Friday 27 February 2009

Rating Distribution - Challenger at 1505+?!

While I finish re-writing the stats presentation part of my code, I thought I would put up some data that people may find interesting. (I am also fed up of people claiming that far more players are at 2k+ due to the new system without ANY evidence whatsoever)

I have pulled together the data from 7 EU BGs using a mix of 2s and 3s to plot a rating distribution, the reason the sample size is relatively small is because I only wanted to use listings with between 3000 and 4700 records. Any less and the sample size starts to get a bit small (hello Ruin 5v5 with 284 (!) active teams), any larger and they likely don't have all the teams.

Please note that while I have specified values these are approximate only as they are averages, and the actual rating reqs for each title will vary by BG.

So, the charted ratings look like this:



As you can see, the top 0.5% has quite a spread, spanning 500 rating, whereas 25% of players are in the 245'ish spread between 1505 and 1750.

In addition 50% of players are at or below 1465 rating. With some of the Hateful gear requiring Rival status it is no wonder that more and more players are dropping out of the arena, I know I personally am finding matches at 1400-1600 harder than I found matches at 1700-1800 in previous seasons.
In part this is down to meeting a huge number of players who outgear us in Naxx 25 welfare gear which allows for people to be dropped almost instantly, however it is also because a lot of the 1600+ players from last season are in the 1400s. (and of course in part because I suck :))

The recent rating changes seem to be clumping people together in the low-mid range, and stretching it out as you go up the scale. This is why many people labour under the assumption that far more people are at 2k than before. They aren't, it is just that the spread is now higher at the top end.

Population distribution shows a similar trend:



You can see here a definite clumping in the lower-mid range, with the top end being extended further out.

I will see if I can pull out some comparisons on actual numbers compared to season 3. (I never bothered with season 4)

Friday 20 February 2009

Minor update

Just a quick note to say that I have re-written the parsing code and am parsing the EU BGs at the moment. However, as there are half as many teams in the EU-EN BGs compared to season 4 (never mind s3 etc.) I have decided to add the French and German BGs to the parse as well.

Now I just need the armory to be reliable so I can parse the bleeding data without much intervention. :)

Tuesday 3 February 2009

Updates planned.

OK, chances are that I am talking to myself as Arena has turned into a shitfest (especially for players like myself, the so-called 'bads' that are a target of derision and mockery even though we are probably far more handsome going off the RL pics of the majority of the sponsored players) and anyone who did find their way to this blog has likely given up when they saw it was dead for 6 months.

The issue with my previous data collection was that it was time consuming, boring (like Warlock Druid vs Warlock Druid, or DK+Pally vs DK + Pally boring) and all round too slow.

So, I had every intention of re-writing my code to be faster, sleeker and more informative. But I have been too lazy, partly due to work commitments, partly due to family and mainly cos I was getting bored of WoW. That said I am finally getting curious enough about the populations of s4 vs s5, rise of the DK/Pally and the fall of the lock to want to poke around a bit and find out more.

It wont happen overnight but over the next month or so I want to be providing more regular updates, and a more interactive way of looking at the data.

So if you still check in now and again I hope to have something soon(tm)!