Virtual Indianapolis 500 Winner Announced

Thanks to an anonymous reader for pointing to this Codemasters press release announcing the winner of the virtual Indianapolis 500, ahead of the actual race this weekend. Apparently Sam Hornish Jr. surged ahead at the last minute to win this pre-race simulation, which was staged using Codemasters' newly released IndyCar game for Playstation 2. Is there any reason to hope that these sports sims (racing, baseball, football, soccer, and so on) have a real possibility of predicting the outcome of matches, or entire seasons?

Comments

Pole Position!

[Anonymous Coward]

First Post ;)
I hope I don't get filtered out by the TrendMicro virus scanning bug...

Stop with the pithy commentary

[Anonymous Coward]

I enjoy the new Games section on Slashdot but simoniker really needs to improve his posts and stop making those one sentence comments at the end of posts that are often blatantly obvious and quite frankly annoying and add nothing to the discussions. Here are some examples:

On Indy 500 game:
Is there any reason to hope that these sports sims (racing, baseball, football, soccer, and so on) have a real possibility of predicting the outcome of matches, or entire seasons?

On Matrix game:
More proof, if needed, that games are seriously big business.

On online games:
Opinions?

On Sega dropping prices for games:
Seems like good news for gamers.

Re:Stop with the pithy commentary

[Anonymous Coward]

Whiner. You probably wet the bed, too.

Re:Stop with the pithy commentary

[KDan]

Damn, I had mod points 3 days ago. +1 to the parent.

Daniel

Statistics

[Samus]

I used to work for a company that did sports simulations on a web site. Most of the engines were based on statistics. Back in March when the maddness hit I used the website to fill out some picks for a pool. I ran each possible game several times and then picked the one that consistently won. It came close and even managed to pick some upsets. If the other team would have won the last game I would have gotten 3rd place in the pool and a bit of cash. So while the current capability is not there if the statistical model is good enough you can get a good idea of what might happen.

Racing games are fun

[emo boy]

but I just need to find a steering wheel that actually stays connected to my desk. Can anyone suggest a steering wheel that works well with most any desktop?

Depends on how accurate the models become

[Hollinger]

And I don't mean the 3D meshes. I read in the hard-copy of Wired some time last year that in Hong Kong (I think) companies are spending a year developing software models of horse races, taking into account all sorts of little variables, and using them to figure out bets. They're so successful, they generally have to launder the money, and place bets through various 3rd parties in order to stay "in business."

I could see the same thing happening with off the shelf software, maybe, if you could make the software deal with stats over the span of a season (more or less) in realtime. e.g. Guy breaks his arm, software mirrors that in the virtual world, or Guy hits 40 homeruns half-way through a season, Virtual Guy more likely to hit homeruns.

They've been doing this for a few years...

[analog_line]

Sony has done it the last eight years with their Gameday football game, like the week before the Super Bowl. I believe they get the captains or each team (not sure of the selection process) to play each other. So far it's predicted the outcome eight out of eight times tried.

http://psx.gamezone.com/news/01_29_03_03_08PM.ht m

Re:They've been doing this for a few years...

[Albert Pussyjuice]

That's not a good predictor to use at all. Basically, you're taking event X, which is related to event Y in a few ways but in many other ways is quite different, to predict event Y. If they were using the demo mode to let Gameday play out the game itself, then it would be a more sound predictor because then the behavior of the two teams would be based on the statistical models represented in the game. As it is, the individual players who are representing their teams are not indicative of how well the team meshes as a cohesive unit. And the 8 out of 8 factoid is such a small number that it's insignificant. There are over 200 regular season games. If Sony were to let a player from each team simulate the game before it was played, I doubt that you'd get overe 50% accuracy.

Re:They've been doing this for a few years...

[analog_line]

Actually, it is a better predictor than mere statistical models. Odd things happen on the field. You can't statistically model a clutch player or someone who folds under pressure. Having an actual person playing the game makes it more real, frankly. A computer can't out-think itself. A computer can't improvise like a human player can. If statistical models were all that mattered, gambling would cease to exist within days, because everything could be modeled.

Of course it's not perfect, but 8 of 8, each one with a completely new game, is some pretty good anecdotal evidence as far as i'm concerned. Yeah, at some point it's not going to predict the winner. Nothing is perfect, but it's certainly interesting and worthy of taking a look at.

Re:They've been doing this for a few years...

[Albert Pussyjuice]

No, it's not a better predictor. Yes, it introduces the possibility of random chance but that randomness is not a reflection of the real game in any way whatsoever. There is a huge difference between video game football and real football and using statistical models of football games will be a much better predictor over the long run.

8 out of 8? 8 tries and you're terming it a success? Discrete math tells us that eight trials is not going to cut it for this type of event. And if you want to know why gambling remains as an industry it is because people love to take the odds - and most times lose. The favored team will, the majority of the time, win. Statistical models exist and are used all the time to determine odds.

Also, when you say "You can't statistically model a clutch player or someone who folds under pressure", I assume you are claiming that with this inclusion of human players that you are modeling this. This is not the case. First of all, the players aren't under nearly as much pressure in these games. There are people watching but there's not a stadium full screaming at them. The game is not being broadcast around the world. The clutch player you mention is not even neccescarily playing the game. Sure, the in-game representation is there, but you're negating the possibility that the game models a player's ability to step up in the big game. In short, only two players are involved in this game, which is much, much smaller than the actual number of players involved in the game and therefore this process only creates an element of randomness that is not stemming from reality because the two players are not interacting with the other X players not represented. Therefore, as I said, it's not a good predictor. You may wish to check out this book to get a better idea of why just including randomness, does not a good model make.

Game Leagues for Their Own Sake

[fuzzybunny]

I understand the allure of 'replicating' or 'predicting' real-world sports. That's been around since way before even our Mac II baseball games where we shared teams by diskette that we'd created with real players.

The predicting side of this is probably only of interest for people involved in sports betting and serious fanatics--but given the rise of game leagues, online competition has proven pretty well that it can stand on its own, hasn't it?

The first example of this that I'm aware of is the International Netrek League.

I don't see why not

[deacon_jay]

This monkey has a perfect record picking winners in the Stanley Cup Playoffs.
And I'm sure that sport simultations are more sophisticated than a monkey spinning a wheel, right?

Not for racing, but maybe for other sports

[Bravo_Two_Zero]

In other sports, past performance may be a better indicator of future performance, even with slightly different environment variables. There are exceptions (numerous, I'd think) like Tom Glavine's first outing as a Met in windy, 38-degree weather (no moisture on the hands for the nuanced friction of a typical release). Rain has a similar effect, but it's more obvious.

And you could say, in football, the Dolphins don't win in the cold, the Packers win in the snow and so on. There, even the changing environment could be figured into a simulation.

But racing is more impacted by microclimates, and those can hardly be predicted for weather forcasting. Particularly at Indy, where the cars don't even use brakes in normal operation (speed primarily drops as the cars drift in the corners or from the associated throttle lift), the simulation goes well beyond the car, track or driver.

And even there, besides the obvious mechanicals, the amount of crud on the track (oil, fuel, rubber, water, carbon fiber threads) and the different rubber compounds available for the tires, the driver is often the least predictable. A sim can say "Schumacher drives this way on this track," but who could have predicted the actions of the unpredictable, talented amateur Kimi Raikkonen? He had no base of data from which to draw, but he scored points as a rookie.

The longer races allow for an exponential increase in all of the above unknown factors. The microclime on turn three changes hour to hour. The new C-compound Pirellis leave more crap on the outside of the hairpin than expected. Billy Boat has a minor case of the flu that is tiring him more rapidly.

So, I'd have to allow that, given a large number of weather scenarios (temp, pressure, wind, sun/shade, precip) combined with weighted random factors for mechanicals and driver behavior, you might predict a portion of a race, like the standings after the first few laps. By feeding live data during the race, you might come closer to predicting the outcome as well.

But remember, if it's possible, a Formula 1 team would likely already be doing it.

No.

[mcSey921]

Is there any reason to hope that these sports sims (racing, baseball, football, soccer, and so on) have a real possibility of predicting the outcome of matches, or entire seasons?

No.

Doubtful....

[Monkeylaser]

The deciding factor in a lot of sports, cliche though it is to say, is a will to win. You can't quantify heart, with a statistical breakdown. You can't predict a quarterback is going to stay in a game with a broken leg, (as Donovan McNabb did), or when someone just decides to throw in the towel due to psychological reasons.

Using statistics will get a reasonable margin of error, but if we always knew a winner was foreordained in any sporting contest nobody would gamble on them anymore.

Also, it'd be tough to take into account the factor of coaching. Sometimes you'll get a coach like Buddy Ryan who will actually put out bounties on opposing players, hoping that a vicious defense is going to hit them right off the field. As long as there is a human at the helm of a sports team, it will be very difficult to predict what even the most talented squad of individuals can do.

Re:Doubtful....

[Matrix272]

I wish I had mod points... I'd mod you up.

Predicting sports by playing games may seem like a worthwhile and attainable goal, but in the end, it's impossible to 100% accurately predict the actions of human beings. If you do happen to predict who will win every race in a season, what's next? Predicting the place of every driver? If by some miracle you get that right, why not just start predicting everyone's life, from the moment of their birth, to the moment of their death? Theoretically, if you can 100% accurately predict human actions and reactions, and you can calculate all the other variables (like weather), you could predict everything that could ever possibly happen to someone throughout their life, and the possibility of it happening... using the same method as predicting races, you could predict, by way of using the highest probability, their successes, failures, and everything in between.

The logical conclusion is that you're trying the calculate the computing power of the human brain, using a relatively simple formula. I may be naive, but I think humans are too intelligent and too unpredictable to quantify their actions the way that you could quantify race results.

Re:No.

[Anonymous Coward]

Let me ask you a question. Do you think that spelling your email like this:

mcsey@[ ]oo.com ['yah' in gap]

dissuades spammers from sending you email? Why not just write: mcsey@yahoo.com ? I mean, if ANY spammers go to the trouble of spidering /., do you think that by not typing mcsey@yahoo.com right away you will make them not spam you?

mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com mcsey@yahoo.com

Oops, sorry 'bout that mate!

I have to disagree with the results

[Mostly Monkey]

The fact that "Sam Hornish Jr. surged ahead at the last minute to win this pre-race simulation" leads me to doubt that this would happen in real life. Computers might be great at computing averages, but when it comes to unexpected occurances they tend not to be as accurate. I would believe the results more if Sam Hornish had consistently led the pack.

Re:I have to disagree with the results

[notque]

I thought the exact same thing when I read it.

It basically came up with a simulation that is so highly unlikely that any hope that it is correct is insane.

I'm sure the reason this happened in the game has an internal setting that cars in back will natrually do slightly better than cars in front so that the game is competitive and not 500 laps of left turns.

totally

[chill182]

Video games can totally be used to predict events. Just the other day I played Battlefield 1942 and it turns out the Allies would win in a war against the Axis. Who would have predicted that?

I'm just waiting for...

[Enrico Pulatzo]

ESPN2 to start broadcasting Madden games. My dad even watched me play Madden 2002 for a few hours after I got it, and he HATES videogames.

Only In the Virtual World

[Dr. Wu]

Hmm,

The simulation must have been based on last year's cars. Since Sam Hornish is driving a Chevy, which are conservatively down 40-50 hp on the Honda and Toyota-powered entries, and considering that no Chevy qualfied within 5 miles of the pole speed, the only sure bet on Sam Hornish for this weekend's race is how long it take for him to be lapped.

But then again, it is virtual reality. As evidenced by the stands packed full of fans in the video game.

Dr. Wu

Madden has been accurate the past 5-7 years on SB

[whirred]

I read somewhere that EA has used Madden to predict the superbowl, and they were correct something like 6 out of 7 times.

If enough variables can be factored in, and this probably won't happen with consumer level games or software any time soon, I'd imagine that software could be developed capable of predicting a high percentage of games correctly. They will never be able to get 100% accuracy, but 90% is possible in our lifetime and enough to win some nice money (60% to 70% accuracy) shouldn't be too hard to achieve with a nice beowolf cluster.