I'm cross posting this from the current ECM feedback thread so people can see some of the numerical effects of ECM now that it has been out for several weeks and the meta game has had a chance to evolve to counter it. This is also data taken more than a week after the TAG range boost patch so that is factored in too and is now helping to counter ECM:
The simple question is - is ECM overpowered enough that having 1 more or less than the other team is enough to predict the outcome of the match.
After all if the whiners would shut up and just counter the ECM it would hardly matter and they would beat the other team just as often as not right?
Well, let's see what the numbers say:
Methodology: I dropped into PUG games exclusively with 1 friend on skype - he helped me count up ECM and TAG instances during the games.
(Out of curiosity I also counted TAG systems on each team to get a gage of how smart of an idea taking LRMs into a PUG match is)
Our side always has 1 ECM since I piloted a trollmando 2D or Craven 3L for the duration of the tests.
Us: ECM 2, TAG 0, Them: ECM:3, TAG 1, Outcome: We Lost Matched Expectation: Yes
Us: ECM 1, TAG 0, Them: ECM:3, TAG 0, Outcome: We Won Matched Expectation: No
Us: ECM 1, TAG 0, Them: ECM:1, TAG 2, Outcome: We Won Matched Expectation: Inconclusive(1ECM each)
Us: ECM 1, TAG 0, Them: ECM:0, TAG 0, Outcome: We Lost Matched Expectation: No
Us: ECM 2, TAG 0, Them: ECM:0, TAG 0, Outcome: We Won Matched Expectation: Yes
Us: ECM 1, TAG 0, Them: ECM:3, TAG 0, Outcome: We Lost Matched Expectation: Yes
Us: ECM 2, TAG 0, Them: ECM:1, TAG 0, Outcome: We Won Matched Expectation: Yes
Us: ECM 2, TAG 0, Them: ECM:0, TAG 0, Outcome: We Won Matched Expectation: Yes
Us: ECM 1, TAG 0, Them: ECM:2, TAG 1, Outcome: We Lost Matched Expectation: Inconclusive(2ECM each)
Us: ECM 2, TAG 0, Them: ECM:0, TAG 0, Outcome: We Won Matched Expectation: Yes
Us: ECM 1, TAG 0, Them: ECM:2, TAG 0, Outcome: We Lost Matched Expectation: No
Us: ECM 3, TAG 0, Them: ECM:3, TAG 0, Outcome: We Lost Matched Expectation: Inconclusive(3ECM each)
Us: ECM 3, TAG 0, Them: ECM:1, TAG 0, Outcome: We Won Matched Expectation: Yes
Latest 5 games: Done 1 January 2013
Us: ECM 1, TAG 0, Them: ECM:1, TAG 0, Outcome: We Won Matched Expectation: Inconclusive(1ECM each)
Us: ECM 1, TAG 0, Them: ECM:2, TAG 0, Outcome: We Lost Matched Expectation: Yes
Us: ECM 1, TAG 0, Them: ECM:1, TAG 1, Outcome: We Won Matched Expectation: Inconclusive(1ECM each)
Us: ECM 4, TAG 1, Them: ECM:2, TAG 0, Outcome: We Won Matched Expectation: Yes
Us: ECM 3, TAG 0, Them: ECM:2, TAG 0, Outcome: We Won Matched Expectation: Yes
Evaluation: 5 matches had equal numbers of ECM and are marked inconclusive, leaving 13 matches. Of these 3 matches went against the hypothesis while 10 were for it. This is a small sample set but using the methods for binary outcomes on page 10 of this document (dealing with binary outcomes in a finite population)
http://classes.soe.u...nter03/h5m3.pdf I calculate the 1 sigma uncertainty at 1.52 matches.
This means that the outcomes could be wrong by up to 2.30 sigmas before the hypothesis would be neutralized/wrong.
Assuming a normal distribution etc. this gives erf(2.30/sqrt(2))= 0.979
This means that even with this small sample set I can state that
ECM is a match decider with 97.9% certainty. The other way of stating this outcome is that you are 3.3x more likely to win the match if your side has a larger number of ECMs on your team. I expect this latter number to reduce with a larger data set, but so far ECM seems to be the best way of predicting the outcome of a match - I would gladly bet money using this.
I will keep track of more of my matches to tighten these bounds but so far
I can state with (97.9%) confidence that the team with more ECMs will tend to win the match, all other factors being equal. If you think that this is the right way for a 1.5 ton 2 slot piece of equipment to be balanced, please stop using my air, there are other more deserving people who need to breathe.
Tag should also not be expected to be there in the vast majority of matches so do not take LRMs expecting a TAG to be around to help unless you take it yourself.
Edited by Tolkien, 01 January 2013 - 05:54 PM.