bacchuspup

As proposed, this poll will only be as good as the lists that people give you, and it won't take into account games left off the list. Um... I think any poll is only as good as the lists people provide. I mean, seriously, garbage in, garbage out. In a perfect world, everyone would give me a list of every 2600 game they've ever played, in order from best to worst, and that would allow me to complete the most accurate poll possible. Look, I'm going to start with the top 10 lists provided on the other polling site. Those of you who understand the algorithm I'm using will also understand the limitations that these top 10 lists provide. I'm fairly convinced that these limitations won't, in the long run, affect the final results. While it's true that people listing their top 10 doesn't tell you what these people thought about games outside of their top 10, it does offer a basis for comparisons of some of the best/most popular games Atari 2600 has to offer. Hopefully, I'll be able to incorporate more complete lists, like the one Chris provided to me, as people view the results. But the fact remains that what this system does is compare each game against every other game, and even lists that exclude crappy games will contribute to the accuracy of the list. Let's see what my results are after I work on it this weekend. It may convince people to submit complete lists to me, either to bolster their favorite games or to lower the ranking of a game that is incorrectly high due to the limitations of the lists that people provide to me.

...

This is exactly why I don't think this will work. If I don't like Pitfall, for example, and leave it off my list, then it has no losses. If it doesn't even make my list, then it should have more losses than the games that did make my list, not fewer losses. And that's why I encourage everyone to give me a list of ALL the games they've played, and not just the ones they like. As I've said (repeatedly) the top 10 lists that currently exist are a starting point, and the comparisons they create are no less valid just because they don't include comparisons of games people didn't like. It is true that the more COMPLETE lists I get, which include ALL games played in order, not just favorites, the more accurate my list will be. My guess is that once I've gotten farther in entering the data, and listed them, people will come forward and say, "why is E.T. so high?" To which I will say, "if you didn't want E.T. so high, you should include it at the bottom of your list." At the end of the day, I think we're going to get something very accurate.

Thanks Chris for your contribution. I'm going to work on it this weekend. It doesn't matter how long someone's list is. Say one person puts E.T. number 1 on a list of 200 games. That means that E.T. would get one "mini-win" against each of 199 games. It doesn't mean that E.T. would get 199 points. It just means that person's comparison would count in 199 separate comparisons. If two other people contribute a list of three games, each Adventure, then Pitfall, then E.T. E.T. would lose both matchups against Adventure and Pitfall 1-2. No matter how long someone's list is, they're still really only making 1 comparison between E.T. and every other game on their list, and that 1 comparison bears no greater weight than anyone else's comparison. Maybe the best way to explain this is to pretend that you had a website that allowed you to make a hundred comparisons of certain games. One of the comparisons is E.T. vs Pitfall. One person gets on and goes through all hundred comparisons, and every time he sees E.T., he makes sure that he has E.T. winning. Two people get on and pick Pitfall, and don't fill out any of the rest of the comparisons. Well, Pitfall is going to get a win head-to-head against E.T., regardless of how many other comparisons the E.T. lover makes. Only if the E.T. lover manages to find a game where his one comparison will hold up (i.e. nobody else enters their vote for that game) will E.T. love get any advantage to his votes. The algorithm can have warped results if not enough people are comparing a given game. Say someone makes up ten non-existant games, and puts E.T. on the top of the list. Since nobody else would be including the non-existant games, E.T. would get ten points. However, you can adjust for this problem by requiring each game to have at least, say, 5 people include it in their list before you consider it. This is the kinda thing that will be easier to explain once I've worked on entering people's lists, and once more people participate.

Actually, Lucifer's point explains why I need a new poll. The other poll's list of everyone's top 10 is a decent starting point, but the fact that people are putting down only the top 10 could skew the results. He's right - let's say 10 people don't like E.T., so they don't include it in their top 10. One guy has an irrational love of the game, and puts it at #1. Because nobody else bothered to put the game in their list, E.T. would have an improperly high ranking. Another example - so far, I've entered 10 people's top 10 in my matrix. Xeophobe is one of the worst games so far at 0-10. But that's not really accurate - the person who put Xeophobe at #10 thought it was worse than 9 other games, but better than every other game in Atari's library, so Xeophobe should be winnng a lot more comparisons. That is WHY I keep saying that I need people to give me new lists, and that the longer these lists are (including all games played, from best to worst), the more accurate my poll is going to be. If those 10 people who didn't like E.T. expanded their top 10 list to include all games played, and put E.T. at the bottom, the fact that one person puts it at #1 wouldn't affect E.T.'s ranking (it would still lose all individual comparisons 1-10). That removes the flaw Lucifer recognized. Again, I encourage people to go to bestrollercoasterpoll.com and read what people say about the system. It's not entirely full-proof, but I still maintain it's more accurate than a popularity contest - especially so if people take it seriously and offer full lists of best-to-worst. I recognize it's harder than just putting in a top 10, but it's certainly harder for me to maintain the matrix than a more simple poll, and in the end, I think it will be worth people's time. Besides, it's gotta be kinda fun to figure out your whole collection, from best to worst. Maybe it's just me, but I love lists.

The algorithm is the same as the one used in bestrollercoasterpoll.com. To see their explanation of it, go to http://www.ushsho.com/howitwrk.htm. It essentially looks at how people view each game vs. each other game. When more people put Game "A" higher on a list than Game "B," Game A gets a win against Game B. Wins are compared against losses, and the game ranking is based on the highest win-loss percentage. That's why it's not a popularity contest. A thousand people may have played Pitfall, but if only twenty people have played both Pitfall and a homebrew called Moogle, and if 15 out of 20 people put Moogle higher on their list, Moogle will beat out Pitfall in the head-to-head. If Moogle and Pitfall beat out all the other games, Pitfall will still have one loss to Moogle's undefeated, so Moogle would be ranked higher. That reflects that people who have actually played both games prefer Moogle. You could never have Moogle win in a poll based on voting points. Let me make something clear - I am encouraging anyone and everyone to give me your best-of list. The longer the list, the more accurate the head-to-head wins and losses are going to be. And for those of you who already participated in the other poll, I can take that into account when you do my poll so that I only count your prior list once. It's going to be pretty cool once I get a few dozen lists into the matrix, and pretty accurate.

Hello again, Okay, I have my matrix set up, and I've tested it by entering the first five top 10 lists from the other poll. Each "win" for a game reflects each instance where one game has been ranked higher than another game more times than it's ranked lower (e.g. if 6 people rank Asteroids higher than Pitfall II, and 3 people rank Pitfall II higher, that's a win for Asteroids and a loss for Pitfall II). These are EXTREMELY preliminary. The more lists I get with the more games (and therefore the more comparisons), the more accurate this list is going to get. However, here's what I have so far. Game (Win-Loss) Asteroids (18- 0 ) DragonStomper (10- 0 ) Demon Attack (9- 0 ) Seaquest (12- 1 ) Pitfall II (18- 3 ) Adventure (16- 3 ) Keystone Kapers (16- 3 ) River Raid (22- 6 ) Super Breakout (11- 3 ) Pitfall (14- 4 ) H.E.R.O. (7- 4 ) Circus Atari (15- 9 ) Missle Command (9- 7 ) Frogger (10- 8 ) Bump n Jump (5- 4 ) Q*Bert (6- 5 ) Star Wars Arcade (6- 5 ) Combat (9- 9 ) Rescue Terra I (5- 6 ) Warlords (4- 5 ) Tapper (4- 7 ) Plaque Attack (3- 6 ) Air Sea Battle (3- 8 ) Frogger II (3- 8 ) Millipede (4- 11 ) Jr. Pac-Man (2- 7 ) Frankenstein's Monster (2- 9 ) Joust (2- 9 ) BMX Airmaster (1- 8 ) Boxing (1- 8 ) Ice Hockey (1- 8 ) Pac-Man (1- 10 ) Megamania (0- 9 ) Defender (0- 9 ) Jawbreaker (0- 9 ) Montezuma's Revenge (0- 10 ) Xenophobe (0- 10 ) Donkey Kong (0- 11 )

Hello all... I'm brand new to the forum, and I don't want to step on anyone's toes here, but I'd like to develop another list of the best games for Atari 2600. I know one's already going on, but I've got a somewhat different take on how to do this kind of list. The way the current top 100 list is being developed is by points for every game on someone's top 10 list. That's valid in some ways (and I give the guy who started it huge props), but it certainly rewards more popular games at the cost of actual quality. For example, Communist Mutants may be better than Pitfall (I'm not saying it is, just bear with me), and everyone who actually tries Communist Mutants would say it's better. But if you have 50 people putting Pitfall at number 5 on their list, and 10 putting CM at number 1, Pitfall comes out way ahead (250 to 100), and that's not really accurate. There is a way to fix this - if you go to Bestrollercoasterpoll.com, you'll see a methodology that removes popularity from the polling. It basically only gives an item points for "winning" against other items. So if you have five people who have tried CM and Pitfall, and all five people put CM ahead, then CM gets a win and Pitfall gets a loss, no matter how many people have tried only Pitfall. It's a bitch to put this kind of list together - it requires a matrix and some tedius excel work - but I'm willing to give it a shot. I just think it more accurately determines which games are "best," whereas the current poll determines which games are "most popular" - neither are more or less valid, but that's what they do. Here are the rules for my poll: 1) Make a list of as many games as you want, and put them in the order of "best of." It doesn't matter if you include homebrew games or supercharger games, and it doesn't matter how many games you want to include. The more the better. Just rank them from 1) to 20), or 30), or as many as you feel like listing - as long as they're in order of best to worst. 2) The only thing I ask is that you ONLY include games you've actually played, not that you've just heard about. That makes the system more accurate. I'm going to start the "best of" list by incorporating the lists people provided in the current poll. It's not going to get me where I want to go (because most people separated homebrews, superchargers, and originals), but it's a good starting point. I will keep an eye out for people who participate in both polls, so I can make sure I don't count anyone twice. I'll put my initial results up as soon as I can. I think it'll be interesting to see what differences there are in the two polls. Thanks all!

Hello all... I'm brand new to the forum, and I don't want to step on anyone's toes here, but I'd like to develop another top 100 list. The current poll (and I give the person doing it serious props) is really a popularity contest. The more popular the game, the more points it's going to get. But games that haven't been played by as many people simply aren't going to get as many points, even if they're better games. There is a way to fix this - if you go to Bestrollercoasterpoll.com, you'll see a methodology that removes popularity from the polling. I think it more accurately determines which games are "best," whereas the current poll determines which games are "most popular" - neither are more or less valid, but that's what they do. I'm going to start the poll in a separate thread; I hope some people will come check it out.