Win Probability Function? (1 Viewer)

[eighsse2]

I don't know if or how other platforms do this, but on CBS, they will put a win probability % on each fantasy game, and it will change live as games go on. I tried to come up with a function to predict win probability, specifically for games that have not yet started (I really only tried to make the function mirror the results of CBS's own function, so not very scientific on my part). But I'd really like to come up with a function to mirror their live predictions too (you know, like if 4 of 8 players have already played on each side, there is less expected variability in the numbers versus the beginning of the game, not to mention one of the teams is then leading by some number of points).

Long story short, my function for beginning of game win probability, where a is the projected score for Team A, and b is the projected score for Team B, is as follows:
team A probability = a^3.5 / (a^3.5 + b^3.5)

Any mathemagicians out there with ideas for a good realtime mid-game function that takes 4 variables (each team's current score and each team's remaining projection)? Or even just a the type of mechanics the function should have, without necessarily fine-tuning the constants.

This is obviously a very inexact science no matter what. But something that is decent enough, you know?

 

[ignatiusjreilly]

I was told there'd be no math

 

[eighsse2]

I was told there'd be no math

Somebody lied. When we figure out who it was, they will be dealt with.

 

[Rubi]

It took me three years to pass algebra, I'll pass on this one

 

[Leroy Hoard]

FantasyPros has a my playbook function that gives each sides probability of winning a matchup and changes as the games progress.

From an NFL team perspective, most gambling sites have live betting odds that could be converted to % chance of winning that game, at least from the gamblers point of view.

 

[eighsse2]

FantasyPros has a my playbook function that gives each sides probability of winning a matchup and changes as the games progress.

From an NFL team perspective, most gambling sites have live betting odds that could be converted to % chance of winning that game, at least from the gamblers point of view.

Thanks for the tip, is that only for paying subscribers? I can access a couple elements of "my playbook" but not certain things without a subscription.

 

[Leroy Hoard]

FantasyPros has a my playbook function that gives each sides probability of winning a matchup and changes as the games progress.

From an NFL team perspective, most gambling sites have live betting odds that could be converted to % chance of winning that game, at least from the gamblers point of view.

Thanks for the tip, is that only for paying subscribers? I can access a couple elements of "my playbook" but not certain things without a subscription.

It's free. Have to go to "matchup" for each team.

 

[eighsse2]

FantasyPros has a my playbook function that gives each sides probability of winning a matchup and changes as the games progress.

From an NFL team perspective, most gambling sites have live betting odds that could be converted to % chance of winning that game, at least from the gamblers point of view.

Thanks for the tip, is that only for paying subscribers? I can access a couple elements of "my playbook" but not certain things without a subscription.

It's free. Have to go to "matchup" for each team.

Hey thanks! Never even paid attention to that tab.

 

[Chaka]

Is this an Onslaught thing?

 

[rockaction]

Is this an Onslaught thing?

Proxy Chaka?

I was too late to the FFA to really remember, but I can do that.

 

[Runkle]

I don't know if or how other platforms do this, but on CBS, they will put a win probability % on each fantasy game, and it will change live as games go on. I tried to come up with a function to predict win probability, specifically for games that have not yet started (I really only tried to make the function mirror the results of CBS's own function, so not very scientific on my part). But I'd really like to come up with a function to mirror their live predictions too (you know, like if 4 of 8 players have already played on each side, there is less expected variability in the numbers versus the beginning of the game, not to mention one of the teams is then leading by some number of points).

Long story short, my function for beginning of game win probability, where a is the projected score for Team A, and b is the projected score for Team B, is as follows:
team A probability = a^3.5 / (a^3.5 + b^3.5)

Any mathemagicians out there with ideas for a good realtime mid-game function that takes 4 variables (each team's current score and each team's remaining projection)? Or even just a the type of mechanics the function should have, without necessarily fine-tuning the constants.

This is obviously a very inexact science no matter what. But something that is decent enough, you know?

Seems doable in Excel. I calculate pre-game win probability by e^(-0.16*s)/(1+e^(-0.16*s)), where s is the point spread. I think I got that from the Wizard of Odds, using historical NBA data (I find NBA probability better for FF because NBA scoring frequently hits triple digits, and, scoring isn't in bunches of 3s and 7s in basketball, so it's a better fit to the range of FF scores).

I have an excel spreadsheet that takes in as input the projected points for every player on the rosters (current week plus forward-looking projections per week), finds the optimum lineup each week for each team, calculates the point spread, then figures out win probability for each game with the above formula.

Then, with a win probability for each game on the entire league schedule, it Monte-Carlo's the season 10,000 times and comes up with each team's odds of making the playoffs (as finishes are dependent on all the other teams' games, too, not just your own... hard to make up ground if the first place team has a cakewalk the rest of the way).


Anyway, adjusting the above for in-game stats should be rather easy, all you'd do is substitute the actual scores for the projected scores after the slate of games?

 

[eighsse2]

I don't know if or how other platforms do this, but on CBS, they will put a win probability % on each fantasy game, and it will change live as games go on. I tried to come up with a function to predict win probability, specifically for games that have not yet started (I really only tried to make the function mirror the results of CBS's own function, so not very scientific on my part). But I'd really like to come up with a function to mirror their live predictions too (you know, like if 4 of 8 players have already played on each side, there is less expected variability in the numbers versus the beginning of the game, not to mention one of the teams is then leading by some number of points).

Long story short, my function for beginning of game win probability, where a is the projected score for Team A, and b is the projected score for Team B, is as follows:
team A probability = a^3.5 / (a^3.5 + b^3.5)

Any mathemagicians out there with ideas for a good realtime mid-game function that takes 4 variables (each team's current score and each team's remaining projection)? Or even just a the type of mechanics the function should have, without necessarily fine-tuning the constants.

This is obviously a very inexact science no matter what. But something that is decent enough, you know?

Seems doable in Excel. I calculate pre-game win probability by e^(-0.16*s)/(1+e^(-0.16*s)), where s is the point spread. I think I got that from the Wizard of Odds, using historical NBA data (I find NBA probability better for FF because NBA scoring frequently hits triple digits, and, scoring isn't in bunches of 3s and 7s in basketball, so it's a better fit to the range of FF scores).

I have an excel spreadsheet that takes in as input the projected points for every player on the rosters (current week plus forward-looking projections per week), finds the optimum lineup each week for each team, calculates the point spread, then figures out win probability for each game with the above formula.

Then, with a win probability for each game on the entire league schedule, it Monte-Carlo's the season 10,000 times and comes up with each team's odds of making the playoffs (as finishes are dependent on all the other teams' games, too, not just your own... hard to make up ground if the first place team has a cakewalk the rest of the way).


Anyway, adjusting the above for in-game stats should be rather easy, all you'd do is substitute the actual scores for the projected scores after the slate of games?

This is awesome.

But 2 things.

1) Regarding your final sentence, wouldn't it make a big difference how much of the score was already set in stone, and how much was left to be determined? For instance, at the beginning of the game, if the spread is projected at 15, then maybe the odds are 63-37 (or whatever). But if there is only one playwr left, with only 10 minutes to go in his game, and the spread (projected final margin) is 15, the odds are probably easily beyond 95-10. Huge difference as the fantasy match goes along. Unless I'm misunderstanding.

2) Really love the season projection thing, I wanted to do something like that for projecting draft picks. But are you saying even at the beginning of the season, you would get individual weekly projections for every week of the season? Where would you find that? (Obviously, the further out the projections are, the less reliable they're going to be, but even just working projections would be nice.)

 

[JohnnyU]

MFL does this

 

[Runkle]

I don't know if or how other platforms do this, but on CBS, they will put a win probability % on each fantasy game, and it will change live as games go on. I tried to come up with a function to predict win probability, specifically for games that have not yet started (I really only tried to make the function mirror the results of CBS's own function, so not very scientific on my part). But I'd really like to come up with a function to mirror their live predictions too (you know, like if 4 of 8 players have already played on each side, there is less expected variability in the numbers versus the beginning of the game, not to mention one of the teams is then leading by some number of points).

Long story short, my function for beginning of game win probability, where a is the projected score for Team A, and b is the projected score for Team B, is as follows:
team A probability = a^3.5 / (a^3.5 + b^3.5)

Any mathemagicians out there with ideas for a good realtime mid-game function that takes 4 variables (each team's current score and each team's remaining projection)? Or even just a the type of mechanics the function should have, without necessarily fine-tuning the constants.

This is obviously a very inexact science no matter what. But something that is decent enough, you know?

Seems doable in Excel. I calculate pre-game win probability by e^(-0.16*s)/(1+e^(-0.16*s)), where s is the point spread. I think I got that from the Wizard of Odds, using historical NBA data (I find NBA probability better for FF because NBA scoring frequently hits triple digits, and, scoring isn't in bunches of 3s and 7s in basketball, so it's a better fit to the range of FF scores).

I have an excel spreadsheet that takes in as input the projected points for every player on the rosters (current week plus forward-looking projections per week), finds the optimum lineup each week for each team, calculates the point spread, then figures out win probability for each game with the above formula.

Then, with a win probability for each game on the entire league schedule, it Monte-Carlo's the season 10,000 times and comes up with each team's odds of making the playoffs (as finishes are dependent on all the other teams' games, too, not just your own... hard to make up ground if the first place team has a cakewalk the rest of the way).


Anyway, adjusting the above for in-game stats should be rather easy, all you'd do is substitute the actual scores for the projected scores after the slate of games?

This is awesome.

But 2 things.

1) Regarding your final sentence, wouldn't it make a big difference how much of the score was already set in stone, and how much was left to be determined? For instance, at the beginning of the game, if the spread is projected at 15, then maybe the odds are 63-37 (or whatever). But if there is only one playwr left, with only 10 minutes to go in his game, and the spread (projected final margin) is 15, the odds are probably easily beyond 95-10. Huge difference as the fantasy match goes along. Unless I'm misunderstanding.

2) Really love the season projection thing, I wanted to do something like that for projecting draft picks. But are you saying even at the beginning of the season, you would get individual weekly projections for every week of the season? Where would you find that? (Obviously, the further out the projections are, the less reliable they're going to be, but even just working projections would be nice.)

For #1, if you want to get that deep into it, you're going to have to come up with a much more complicated function, sure. You'll need to somehow get expected points for the remainder of the game based on historical score/down/distance/time data that goes beyond what a layman can do, probably would need those advanced AWS stats or whatever.

As far as #2, a few platforms provide full-season projections. You can look them up on CBS for example. Go to your league's URL plus /stats/stats-main and there's a drop down for each week's future projections and a summary 'weekly scoring' selection that'll give you the projected points for each player for the rest of the year. There's a gray hover box on the bottom of the screen to export them all to excel too.

 

[eighsse2]

I don't know if or how other platforms do this, but on CBS, they will put a win probability % on each fantasy game, and it will change live as games go on. I tried to come up with a function to predict win probability, specifically for games that have not yet started (I really only tried to make the function mirror the results of CBS's own function, so not very scientific on my part). But I'd really like to come up with a function to mirror their live predictions too (you know, like if 4 of 8 players have already played on each side, there is less expected variability in the numbers versus the beginning of the game, not to mention one of the teams is then leading by some number of points).

Long story short, my function for beginning of game win probability, where a is the projected score for Team A, and b is the projected score for Team B, is as follows:
team A probability = a^3.5 / (a^3.5 + b^3.5)

Any mathemagicians out there with ideas for a good realtime mid-game function that takes 4 variables (each team's current score and each team's remaining projection)? Or even just a the type of mechanics the function should have, without necessarily fine-tuning the constants.

This is obviously a very inexact science no matter what. But something that is decent enough, you know?

Seems doable in Excel. I calculate pre-game win probability by e^(-0.16*s)/(1+e^(-0.16*s)), where s is the point spread. I think I got that from the Wizard of Odds, using historical NBA data (I find NBA probability better for FF because NBA scoring frequently hits triple digits, and, scoring isn't in bunches of 3s and 7s in basketball, so it's a better fit to the range of FF scores).

I have an excel spreadsheet that takes in as input the projected points for every player on the rosters (current week plus forward-looking projections per week), finds the optimum lineup each week for each team, calculates the point spread, then figures out win probability for each game with the above formula.

Then, with a win probability for each game on the entire league schedule, it Monte-Carlo's the season 10,000 times and comes up with each team's odds of making the playoffs (as finishes are dependent on all the other teams' games, too, not just your own... hard to make up ground if the first place team has a cakewalk the rest of the way).


Anyway, adjusting the above for in-game stats should be rather easy, all you'd do is substitute the actual scores for the projected scores after the slate of games?

This is awesome.

But 2 things.

1) Regarding your final sentence, wouldn't it make a big difference how much of the score was already set in stone, and how much was left to be determined? For instance, at the beginning of the game, if the spread is projected at 15, then maybe the odds are 63-37 (or whatever). But if there is only one playwr left, with only 10 minutes to go in his game, and the spread (projected final margin) is 15, the odds are probably easily beyond 95-10. Huge difference as the fantasy match goes along. Unless I'm misunderstanding.

2) Really love the season projection thing, I wanted to do something like that for projecting draft picks. But are you saying even at the beginning of the season, you would get individual weekly projections for every week of the season? Where would you find that? (Obviously, the further out the projections are, the less reliable they're going to be, but even just working projections would be nice.)

For #1, if you want to get that deep into it, you're going to have to come up with a much more complicated function, sure. You'll need to somehow get expected points for the remainder of the game based on historical score/down/distance/time data that goes beyond what a layman can do, probably would need those advanced AWS stats or whatever.

As far as #2, a few platforms provide full-season projections. You can look them up on CBS for example. Go to your league's URL plus /stats/stats-main and there's a drop down for each week's future projections and a summary 'weekly scoring' selection that'll give you the projected points for each player for the rest of the year. There's a gray hover box on the bottom of the screen to export them all to excel too.

I think I made #1 sound a little too serious than I meant it too. I didn't mean anything situational, just based the amount of players left or "player minutes left". Just something that says, "okay, so the expected victory margin is 15. But wait, how much time is left? If the match hasn't even started, the odds should be closer, because there is so much variation that can go on in the match. But if there are only 2 players left to play, the odds should be much heavier for the favorite because the game is much more set in stone." Just like a 10-point lead in the first quarter of an NFL game means a lot less than a 10-point lead in the fourth quarter.

Anyway it might not matter, because I think the fantasypros odds might be more intelligent than the cbs ones, which is all I'm really looking for.

As for #2, thank you, this is very helpful. I guess I don't explore enough!

 

[zamboni]

Is this an Onslaught thing?

This guy says yes