Frequently Asked Forum Questions | ||||
Search Older Posts on This Forum: Posts on Current Forum | Archived Posts | ||||
Re: Your analysis is grossly incomplete | |
Posted By: Hoovaloov | Date: 1/17/12 4:54 p.m. |
In Response To: Your analysis is grossly incomplete (RC Master) : Lower bound, upper bound, 1st quartile, median, mode, 3rd quartile, : inter-quartile range, upper bound, range. Lower and upper bounds are easy to spot, just look at the top and bottom of each column of data. : Mean average is susceptible to unfair skew from extreme values. Crunch the
Median and Mode don't perform well in such a small sample size. For example, in the Long Range 85% bloom data set, the Mode is 7 shots, but that only occurs 5 times out of the 25 kills. That throws out 80% of the data if you look at Mode. Median suffers the same problems in a small sample size. To truly have a good Median, I would need a much larger sample, and this analysis already took forever. But if you would like to crunch the numbers further, I encourage it. I have provided the raw data, and I think you can download your own version of the spreadsheet under the File menu of the Google Doc. I would even be willing to provide the gametypes and maps I used if you want to add data to this set. Regardless of averages/medians/modes/etc., if you look at the number of paced wins for each range in 100% and 85%, it is clear that 85% increased the number of paced wins. Below, I explain that calculating the number of paced wins is a viable statistic, regardless of how they are sorted. : To be clear, simply running them off side-by-side from quickest to slowest
Interestingly enough, instead of comparing fastest paced and fastest spammed all the way to slowest paced and slowest spammed, if I reverse one data set and compare fastest and slowest all the way to slowest and fastest, it results in the exact same amount of paced wins for every data set. Since any pacer's times CAN go against any spammer's times, and since the method of sorting did not change the results, I decided to leave it sorted. Therefore, my method of determining the amount of pacing wins is viable. : I'd also like to see you run the numbers for a TU Needle Rifle alongside
I agree, the Needle Rifle is its own strange beast. But since the DMR is the most common starting weapon in Matchmaking, and NR starts are only in gametypes with Elites, I determined that I should leave out the NR. Plus, this study took a long time as it was. :) : Additionally, 'pacing' and 'spamming' aren't the only two ways to use a DMR
I realize that purely pacing and purely spamming aren't the only ways to use the DMR. But if I started mixing in spammed shots to the paced, and paced shots to the spam, it would muddy the results. In fact, the absolute fastest way to kill someone is likely 3 spammed shots to the body, a short pause, another shot to the body, followed by a quick headshot. Or some other combination of spam and pace. : Your data is also incomplete. You mention that "I also compared the
The data is not incomplete, I just didn't spell out everything for you. In either bloom setting, 5 shots is the minimum it takes to kill without any missed shots (4 body + 1 headshot). In 100% bloom, 8 shots is the maximum it takes to kill without any missed shots (8 bodyshots + 0 headshot). So anything under 8 resulted in a headshot in 100% bloom. Anything 8 or over doesn't really matter, headshot or not, since that last bullet would kill either way. In 85% bloom, 7 shots is the maximum it takes to kill without any missed shots (due to damage bleedthrough). So anything under 7 resulted in a headshot in 85% bloom. Anything 7 or over doesn't really matter, headshot or not, since that last bullet would kill either way. Since I provided the number of shots for each kill, it's a matter of simple math to figure out how many shots missed for any giving data point. Just subtract 8 from 100% bloom results, and 7 from 85% bloom results. But the number of headshots is not very important if one method or the other is winning the majority of the battles. And I've shown that 85% bloom increases the number of pacing wins for each range. : Furthermore, your methodology is incomplete: are these kills zoomed,
All shots were fired unzoomed across all ranges. Why would I even think to change the zoom? I showed my setup in the picture. Yes, as I stated in my Methods, I aimed at the head. I do not think it matters how big the size of the target is as long as I used the same target for all tests. If I used a bigger target, it would probably dampen the results since the whole point of the test is to determine accuracy, and aiming for a big target makes all the shots more likely to hit. So a small target would more clearly show if 85% bloom was increasing or decreasing the accuracy of spammed shots. : Finally, average ping across all matches was quoted as 100ms in Halo 3 - or
This test was performed in an XBL lobby, but since the game included just me and a guest (clearly visible in the picture), then the game was hosted on my Xbox, the same as a Local Lobby. Since Reach is a completely rebuilt engine, we have no reason to assume that anything is the same as Halo 3, even online "tolerances." But I agree there is some amount of "tolerance" for headshots that can result in simultaneous deaths. Since I do not know this tolerance, we can't determine how it affects the results. : P.S. the Medium range, 85% bloom, pacing category is not sorted. Thanks for the heads-up, I have fixed it. (Notice that sorting it didn't affect the amount of pacing wins!).
|
|
Replies: |
The HBO Forum Archive is maintained with WebBBS 4.33. |