Leaderboard: August 29th, 2013, 5:20pm: penda: 1287 played, 769 wins, 518 losses: Win ratio of 59.75% Heretiik: 1308 played, 784 wins, 524 losses: Win ratio of 59.93% Tedium: 633 played, 362 wins, 271 losses: Win ratio of 57.18% osomoranno: 691 played, 390 wins, 301 losses: Win ratio of 56.43% Vakaz: 731 played, 446 wins, 285 losses: Win ratio of 61.01% Mindful: 579 played, 342 wins, 237 losses: Win ratio of 59.06% Lance: 937 played, 490 wins, 447 losses: Win ratio of 52.29% Zureiya: 375 played, 233 wins, 142 losses: Win ratio of 62.13% Attog: 725 played, 427 wins, 298 losses: Win ratio of 58.89% Nem-x: 269 played, 158 wins, 111 losses: Win ratio of 58.74% Average percentage of games won in top 10: 58.54% What does this mean? 1) The matcher probably does a pretty good job of matching you to “even” opponents, such that there is no player climbing to the top through a series of landslide wins against misranked players. 1a) Or it could mean that all opponents count as “even” opponents. If the game is mostly random (80% random, 10% skill, 10% gear), then a player who maxes out on their skill and gear category would only win 60% of their games. With Zureiya winning more than that (62%), we can tell that the combination of skill and gear represents at least ~21% of the game as opposed to luck. However, since Zureiya is unlikely to have won 100% of the games that hinged on an element of skill or luck (no offense), that means it is likely that more than 21% of the game is skill and gear based. My non-mathematical guess is that this number could be in the 30-50% range. Finding a few players with large numbers of wins and larger percentages won could indicate a stronger element of skill and gear...or it could represent the matcher not working as well as we previously thought. 2) While determining the ratio of luck to skill/gear is more difficult than I know how to do, we can assess from these results that the game is not entirely random. There is definitely an element of skill/gear at play. 3) We could determine the amount of skill in the game by running a tournament week where everyone plays with the same characters and the same equipment set ups, totally fixed. If players won a higher percentage of wins on a statistically significant number of games, that would indicate the amount to which skill influences the game and the amount to which luck influences the game. If, for example, in a fixed gear tournament, the win/loss ratios values ran from 70%/30% to 30%/70%, we would have reason to believe that the “skill” column represents about 40% of the game when opposed to luck. The reason this is approximate and not fixed is because even the most skilled (70%/30% player) may not be winning 100% of the match ups determined by luck rather than skill. When you add gear into the mix, sorting out the math gets more difficult. End note: Don't feel bad about losing when they block your 30 damage overwhelming assault on a 6 with an unreliably block then kill you with a 30 damage overwhelming assault after rolling a 1 on a parry
I can't seem to edit this, but the sentence that contains: "element of skill or luck (no offense)" should read "element of skill or gear (no offense)"
While I applaud the effort, how do you account for wins against the AI (which is not uncommon at extreme rankings)?
I think you're making some significant analytical errors. They have 60% win rates because thats what it takes for them to GET where they are (the top). Assuming that 1) they are not at the very top OR 2) they played for a lot longer than for a lot more matches. Their win rates will move towards 50%. Essentially, the extra 10% wins comes from the time when they are moving up the ranks. Assuming that there is a large enough population and good matchmaking, you win rate should be always be 50% after a sufficiently long time playing since you stop gaining ranks at some point and start winning 50% of your battles against people of equal skill/gear. With the current situation as it is, 60% win rates doesn't actually tell us much. In fact, assuming matchmaker works properly and there is no individual thats so much better than everyone else, long term 60% win rates are a abnormality. As it is, it exists because the top players don't have that big a pool of similarly skilled opponents to fight again (so imperfect matchmaking) and the short beta period prevents their initial gains from being entirely removed. Thats... not it. Totally not it. You're using some really weird logic here. If its 80% random and 20% skill/items. Then your win rate should still be 50%, because the computer wont matchmake you against people you have a 20% advantage over. this stat in no way imply what you end up assuming. I don't even know how you took that leap. The logic you're using is so weird i don't even know how to accurately articulate where you went wrong. So I will just use an exmaple: chess is close to 100% skill based. No luck. If you try to use the same sort of logic you did for Card Hunter, you will be expecting insanely skewed win rates. That sort of win rate doesn't exist. Yes to the former, no to the latter. Again, the AI matchmakes. If you have a gear + skill rating of 100 (giving a value to an normally difficult to quantify variable for illustration), you are not going to be fighting people with a gear + skill rating of 10. 2 people with the same gear + skill rating will end up winning 50% each. The more statistically significant your number of games are, the closer to 50% it will be. The only way to get an idea of how it goes is to remove both AI matchmake AND gear restrictions. Just having the same gear isn't enough, since some player play certain styles much better. You cannot be certain that the fixed gear options for a 2 warrior group is not better than for a 2 warrior + wizard group. Essentially: With a decent AI matchmaker (and its not that easy to mess up coding one), win rates don't matter and doesn't tell you anything.
@Blindsight: I don't, with regards to wins against the AI, but I imagine they would fall into the "wins based on skill" category. @Jotun: Interesting. I'm not sure I entirely disagree with you, but I think we're working from different premises. For example, you say: "Their win rates will move towards 50%." This seems to imply that you think that the high end players are generally equal in the skill and gear factor (you also say "there is no individual thats so much better than everyone else,". (Or that the game is 100% luck). I believe in the strength of the matchmaker, but there has to be players that are better and worse, even marginally so? When you move to your chess example, you say "Those win rates don't exist." I imagine you mean in both of chess and card hunter, but I think with chess those win rates could exist. If Bobby Fischer was hanging around Chess.com, even with a strong matchmaker, I suspect his win/loss ratio would be extreme. In card hunter, an element of luck mitigates the win/loss ratio. Determining how much of the win/loss ratio is mitigated by luck, and how much is mitigated by a matchmaker doing a very good job pairing people is difficult to say --- but interesting to think about. You'd also be fair to point out that with the number of registered users we have, we might not have a Bobby Fischer --- but surely there are some players that are better and worse? I think the major reason we have different ideas here is that you seem to believe the matchmaker is so strong that it will match people evenly all the time --- but I believe there is enough subtle variance in human play and gearing that this is not the case, and wouldn't be the case with 100000 players with 100000 matches each. Some people are just going to be better. Some of them by a good margin.
Yes, him. Singular. And that win/loss ratio will drop back down to 50% when other grandmasters start playing. For all intends and purposes, focusing on an abnormality that affects the top 1-10 players makes no sense. No it doesn't. It has no impact whatsoever. luck is 50-50. its net effect on ratio is zero. I will honest here, you don't seem to understand how statistics/elo ratings and probabilities work. You points aren't just wrong, they literally make no sense. You assumptions (like the prior mentioned luck on win ratio) are nonsensical. Your talk about variations in human play and gear has no impact whatsoever with how a matchmaking system works. The 2 are entirely unrelated concepts. I frankly don't even know how to explain this. Matchmaking isn't some sort of smart code working out parings using algorithms, its a dumb system that literally CANNOT screw up given large number of games. If the basic underlying elo system is flawed, then matchmaking will be inefficient. That doesn't mean it won't eventually pair you up with the right people. The statement "your relative skill won't be reflected because of poor matchmaking system" makes no sense from a statistical, programming or logical sense. If there is enough players and you play enough games. You WILL be matched evenly. No if, no buts. Its mathematically impossible not to. The matchmaking system literally keeps on moving you till you get 50%. It will not stop till it happens, and it will happen sooner or latter (depending on efficiency). You can only gain extra win rates during your climb. The moment you reach equilibrium, you start achieving a 50-50 win rate. Then your prior bonus win rates matter less and less ans you play more and more. The only way to not reach equilibrium eventually is to continuously grow relative to other players and eventually become and maintain a skill level thats hight than literally everyone else the computer could pit you against. this is the statement i'm totally puzzled about. What do you even mean? the 2 are not related at all. do you mean a player could play at 1200 and 1400 on different days? then he would just be a 1300 player and his 1200 and 1400 days would cancel out. Give me one scenario in which "human variance" affects matchmaking/win rates. You can't just throw out words without explanation.
The skill needed to beat the AI is far below that needed to beat a relatively skilled human. Knowing one trick can get you a win on any of the AI. Throw a holy nimbus and walk into the middle of the enemy team. They waste all of their attacks, go beat them up, rinse, repeat. Aside from it not really being that skill based, you can also predictably fight the AI once per day (if playing at the right times) and achieve a much higher win ratio than you deserve. I also just want to say that I side with Jotun on most of what he has posted here thus far. You either are not being clear in your definitions, or seem to not understand how the system works. In order to have a >50% win ratio you would either need to have a a very small fight sample, a disparity in temporal matchmaking (ie. top two players will never match up because they don't play at the same times), fight the AI a lot, and/or be able to beat all of the other top players consistently -- which is very possible with a small population and disparity in temporal matchmaking.
As one of the people on this list I would like to add my thoughts. First, the limited amount of players has made it easier for me to climb the ladder. I shot up from 1400 to 1600 in a relatively short time because I was consistently playing (and beating) the same few players. Obviously it was a mixture of luck and skill that got me there. I have since turned off the AI option and am still on the list but a lower position due to a string of losses. Again, luck and skill were both involved. As far as win percentage, that will of course even itself out once there are more people of a similar skill level playing. Given a large enough player pool there is no way for a group of players to consistently be at 60%.
As a further point towards more players evening it out i was at a 92% win rate during the closed beta with something like 300 games played. Yes of course that was a lot of bot games since the player base was small i would estimate about 60-70% but still. Within the first few weeks of more people joining i was down to 80 and then later 70. The more people playing at your actual level present in the ladder the less screwed your win percentage will be its obvious.
I guess this is what's confusing me: "The moment you reach equilibrium, you start achieving a 50-50 win rate." Maybe because I play at off-peak times and there isn't enough people playing, but sometimes I get people ranked 100+ points above me, and sometimes I get people ranked 100- points below me. Some of these matches are incredible landslides --- some of the players near the top (but at equilibrium) are going to face a disproportionate number of people below them, unless the matchmaker sets them to fight the same few people all the time. I don't know...is anyone on the leaderboard with a large number of games played at a 50/50 ratio? I scrolled through a few pages and didn't find anything...is the answer that merely none of the people with 1000+ games have happened to reach equilibrium yet?
Part of this is that you play at off peak hours, but also that there is such a small population of players. I remember being matched up with someone (IIRC) over 500 points above me. When there are only 3 people playing, at the time, this can happen. Ranking systems just can't work to make balanced games with so few choices to match you up against. When we get out of beta and there are a lot more players this sort of thing will happen much less frequently.
