Hello again my avid (and mostly imaginary) readers, it is again time to talk about how to get the most bang-for-your-buck from the arena system. There has been some huh-bub in the forums over Kalgan's post regarding the Elo system (which, for those who are unfamiliar, is the queuing and rating requirement system to promote a competitive and fair environment in the arenas). Kalgan makes mention that they have a PhD in-shop with Blizzard's developers who "did his dissertation on competitive ranking systems" (source). The reason he brought this tidbit of information to light was because there is a new and shiny "exploit" that has started to appear, though Kalgan stands firmly behind Elo and claims that this new "exploit" is nothing more than a way to increase the number of wins necessary to get to rating plateaus (i.e. 1850, 2000, etc).

The exploit works like this: first, have a 1700 2s team. Sounds easy right? Have your team mate drop the team and rejoin, knocking his personal rating down to 1500. Next, give him ownership while you drop from the team and rejoin, knocking your personal rating to 1500. Here is where the interesting quirks of the Elo system take effect. Queue up! The theory here is that because your personal rating is more than 150 below the team rating, you will be matched as if your rating is the average of your two personal ratings (1500). Presumably, if you have a 1700 team rating, then you earned it, and therefore should be facing opponents in the 1700 skill/gear range, but if you are queued as if you were a 1500 team (as this case proves to be), then you would expect to face teams who are of a skill/gear level of 1500. This is where this particular "quirk" gets interesting.

Presumably, if everyone (but you) is playing legitimately, then you (as a truly 1700-skilled team) would be able to make short work of a 1500 team. So, you do this 10-15 times and have an amazingly high win percentage (something like 80-100) and your personal rating is 1700-1750 because you were getting about 15 points per win and 15 per loss; at an 80% win rate, you would net about 120-180 personal rating and your team rating should no longer be 150 above your personal rating because of your high winning percentage and the fact that your team rating was getting prorated for being much higher than the opponents you face. Now, you have another quirk to play with.

You have 1700 personal rating, and your team rating is up to 1800 (you made 200 rating, your team made 100... that's about right after a few losses and the steep prorate). Bummer, now you would be matched against an 1800 team because the lowest person in the queue would be within 150 personal rating of the team rating, so the team rating would be used in match-making (in practice, it would have happened before now, but I am just showing an example). Party's over right? Now you have to legitimately earn those last 50 team rating and 150 personal rating, right? You have completely screwed yourself... had you done it right from 1700, you would have hit 1850 in fewer matches... ... ... but only if you had won in the same fashion you have been winning in while facing scrub teams. That is the big "if" to the argument.

But wait... you did it before... why not do it again!? Except, this time only ONE of you is going to leave and rejoin so that the other can keep his personal rating. Alright, you drop and rejoin, now your personal rating is 1500 and your partner's is 1700 while your team has an 1800. How does that help? Well, right now the averaged personal rating of the members is more than 150 less than the team's. So, your team is now being matched via the Elo System's rule of "150 below means average the two queued members and match on that rating." The average of 1700 and 1500 is 1600, so you will be facing a 1600 rated team (and remember you originally got your team to 1700, so you are good enough to beat 1600 rated teams, most likely)(source). From here, it should be relatively straight forward to see the progression this takes. A team can reasonably easily push one member into the 1850PR/TR level to get their weapon as long as one member bites the bullet and gets fewer points and lower personal rating.

Problems with this system:
Well, for one, if this is becoming the way to power someone to an 1850 rating for the s3 weapon, then all the high-level weapon-sellers will be doing it too. That means that the 1500-1700s will be littered with s3 shoulders teamed with s1 teammates in hopes of getting their 1850 PR/TR for the weapon. If the aforementioned situation arises, then this method will not work because your team will be facing other teams who are doing the exact same thing (which means that your 1850 team, with an 1800 warrior, and a 1500 druid will be facing the exact same except it will be a 2000 team with an 1800 paladin in blues and greens, and a 1500 rogue in full s3).

Conclusion:
Really, this is not the worst situation in the world. I hypothesize that if you play at peak hours, then this is the best thing ever because the 1500 teams you face are more likely to actually be 1500 teams, whereas the off hours will be littered with these types of teams trying to power up against scrubs. Additionally, this is not even cheating, per se, because your are not abusing the system in underhanded ways, you are simply rejoining the team. Even Kalgan states that this is not a form of cheating as far as he is concerned because one would have to play (and win) more games to get to 1850 than you would doing it legitimately (source). What he fails to realize, however, is that he assumes that your winning percentage is constant at all rating levels. He is making the assumption that we would have that same 80% winning percentage in the 1800 brackets that we have using the Elo-reset system in the 1500-1600s, which is just unlikely.



EDIT:
After going onto Wikipedia and reading more about the Elo rating system, I found that this system suffers from a few flaws that were ignored by the Blizzard developers. Initially, the Elo system was created for the Chess circuit so that chess players around the world would have a ranking system by which to set themselves. This is a natural idea, so the Elo system was created, which would state that if you were a 1600 and played against a 1600 and won out-right, you were obviously more skilled than your opponent, so you should be rated higher, and consequently, he should be rated lower.

This is interesting because the adaptation to other games seems like a reasonably good idea. However, if we look at the aspects of chess, it becomes rather clear as to why this cannot (rather, should not) be adopted to WoW arenas. Chess is a one-on-one game that pits skill AND ONLY SKILL against an opponent. Chess is a dance amongst pieces that can all take other pieces (except king cannot take king, etc); that is, there is no chance that when I move my rook down the board and onto your knight's square, he will overcome my advance and defeat my piece in a defensive maneuver. Likewise, there are no auto-attacks that are made for me without my having to play them fully aware of their consequences. Therefore, we can plainly say that the game of chess has no luck associated with it (unless you count your opponent's inability to see your next move... but I would argue that such is skill rather than luck).

Because there is no luck in chess, then it is completely suitable to create a rating system that states when player x beats player y, x is better than y. Logically, the argument holds water, so to speak. However, the moment luck is introduced into a game, such a rating system is dashed to pieces. If, for example, I make a last-ditch effort to take your piece with my piece, and this move will either win the game for me or lose the game for me, and the outcome is based on probability a la "20% of the time, this move will succeed," then we cannot say conclusively who is the better player. Sure, the argument to use such a low-percentage move in such dire straits could (and would) undoubtedly be argued as a gamble rather than skill, but what if the move had a 50% chance to succeed? Could we then argue that player X's defeat was due to lack of skill when player Y quite obviously had a similar chance to win? Such would be the arguing over whether flipping a coin the way the flipper wants requires skill. Quite clearly, adding any luck-based components to a game diminishes the ability to, without bias, state who is a better player.

Another reason that the Elo system is inadequate for use in the arena rating system of WoW is because the Elo system accounts for ties and close matches. Yes, the arena does account for ties as well, I have seen screenshots of zero gain matches that ended in a no-decision because sinister strikes landed at the "same" time. However, in chess, if two players play amazingly well and end up with just kings left, and one player finally corners the other player into no possible moves, then this is a narrow victory for the cornering player. It could be argued that such a victory was as close as victories come, and therefore the players were pitted equally against one another. So, their ratings will change very little after the match is concluded because of the nature of the victory. In this example, 'close' matters in both horseshoes and chess. The arena system does not have a way to determine close matches as "near-draws" and award points based on narrow victories or losses. If we play an arena match that lasts 5 hours and ends in a narrow victory with one player killing the other while keeping 2 hit points at the end, with comparable healing and damage done (etc etc), then one could argue that such a match (epic though it may be) would be a "near-draw" resulting in only 1-2 rating differences between the two teams because they were obviously near-equals with regard to one another (so long as their respective ratings were equal).

Personally, I feel the need for a new "fair" rating system that 1) is impossible (read: extremely difficult to do, or easily noticed and punishable) to exploit/game, 2) renders fair judgments and rewards between losers and winners, and 3) pits skill against skill, rather than gear against gear. I do not have any ideas for such a system, but I will have about 2 weeks (taking a vacation) to mull it over and talk with my brother about it. Such a system would need to be WoW-specific, and therefore take a grand scheme of things into account when determining rating. For instance, two undead rogues should always be able to defeat a warlock+shadowpriest team... every time if the two rogues are not incompetent. This is due to the nature of the rogue class having magic immunity, fear immunity (as undead and the pvp trinket), and amazing damage output against low armor classes. Sorry for the tangent here, but I know first hand how annoying it can be to go against a counter-comp team and lose 18 rating with no chance at winning, and I think that these scenarios should be addressed and accounted for the in fair rating system for the arena.

Oh well, if I come up with anything good, I will post on it back here.