Percentages with Poisson

[roasthawg]

Haven't messed with Poisson in awhile... just need a quick refresher. Is Poisson the correct distribution to use to derive win percentage for the over/under in a baseball game where I have the actual O/U and my expected runs scored figure? Thanks!

[yoyoyo]

I'm not completely sure, but I don't think Poisson would be used with O/U.

I was under the impression that Poisson was used for things that are achieved one at a time. Free throws made, number of hits, number of strikeouts, etc. And not so much for things like RBI's, points scored, etc.

I don't know much about it, though, so I could be completely off.

[JVP3122]

I'm not completely sure, but I don't think Poisson would be used with O/U.

I was under the impression that Poisson was used for things that are achieved one at a time. Free throws made, number of hits, number of strikeouts, etc. And not so much for things like RBI's, points scored, etc.

I don't know much about it, though, so I could be completely off.

Correct. Also, it's generally used for rare events.

[roasthawg]

Runs are scored one at a time though in a baseball game, correct? Would love to hear some other ideas on how to calculate win probability given the actual o/u offered at sportsbooks and my expected o/u though if Poisson is not the tool to use here.

[yoyoyo]

Runs are scored one at a time though in a baseball game, correct?

Right, but they can be scored in bunches. 1 at bat can score more than 1 run.

[Tackleberry]

Runs are scored one at a time though in a baseball game, correct? Would love to hear some other ideas on how to calculate win probability given the actual o/u offered at sportsbooks and my expected o/u though if Poisson is not the tool to use here.

I am in the same boat.

Total runs in baseball does not fit the criteria for the poisson distribution to apply.

[roasthawg]

So the over/under of a game is set at 8... my calculations have the expected runs scored at 8.5. How do I translate this to win probability in order to plug it into Kelly? That's what I'm trying to figure out here.

[Rich Boy]

Just use the value of 0.5 runs, historically. Or you can use what Pinny or 5dimes has for the difference between 8 and 8.5

From my database, the difference between 8 and 8.5 is roughly 3.5%

According to Pinny, the difference between 8 and 8.5 is 3.8%

I wouldn't use poisson to calculate MLB game runs since it does not increase one at a time

[roasthawg]

Just use the value of 0.5 runs, historically. Or you can use what Pinny or 5dimes has for the difference between 8 and 8.5

From my database, the difference between 8 and 8.5 is roughly 3.5%

According to Pinny, the difference between 8 and 8.5 is 3.8%

I wouldn't use poisson to calculate MLB game runs since it does not increase one at a time

I'm hoping to arrive at some sort of an equation where I can plug in the numbers. In actuality my expected runs scored figure would rarely be 8.5 exactly... more like 8.7634522. I'm trying to plug that number and the actual o/u into an equation or formula that will give me win probability.

[Tackleberry]

Check out buby74's post in the thread I started, could prove helpful

[roasthawg]

Check out buby74's post in the thread I started, could prove helpful

Definitely an interesting post... I'm not trying to reinvent the wheel though as I already have a pretty accurate figure for expected runs scored. Just trying to figure out how to plug it into kelly accurately in terms of the win probability.

[Tackleberry]

Could you not use your figure for runs scored to come up with the runs per inning value, use that along with the info in the other post to derive the probabilities required to run a simulation and price any O/U line which would then give you what you need to determine your appropriate kelly stake?

[uva3021]

run a simulation, in excel use norminv and rand() as the probability and the total as the mean for both the listed total and your prediction