By analyzing the Wizard's data from 2013-2017 in his 'NBA betting' article, I noticed a huge discrepancy in some of the probabilities for certain spreads over that time period with a decent sample size of on average 350 games. The ACTUAL probability shows around a 5 to a 8.5% edge over the implied probability over this data set depending on which spread you choose in the chart. Now with that being said, this is comparing the edge vs a FAIR LINE, which is almost never the case when a bookie puts out a line, but that doesn't mean there still isnt value, supposedly at least. For example, looking at the +5.5 point spread(the spread with the biggest edge) from 2013 to 2017, in a 336 game data sample, the fair line or implied probability would be 29.3% or +241. The Timberwolves the other day vs the lakers was +5.5 and +200 on the moneyline with an implied probability of 33.3%. Although the bookies line is worse than the fair line price, the ACTUAL probability for this spread from 2013 to 2017 according to the Wizard's chart shows a probability of 37.8%, so still a 4.5% edge (37.7 - 33.3 = 4.5). You even had the Bucks at the Warriors the other day at +5.5 with a +220 line at a 31.3% implied probability and using the Wizards Chart again, we get a 37.8% probability for the +5.5 spread momeyline bet, which means there is about a 6.5% edge on that game and what do you know, the Bucks went into Oakland and not only WON but BLEW OUT the best team in the league, which is very interesting.

My problem with this is the following. After really thinking about it more and what Im really doing as a whole regarding collecting historical data in sports, I started to ask myself some questions regarding if this is even valuable information to make a sports bet truly +EV. Backtesting historical game information even IF it has shown to be +EV over a 350 game sample size or whatever size for that matter, doesnt mean ANYTHING. Each game is independent regardless of what happened in the past, especially for the data set I illustrated above when it is based off random teams as a whole. I mean if sports are truly random as everyone in the world including me throughout my life always believed to be true, how can this type of information or ANY backtesting information REGARDLESS of the sample size ever be helpful in the LONG RUN? Each game is independent just like the spin on a roulette wheel. It is impossible for anyone to predict a single game when there are so many moving parts each and every game such as players feeling sick without telling anybody, things going on in their lives behind the scenes, added motivation because a reporter or someone said something to them, or maybe just not in the mood that game, angry with teammates in the locker room, etc. Even if you dont use any backtesting information and just try to analyze each and every game, you still have no idea what will actually happen on that single night. The only reason bookies make money is because theyre ripping you off on the payout both on the moneyline and the spread with equal action on both sides. That has no subjectivity of it like trying to predict a game. They are guaranteed to make money as long as there is equal action on both sides.

If sports are truly not random and -EV, that would simply ruin sports for me as a whole, one of the things I enjoy the most. Because if it's not random, that would mean it's rigged or bias in some way that could make historical information relevant and that would make me lose all interest if I found out it's basically a programmed event regardless if I can profit from it.

Quote:ZK...I mean if sports are truly random as everyone in the world including me throughout my life always believed to be true, how can this type of information or ANY backtesting information REGARDLESS of the sample size ever be helpful in the LONG RUN? Each game is independent just like the spin on a roulette wheel. It is impossible for anyone to predict a single game when there are so many moving parts each and every game such as players feeling sick without telling anybody, things going on in their lives behind the scenes, added motivation because a reporter or someone said something to them, or maybe just not in the mood that game, angry with teammates in the locker room, etc...

To me, this looks like you're saying sports are random (50/50 chance) like a roulette wheel. Something that can't be predicted.

Quote:ZK...The only reason bookies make money is because theyre [sic] ripping you off on the payout both on the moneyline and the spread with equal action on both sides...

Are you saying because there's equal action on both sides the casino has an advantage? Or in other words, the odds they offer, in it of themselves, isn't +EV to the casino, but only as a whole?

How about this -- if you truly believe sports are random (50/50 outcome), you can book my bets. I'll lay -105 on whatever bet I want to make, and since they are 50/50, you'll have a 2.5% advantage. Okay, that offer was a bit facetious since I don't think you'd take it (but if you do, LMK), but would you bet some trashcan team at +105 odds vs the Warriors (or whoever's the best team)? If it's 50/50, you'd have a 2.5% edge. There are bets all the time that are greater than +105, why not bet anything and everything that's longer than +100 odds, because it's 50/50?

Sports are not 100% predictable, as in, for anyone to say with absolute certainty X team has a Y% chance to win or a Z% chance to cover. I think part of the problem is you may be viewing 'prediction' as something which is certain, when a prediction is really just that but with odds attached.

Quote:ZK...If sports are truly not random and -EV, that would simply ruin sports for me as a whole, one of the things I enjoy the most. Because if it's not random, that would mean it's rigged or bias in some way that could make historical information relevant and that would make me lose all interest if I found out it's basically a programmed event regardless if I can profit from it...

Sports aren't random because there's almost always going to be a favorite. That doesn't mean it's rigged or biased, at least not in the way I think you're thinking about it.

Over two dozen teams of almost seventy data analysts given one real world set from the year

2014 could not develop a valid response.

Just because the data shows certain bets in the past were +EV doesn't mean that can't change for one reason or another.

it's not debatable. it's just false.

the great majority of those who do beat sports don't do it by back testing a system although they may use that info

the great majority who do beat sports are able to do so because their opinion or insight has greater validity than that of the general betting public whose opinion is reflected in the lines and the odds

if for some reason they had to bet every game probably no one could beat sports

but they don't have to do that

they cherrypick

It is indeed hard to predict how much Alabama will beat Vanderbilt by, but there is nothing fundamental about either the money line or the sports spread that makes a wager a -EV proposition. Sometimes the bookies will set the line or spread so that it is not 50/50 because they make a mistake, there is something they are not aware of, they place too much weight on certain factors, they intentionally skew the bet because they are trying to balance the action and the public has an emotional favorite, etc.

And as others have said, it is important to understand that in some/most games the spread and/or line is set very well and it is a -EV proposition to bet on those games. You can only win by limiting your bets to those games where the line/spread is not 50/50 -just like you only double in BJ on those hands where the EV is positive and you Wong out when the EV is highly negative.

Your mistake is assuming "100% perfect efficiency in setting the line and spread."

Quote:ZenKinGNow with that being said, this is comparing the edge vs a FAIR LINE, which is almost never the case when a bookie puts out a line, but that doesn't mean there still isnt value, supposedly at least. For example, looking at the +5.5 point spread(the spread with the biggest edge) from 2013 to 2017, in a 336 game data sample, the fair line or implied probability would be 29.3% or +241. The Timberwolves the other day vs the lakers was +5.5 and +200 on the moneyline with an implied probability of 33.3%.

You can see that after I smooth out the ups and down, my probability of a 5.5-point underdog winning is 0.293188. So, if you bet them at +200, you would an EV of -12.04%. Negative indeed.

Betting the NBA

Backtesting data sets in sports is also completely flawed because its not comparing apples to apples and theres way too many moving parts happening during that sampling that can lead you to making false conclusions. NO ONE knows whats going inisde these players heads or the coaches heads, or who might get injured, or who might sit out a lot of minutes after slightly pulling a muscle because the playoffs are coming up, all of which can skew any data analysis or whatever you do to handicap the match for that single night.

So back to favorites and underdogs. Sure, people know that each game there is a bias of who might win, but you're getting ripped off on the payout and in the long run it's impossible to know how many times a certain favorite can really win over that type of underdog even if the historic data set probability you put together show it to be +EV vs the implied probability of the line put out by the bookies. The reason being is like I mentioned above, these historical probabilities are eadily skewed with too many moving parts during that period that give a false conclusion that a data set might be outperforming, but thats false. We then cluster these probabilities and think a bet is +EV, but dont realize what actually happened game to game for that data set to outperform.

Bookies at the end of the day just want equal action on both sides and is why the line gets adjusted up and down during the day when theres too much action on one side vs the other as im sure all of you know already. Bookies set the spread and moneyline based on public perception that will get the most even action as possible and if it doesnt work out, they adjust the lines up or down during the day to guarantee the vig/profit.

The only way to make money in sports is if you're a bookie or if it's rigged and you have inside information on one of the best players on the team or the refs throwing the game or other inside information that no one knows about such as someone who will only play half the game cause theyre sick or whatever. Thats completely different and comes down to inside information. With that information you can make a killing and also use implied probability to escalate your winnings even more. Im not debating any of that. Im debating the fact it's -EV for anyone who tries to 'handicap' a particular match. In the long run, youll lose out because no one knows the future long term probability of a single random event that changes game to game over time. Handicapping a match is not a linear progression. Sports arent linear. It is always changing and way too many moving parts game to game.

Quote:WizardYou can see that after I smooth out the ups and down, my probability of a 5.5-point underdog winning is 0.293188. So, if you bet them at +200, you would an EV of -12.04%. Negative indeed.

Betting the NBA

The 29.3% wasnt the actual probability that was gathered during the data set. The actual probability over time showed a 37.8% win rate for 5.5 point dogs. So at +200(33.3%), it was a 4.5% edge.

None of this even matters though because I dont believe in backtesting subgroups of data in sports anymore. The more I thought about it, it just doesnt make sense to me, especially when these subgroups of data clustered together had so many moving parts throughout the sample size. Most backtests done in sports is just a bunch of clustered data that wasnt apples to apples. One cant then just group it all together and say it's +EV. I think this is where many sports bettors get in trouble. Even single game analysis game to game is -EV because no one really knows what will happen 'exactly' on that day regardless of historic stats and matchups.

Sports arent linear or scientific. Nothing but an independent trial with illusions of data sub groups over time that can falsely lead someone to believe there is a 'trend' or a correlation of data going one way. Maybe player's mentalities were different before the game begun or maybe they had different motivations before the game, maybe someone got hurt, maybe they had a feud with the coach for the past week and it affected their performance and the teams performance, maybe someone had a lingering injury, way too many moving parts for backtesting data in sports to ever have any signifcant value in my opinion. Single game analysis, same thing, completely random game to game and just another indepedent trial.