regression towards the mean

[crjohnson32]

Are there any stats experts here that think regression towards the mean plays a role in handicapping?

I was having a conversation with a friend. I had mentioned that I was on the bears +9 and one reason I was feeling confident is Vick's 0-3 career record vs the bears.

He suggested that based on that 0-3 stat, regression towards the mean would point to Vick getting a W against the bears.

What do you guys think? If you got 2 evenly matched teams, team A is 4-0 against team B the last 4 years. Based strictly on that 4-0 record, and not taking into account other factors influencing your decision, would you be more inclined to lean towards team A or team B?

FYI, in this case we concluded Vick is a statistical anomaly and therefore regression towards the mean was not applicable.

[hutennis]

How do you know what is your mean?

But even if you do...

Lets say its 50/50
Like red and black on a roulette table (never mind zeros for now)

You see it in casino all the time.
10 blacks in a raw.
It surely must start getting back to the mean.
You start betting red on your conviction only to be raped by another 8 blacks.
Then its red, red, black, red and then another 11 blacks one after another.

All you got left is to cry "WTF IS GOING ON WITH THIS F***ING REGRESSION????!!!!"

Well, nothing actually. It is there alright. You are just dont understand and misusing the concept.

[Romanov]

quit gambling. there is no mean in your situation. what would the mean be for vick's career victories against the bears? that stat is useless. These teams are totally different than those teams in the past

[crjohnson32]

quit gambling. there is no mean in your situation.

wow....

I have never heard of "regression towards the mean" until my friend mentioned it to me, so it has never influenced a bet.

This is the handicapping think tank. I'm simply asking if regression towards the mean is relevant in sports betting.

Forget the situation I described above, I'm not asking about that particular situation.

[Romanov]

i beg your pardon. i was a little harsh

regression to the mean means that over an extended series of bets in our case, a bettor will tend to win his mean amount of bets. If you are a 50% capper you may go 40-10 over 50 plays but out of a large number of plays you will hit 50%

keep in mind that each bet or event has its own individual probability and does not influence the succeeding bets/events

[podonne]

To say it in a slightly different way, "regression to the mean" is the concept that over a large number of independent trials (games, coin flips, etc...), the "observed" mean will converge on the "true" mean. Look up the Law of Large Numbers on wikipedia for more info.

Your example has two evenly matched teams, but one has recently gone 4-0 against the other. I assume by evenly matched you mean that the result is 50-50, so a coin flip is a good analogy.

Let me ask the question back at you. If you got 2 evenly matched teams a fair coin, team A heads is 4-0 againstteam B tails the last 4 years flips. Based strictly on that 4-0 record, and not taking into account other factors influencing your decision, would you be more inclined to lean towards team A heads or team B tails?

Any other answer other then team A is going to beat team B 50% of the time in the fifth game is wrong. See also: The Gambler's Fallacy.

[ManBearPig]

Google Probability Theory and you will find the answers you seek.

[illfuuptn]

Jesus you're dumb op. If heads goes 40-0 on the first 40 flips then how many more "wins" should heads have over tails after 1,000,000 more flips? Ummmmmmm 40. 40-0=100%. 500,040-500,000=prettydamncloseto50%. The end. Leave.

[Inkwell77]

It's interesting to think about, but you have to take the spread into consideration.
Just because Vick is 0-4 straght up that does not really matter. If he was 0-4 ats it still doesn't really matter, but it means more than 0-4 SU.

With any type of regression towards the mean type of situation you have to know what the true mean is and in sports betting that is almost impossible and you will always have variance so it is extremely difficult.

Also, remember in a sport like football with such a small number of games each season this type of betting is very tough, especially with individual situations like the one you describe.

[crjohnson32]

Jesus you're dumb op. If heads goes 40-0 on the first 40 flips then how many more "wins" should heads have over tails after 1,000,000 more flips? Ummmmmmm 40. 40-0=100%. 500,040-500,000=prettydamncloseto50%. The end. Leave.

I can't imagine what kind of events in your life would lead to the lonely, paranoid state state of mind you would have to be in to write such a hostile response.

[Patrick McIrish]

Absolutely. I also bet teams that are "due".


Sincerely,

The General

[ManBearPig]

I can't imagine what kind of events in your life would lead to the lonely, paranoid state state of mind you would have to be in to write such a hostile response.

It's the culture...For example, I misread a post and responded with something irrelevant and instead of a civil response like thanks, but i was talking about X...I got called stupid, lazy, and a dumbass because I tried to help. It's hard to have honest open conversations these days around here without being called a name in some shape or form.

[babar1000]

It's the culture...I mis-read a posted and responded with something irrelevant and instead of a civil response like thanks, but i was talking about X...I got called stupid, lazy, and a dumbass because I tried to help. It's hard to have honest open conversations these days around here without being called a name in some shape or form.

[illfuuptn]

It's the culture...For example, I misread a post and responded with something irrelevant and instead of a civil response like thanks, but i was talking about X...I got called stupid, lazy, and a dumbass because I tried to help. It's hard to have honest open conversations these days around here without being called a name in some shape or form.

I can't imagine what kind of events in your life would lead to the lonely, paranoid state state of mind you would have to be in to write such a hostile response.

oh but you guys are huge pussies

[suicidekings]

Tough crowd in here

[crjohnson32]

oh but you guys are huge pussies

My heart goes out to you.
I can't imagine the abuse, whether physical, mental, or sexual, you must have endured during the impressionable years of your life that shaped the person you are today.

[illfuuptn]

My heart goes out to you.
I can't imagine the abuse, whether physical, mental, or sexual, you must have endured during the impressionable years of your life that shaped the person you are today.

I'm a college student who has had a pretty easy, fun life to this point. The only reason I direct anger towards people like you is to express my frustration with the useless masses of people from older generations, my own generation, and generations of the future. It most likely would be of great benefit to this planet and many of its inhabitants if you were killed. You just aren't smart enough to be worthwhile. Don't take it personal; you're in a rather large club.

[crjohnson32]

It most likely would be of great benefit to this planet and many of its inhabitants if you were killed. You just aren't smart enough to be worthwhile. Don't take it personal; you're in a rather large club.

Just so we're clear, in your lonely, paranoid, mind, actions worthy of death include:

joining a sports gambling discussion form and posting a thread in a section called "handicap think tank" asking if regression towards the mean has any relevance in sports betting?

what happened to you?

I'm beginning to suspect it was something far worse than I originally imagined.

[crjohnson32]

My God.... did you ever attend "the Second Mile’s" early intervention youth programs in Pennsylvania?

[statictheory]

I'm a college student who has had a pretty easy, fun life to this point. The only reason I direct anger towards people like you is to express my frustration with the useless masses of people from older generations, my own generation, and generations of the future. It most likely would be of great benefit to this planet and many of its inhabitants if you were killed. You just aren't smart enough to be worthwhile. Don't take it personal; you're in a rather large club.

omg, you just put yourself in your own category with that comment. see a therapist

[ManBearPig]

oh but you guys are huge pussies

case and point...grow up and educate yourself about how to not be a complete douche

[uva3021]

stochasticity can not be resolved by past observation

[solobass]

regression to the mean is not nearly as important as handicapping a game as an independent event/probability. also look at things on the team level and that will help. even if you do bet on the eagles your odds of winning your wager are about the same as the odds of vick actually finishing a game.

[AlwaysDrawing]

Over the next xx games, teams would be more likely to hit 50% than 0%, because in general, teams are likely to hit around 50%., and past performance does not indicate future results.

If a team is 0-10, they are likely to cover around 50% of the time in their next game. After a coin comes up heads 10 times, the next time it's still 50/50.

It's a standard gambler's fallacy that teams are due because of past poor performance, or numbers are due because they haven't come up in awhile. Likewise, teams that have done well against the spread recently, or the Bears beating Vick 3-0 or whatever, does not influence the next result. Those "trends" are usually artifacts, results of small sample sizes, or just the result of data-mining.

In short, you should neither gain nor lose confidence in a pick based on past ATS results. Those numbers don't matter, and using them in your analysis will help you as much as it would to pre-flip a coin 3 times so you would be "confident" in another tails because the coin came up Tails, Tails, Tails. Any confidence gained is illusory.

[LetsWin]

In short, you should neither gain nor lose confidence in a pick based on past ATS results. Those numbers don't matter, and using them in your analysis will help you as much as it would to pre-flip a coin 3 times so you would be "confident" in another tails because the coin came up Tails, Tails, Tails. Any confidence gained is illusory.

I believe this is why cappers who use past ATS results as their primary strategy of picking games have results similar to someone flipping a fair coin.

[Rourke]

If you got 2 evenly matched teams, team A is 4-0 against team B the last 4 years. Based strictly on that 4-0 record, and not taking into account other factors influencing your decision, would you be more inclined to lean towards team A or team B? FYI, in this case we concluded Vick is a statistical anomaly and therefore regression towards the mean was not applicable.

Only Podonne in this thread is even talking about the definition of "regression to the mean". In your case, it means that the true underlying probability of Team A beating Team B is less than 100%, even though they've beaten them 100% of the time recently. That's all it is: if a statistic looks like an outlier over a small sample size, it probably is and you should take it with a grain of salt.

FWIW, my models have shown virtually no advantage in using head-to-head in predictions over a general rating, so if I thought the market was taking the 100% into account it could be prudent to bet against that.

[Masu485]

I recently made a similar thread asking about regression here:

http://forum.sbrforum.com/handicapper-think-tank/1423011-does-regression-exist.html

The responses near the end were the most helpful, but I must admit I still don't fully understand.

[Inspirited]

god does not play dice with sports players

evaluate the player and pick winners

only cancel the winner if intuition is telling you that the line is wack

[Pokerjoe]

Absolutely. I also bet teams that are "due".


Sincerely,

The General

That sparked a real LOL from me. It isn't that funny, but it hit me right. NH.

[Pokerjoe]

The nature of the regression to the mean in sports betting is probably the biggest issue a handicapper faces.

Think of it this way: two kinds of RTM, causal, and dilutive.

Dilutive RTM is just law of large numbers. Meaning, a past aberration isn't affecting the future, rather it's impact on an increasing set is being diluted with future results. So you flip a coin 5 times and it comes up heads, heads is batting 100%, but you flip it 100 more times and expand the set of results, and the results are 50/50 in those 100, heads is now 52.38%. The small 5-heads set you started with isn't affected by future results, or affecting them, it's simply being diluted by inclusion with those future results.

More interesting to us is the possibility of causal RTM. That's the theory behind counting the deck in blackjack (the elimination of certain cards leaves a changed deck and thus changed expectation). IOW, the expected future results change because of information revealed by past results. Semantically, this is a more intelligent use of the phrase RTM, because "to regress" is a verb, an action, and law of large numbers RTM is a result, a description, but that doesn't matter, we're here to make money, not theory.

SB lies in between. Consider a league of 32 teams, each exactly equal. They play 5 games. You would expect after that 5 games to have a team at 5-0, and one at 0-5 for that matter, on luck alone. Okay, if you knew the teams were all equal, you wouldn't change your line. But suppose you were the only one who thought they were equal (because you were an awesome handicapper), and that the rest of the market was reacting to results. Then the market would make the 5-0 team a big fave over the 0-5 team, and you'd have a great bet on the dog.

Then again, maybe you aren't an awesome capper, maybe you're not paying attention, maybe the 5-0 team is 5-0 because it's really, really good.

So like I said, RTM is the biggest, constant, base, central issue in handicapping. Do you think "The Broncos are a changed and great team with Tebow, they've gone 6-1, they're underrated!" or do you think "The Broncos are the same team with Tebow, they've just won 6 of 7 coin flips, they're overrated!"

[allin1]

it would be nice to get those useless posts with insults deleted. the think tank is not the place for such a mess. besides that, great topic!

[donjuan]

Jesus you're dumb op. If heads goes 40-0 on the first 40 flips then how many more "wins" should heads have over tails after 1,000,000 more flips? Ummmmmmm 40. 40-0=100%. 500,040-500,000=prettydamncloseto50%. The end. Leave.

LOL u dumb.

[Wrecktangle]

The bashing in this sub-forum has gotten completely out of bounds.

FWIW, think of regression as multi-dimensional averaging, where the regression line is the "average value" for the range of data considered. Maybe this will help. If not, I'm certain the Hezbollah Stat Freaks will have fun at least.

Commence Bashing.