely accurate because of game 1. We've assumed a win rate of 54.89 and a loss rate of 45.11 but have ignored the reality that this game could result in a push. Next post I'll do the same game but provide more accurate data.
I pulled the following game data off the ParlayMaker program yesterday afternoon. The probabilities are of course the same as those on the Ontario web page.
I'm going to use the following 5 games - Win% Edge 1H Baltimore +3.0 54.89 -2.30% 2H Carolina +2.5 53.04 -5.59% 7H Cincinnati +3.5 55.61 -1.01% 10H Miami +0.5 56.75 +1.01% 11H NyJets -0.5 56.67 +0.87% running the parlay list Parlaymaker gives us the following 5-teamer - Win% Edge 1H - 2H 7H -10H - 11H 5.21 +4%
first let's just verify the above
the win probability is calculated as 100 * .5489 * .5304 * .5561 * .5675 * .5667 = 5.2068
the edge is calculated as (20 * 5.2068) -100 = +4.1355
But we know that isn't going to be completely accurate because of game 1. We've assumed a win rate of 54.89 and a loss rate of 45.11 but have ignored the reality that this game could result in a push. Next post I'll do the same game but provide more accurate data.
-
ProlinePlayer
- Site Admin
- Posts: 2076
- Joined: Sun Mar 31, 2013 6:33 am
Re: Pointspread - Push Lines
So now let's calculate everything again but using more accurate numbers. The web site does actually work out all the numbers for each game as win/push/loss but for simplicity does a conversion to straight win/loss. However the file downloaded by the ParlayMaker program contains enough information to recreate the original values.
again ... Win% Edge 1H Baltimore +3.0 54.89 -2.30% 2H Carolina +2.5 53.04 -5.59% 7H Cincinnati +3.5 55.61 -1.01% 10H Miami +0.5 56.75 +1.01% 11H NyJets -0.5 56.67 +0.87% but this time for 1H we'll use the exact win probabilities - Baltimore Win 50.54% Push 7.93% Baltimore Loss 41.53%
Now we can use these values to work out the parlay's expected value correctly
Probability of winning all 5
100 * .5054 * .5304 * .5561 * .5675 * .5667 = 4.7941
Probability of 4 wins and a push
100 * .0793 * .5304 * .5561 * .5675 * .5667 = 0.7522
and now the edge would be
(20 * 4.7941) + (10 * 0.7522) -100 = +3.4040
So the edge as expected is somewhat lower than ParlayMaker predicted but the difference is less than 1%. Also note that for football this is an exceptional case with such a high push rate of almost 8%. Although a drop of +4.1% to +3.4% may seem significant I would expect the drop to remain the same for higher edges. Edges of over +20% are very common for NFL 5-teamers and a drop of 0.7 from something like +34% to to +33.3% is hardly noticeable.
PLP
again ... Win% Edge 1H Baltimore +3.0 54.89 -2.30% 2H Carolina +2.5 53.04 -5.59% 7H Cincinnati +3.5 55.61 -1.01% 10H Miami +0.5 56.75 +1.01% 11H NyJets -0.5 56.67 +0.87% but this time for 1H we'll use the exact win probabilities - Baltimore Win 50.54% Push 7.93% Baltimore Loss 41.53%
Now we can use these values to work out the parlay's expected value correctly
Probability of winning all 5
100 * .5054 * .5304 * .5561 * .5675 * .5667 = 4.7941
Probability of 4 wins and a push
100 * .0793 * .5304 * .5561 * .5675 * .5667 = 0.7522
and now the edge would be
(20 * 4.7941) + (10 * 0.7522) -100 = +3.4040
So the edge as expected is somewhat lower than ParlayMaker predicted but the difference is less than 1%. Also note that for football this is an exceptional case with such a high push rate of almost 8%. Although a drop of +4.1% to +3.4% may seem significant I would expect the drop to remain the same for higher edges. Edges of over +20% are very common for NFL 5-teamers and a drop of 0.7 from something like +34% to to +33.3% is hardly noticeable.
PLP
Re: Pointspread - Push Lines
Hi PLP
This is exactly the reason I was asking in a different thread if you knew how often NHL games end up exactly with a one goal win for the favorite. I can't seem to find the answer to this anywhere , am thinking it is in the 33% range , but would like the exact number if you are able to run it from your database....thanks
This is exactly the reason I was asking in a different thread if you knew how often NHL games end up exactly with a one goal win for the favorite. I can't seem to find the answer to this anywhere , am thinking it is in the 33% range , but would like the exact number if you are able to run it from your database....thanks
-
sharpasitgets
- Posts: 81
- Joined: Tue May 28, 2013 2:36 pm
Re: Pointspread - Push Lines
The probability of a favorite in hockey winning by exactly 1; is the delta between the probability of 1+ (i.e. money line) and win 2+ (i.e. the puck line).
Re: Pointspread - Push Lines
First time reading this thread. Push probability of the NFL 3 seems low at 7.93%. Should be about 2% higher.
PLP, are you using both fav(~10%) and dog(~6%) 3s and averaging them out?
PLP, are you using both fav(~10%) and dog(~6%) 3s and averaging them out?
-
ProlinePlayer
- Site Admin
- Posts: 2076
- Joined: Sun Mar 31, 2013 6:33 am
Re: Pointspread - Push Lines
Other factors are of course used to determine the push rates but why this game was so much lower I really don't know.BTS wrote:First time reading this thread. Push probability of the NFL 3 seems low at 7.93%. Should be about 2% higher.
PLP, are you using both fav(~10%) and dog(~6%) 3s and averaging them out?
You are right that I would normally expect a push rate of very close to 10% for a 3 point line.
PLP
Re: Pointspread - Push Lines
Yes it is. And that number is normally 25% and up, which is huge.sharpasitgets wrote:The probability of a favorite in hockey winning by exactly 1; is the delta between the probability of 1+ (i.e. money line) and win 2+ (i.e. the puck line).