I have been using accuscore as a crutch when I need some more info on a certain game. How it started out was that if accucore say there is a 65% chance I took it but the things went south and I started taking 70% and up chances with accuscore and now it seems to be worse. Is there any thing or anyone that has looked at what percentage that accuscore predicts to be more accurate. Such as on a MLB game 75% actually hits less than a 65% prediction. I know it wouldnt make since in real life but accuscore is just simulating the games.

Post edited by: kuzniart, at: 2008/08/23 16:48

I dont really get that there has not been an answer to this question. I have looked at what accuscore reports for their predictions but what I am asking is that are their predictions that are better than other even though it may be higher a perfect example is yesterday the Angels were 72% favorites over the A's and they got beat so is there a magic number? Does it have to be 75% to be a valid prediction or some other percentage such as 65% or 68% that they typically g4et it right?

There are several ways to use the AccuScore simulations, and while going with the best percentage is one method, it is not a guarantee to win. Picking a game that has one team as a 75% chance of being a winner still leaves a 25% chance of that team losing.

Using two easy examples, I could roll a dice, and I would have an 83.33% chance that the dice would not land on 6.

I could also flip a coin, and I would have a 50% chance that the coin would not land on heads.

Obviously the better bet would be "the dice will not land on 6". However, that does not mean the dice will not land on 6. It just means that if the dice was rolled 10,000 times, it would land on 6 only 1667 times.

Translating that to the simulations, we are saying that a team winning 70% of simulations has won 7000 times out of the 10,000 simulations we ran. That does not mean that this team is a guarantee to win, just that they have a better chance than a team winning 60% of simulations.

A lot of our members find that a better method to use is to see the value the team has. You can take the game line and find out what the Vegas odds are (these numbers can be found on the Live Betting Page, under LW%). By comparing the Vegas line with the AccuScore simulations, you can see which picks will give you the most value.

As an example, tonight's COL @ SF game:

Colorado is winning 35.3% of simulations. San Francisco is winning 64.6% of simulations. However, the Vegas line (LW%) indicates that Vegas gives SF a 70.2% chance, and gives Colorado a 29.2% chance. This shows that AccuScore likes Colorado more than Vegas does (this is called side value.)

So far this season, AccuScore's SV picks are up 4174 units (or 41.74 at one unit per game). That's always been the trend I've followed, and a lot of people have found it to be successful.

Of course you also have to put your own analysis to work. Take the Colorado game for example. Tim Lincecum is my favorite pitcher, so there's no way I'm betting against him, especially at home against a Colorado team that has struck out the eighth most in the majors. But that's just my personal opinion.

I know what 7 out of 10 is but what you said gave me something that might make what I am trying to get to clearer. What I thought accuscore did was take the stats from each player and whatever the situation of the game might be such as weather and come up with a percentage if I am wrong then I guess there would be no reason to use accuscore really because like you said a good team that is playing well is going to beat a bad team that is not playing so hot I dont need anyone to tell me I would be wrong to take a good team and I definitely dont need they to tell me that 10000 times what would be the point of that? If all accuscore is really doing is taking team A against team B and then comparing them 10000 times and comparing the outcome to the odds is that really necessary? Why couldnt you just do it once or twice? would that be more or less accurate I wouldnt think so. I am really not trying to bust anyones balls but dam man I have been using accuscore for awhile and thats why I was asking is there some kind of trend but if all you guys are really doing is flipping a quarter 10000 times and then putting the results against the odds I feel like I have been really putting trust in the wrong thing.

Cuisinart -

Perhaps a bit from the article will help you understand what accuscore really does for a simulation engine:

<snip>
AccuScore forecasts the future of sports, just like the weather forecast predicts the weather. We use a highly complex simulation engine run off of a bank of super computers and then we analyze the results. The simulation engine was originally developed to simulate human evolution and no, I'm not making that up. The engine would simulate from scratch, 10,000 times. By analyzing the data, the probability of any event could be calculated such as genetic mutation, tribal migration, you name it. Complex computer simulation is used everywhere from predicting the weather, the stock market and it is even used to predict government coups. Now AccuScore has brought that level of technology and analysis to the world of sports.

If you're anything like me, your brain is starting to cramp up like those headaches you get when you eat ice cream. You aren't alone. Let's face it, we're geeks. We like sports and math. In fact, I only missed one question on the math part of my SAT and it was a trick question, not that I'm still bitter about that or anything.

So how does it work? Well, for simplicity sake, AccuScore creates virtual players using over 30 different attributes then we put those players on a team which also has attributes such as coaching tendencies, then we let them play a game, one play at a time. We record every possible statistic then repeat the process, 10,000 times. The AccuScore simulation engine even factors in things like wind, rain, snow, field type, injuries. You name it, it's taken into account.

Don't you think the game would turn out differently if it's outdoors and snowing in Green Bay compared to cold but not snowing? So we monitor the weather report, the injury wire and every expert we can find for any minute detail that could affect the outcome of the game, then we re-run our engine.

As you can imagine, this process spits out a whole lot of data so then, in a combination of computer programs and human eyeballs, AccuScore analyzes this data and presents the results to you.

There are plenty of experts who predict outcomes, do you really need a computer too? While AccuScore will never replace the human expert, we just present a different way of looking at the game because the computer has no opinion. No angle. For example, a human expert might say that a player is going to have a huge game this weekend because he's playing against his former team. That makes for a good news story but the truth is, while it happens sometimes, it doesn't happen as often as you think; and many times it doesn't happen at all. Did you see Terrell Owens' stats as a Cowboy the first time he went up against the Eagles? Nothing to write home about, but boy was it the top story the entire week before the game.

The other reason is that we don't give history any value. While we use historical data as one part of creating our virtual players, we don't use historical data to simulate the future like other companies. A great example of this was Jamal Lewis a few years back. Lewis had two games that season that were beyond words. He set the record for individual rushing yards in a game vs. Cleveland, nearly 300, and in his other match-up vs. Cleveland tht year, he ran for well over 200 yards. However; those games were against a horribly weak Cleveland rush defense. While Jamal Lewis was great that year, those two games skewed his average off the charts. His rushing yards average was iver 150 yards per game but did that mean that Jamal would rush for 150 yards every week? Of course not.

AccuScore doesn't care what a player has done in the past because every game is different, every opposing team is different. If the opposing team had an incredible running game, that would mean they would have more time of possession which would mean less carries for Jamal and therefore less yardage but Lewis had nothing to do with that as he doesn't play for the defense. Everything that happens in a game impacts every other part of the game. I am simplifying things but I think you can see how forecasted data can be a valuable tool to help your fantasy team. It's just a different way of looking at the game.

So do we predict the future? Are we a crystal ball? Yes and no. Although often times our data is used to predict a single thing like who will win the game, our data is best used for determining probability. So while our data might show Peyton Manning as forecasted to throw for 287 yards this weekend, that is actually the halfway point in the data, also called the median. It means that of the 10,000 simulations, Manning threw for less than 287 yards about half the time, or in around 5,000 of the simulations and he threw for more than 287 yards the other 5,000.

This data can be helpful but the best usage would be to head over to AccuScore and check out some of our free fantasy football tools. In the Pro Player Analyzer for example, you can pick any stat and any amount. Maybe you want to know the odds of Peyton throwing for more than 300 yards. The Pro Player Analyzer will show you the probability. It might say Peyton has a 39.9% chance of throwing for more than 300 yards at home against the Saints in week one. He may also have a 20% chance of getting more than 30 fantasy points in that game.

I hope all this helps you better understand why AccuScore data is such a valuable tool to have at your disposal.

Just a few other minor tips. You will be seeing a lot of data laid out in charts each week. You may see something like Larry Johnson's forecasted rushing TDs for the week is listed as .82. How can you score 8 tenths of a touch down? You can't. You can read that as Larry Johnson has an 82% chance of scoring at least one touchdown, and that's really good odds but it also shows that Johnson has an 18% chance of not scoring a single touchdown. That's how probability works and you should stick to the highest probability you can.

The other thing to know is we often use forecasted fantasy points to evaluate players. These fantasy points are based on standard performance scoring and your league's scoring system may be different. Only use this information as a basis for comparison.


<snip>

Post edited by: jmanasse, at: 2008/08/29 04:51

I hope the article Jason posted above helps.

I should also note that I posted the "quarter vs dice" example to illustrate how something with 50% odds or 83% odds of winning is not always a guarantee.

Rest assured, the simulation process is much, much, much more complicated than flipping a coin 10,000 times.