Development Blog Update - Game Statistics
MurphyIdiot
NS2 programmer Join Date: 2008-04-17 Member: 64095Members, Super Administrators, NS2 Developer, Subnautica Developer, Pistachionauts, Future Perfect Developer
NS2 programmer Join Date: 2008-04-17 Member: 64095Members, Super Administrators, NS2 Developer, Subnautica Developer, Pistachionauts, Future Perfect Developer
Comments
What i find interesting is how similar the stats are for tram and rockdown (ok they fluctuate... but the other day they were nearly identical :P):
<a href="http://unknownworldsstats.appspot.com/displayendgamestats?map=ns2_tram&version=162" target="_blank">http://unknownworldsstats.appspot.com/disp...amp;version=162</a>
<a href="http://unknownworldsstats.appspot.com/displayendgamestats?map=ns2_rockdown&version=162" target="_blank">http://unknownworldsstats.appspot.com/disp...amp;version=162</a>
Glad to see version 163 fixes these issues >:D
<a href="http://unknownworldsstats.appspot.com/displayendgamestats?&version=163" target="_blank">http://unknownworldsstats.appspot.com/disp...amp;version=163</a>
What we really need is a zip file (just put it up on filefront to save on bandwidth) containing all stats for this patch so we can find out which way to swing the nerf bat.
aside from that, good chart but I rather think marines are overpowered rather than aliens though. Alien abilities are rather gimped, if not broken in general.
aside from that, good chart but I rather think marines are overpowered rather than aliens though. Alien abilities are rather gimped, if not broken in general.<!--QuoteEnd--></div><!--QuoteEEnd-->
You mean kinda like what Steam does from time to time? (asks if you want to participate, then shows you neat graphs and %'s, breaking down all the responders' hardware setups, piece by piece)
Not only would that be cool, but we could break it down further - collect the normal resolution/graphics settings people play at, as well as their average fps over the last couple times they've played. Cross-examine that average fps (or other benchmarking stats) by hardware pieces and you could get some interesting results.
Those pesky lerks!
Not only would that be cool, but we could break it down further - collect the normal resolution/graphics settings people play at, as well as their average fps over the last couple times they've played. Cross-examine that average fps (or other benchmarking stats) by hardware pieces and you could get some interesting results.<!--QuoteEnd--></div><!--QuoteEEnd-->
yup, something like that. I wouldn't mind sending this type of information, if this would help I gladly share it with developers here. it seems to help valve when they do it, we should have this as well if it can further help development. Plus I would love to see statistics of all the data collected, would be great to see what type of computers people using, among other things!
Would it be difficult/cumbersome to allow reporting on a larger group than 100?
It isn't difficult. I am just not sure what our CPU/bandwidth limits will be like. I can probably add an option to return X number of results or have pages of results.
<!--quoteo(post=1829812:date=Feb 4 2011, 08:56 PM:name=Kouji_San)--><div class='quotetop'>QUOTE (Kouji_San @ Feb 4 2011, 08:56 PM) <a href="index.php?act=findpost&pid=1829812"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->I'd love to see those awesome hotspot maps included into this like I've seen on other games, where the most action/kills/building spots and so on. Perhaps even adding data on which powernodes get chomped or repaired the most to get a nice overview of the gameflow of the maps.<!--QuoteEnd--></div><!--QuoteEEnd-->
I have been thinking of such things lately as well. I want to get some other statistics like deaths, framerate, etc that are tied to positions on maps so we can eventually generate heat maps like this. Neat stuff :)
Would it be difficult/cumbersome to allow reporting on a larger group than 100?<!--QuoteEnd--></div><!--QuoteEEnd-->
That's true, but their are ways around that. For example, use multiple trials where one trial = the last 100 games played downloaded at the same time each day. If you did this for a weak, you'd probably get a good series of data which you could then average to get the 'true' values. More data is usually better, but the key is making sure you get as random a sample as possible.
Maybe save how many players were present at round start and end, because that makes a huge difference.
i really like stats archievment in every way. would be nice if you could extend this, so we dont need psychostats etc?
*thumbsup*
How interested are you in having a Blackberry stats client visualizer? I could put one up for you guys.
And on a sidenote, personal statistics on a mobile app would also be pretty cool. Kinda like how some devs do with the BFBC2 api to retrieve soldier stats.
Anyway, PM or MSN me if you´re interested.
Have you tried experimenting with <a href="http://www.varnish-cache.org/" target="_blank">varnishing</a> the output? You could use it to generate a much larger json dump once every... say, 15 min, and anybody requesting the data would just see the latest cached copy. In effect, your webserver would just be serving up a page full of text, which shouldn't cause too much strain at all.
Finding the right balance between keeping your server load down and giving us access to real time data will be an understandably tricky one.
No, forget that. it's wasting more resources than requesting and backing up the json everytime its updated and merging it with the old data.
<img src="http://vicious-violet.com/Misc/OtoyaReactionFace.jpg" border="0" class="linked-image" />
Every game i play the marines win if they got the flamethrower.
But most times the skulks just rush and win the game.
Lerks are very easy with a grenade launcher, shotgun or flamethrower.
Anyway, here goes again:
Raw JSON returns as content-type text/html, but as it’s pure JSON, could you fix the content-type to JSON?
Also, you may want to make it callback-able, for (easy) JS inclusion.
If you need help on any of that contact me.
Every game i play the marines win if they got the flamethrower.
But most times the skulks just rush and win the game.
Lerks are very easy with a grenade launcher, shotgun or flamethrower.<!--QuoteEnd--></div><!--QuoteEEnd-->
well, there is a game length attribute in the data, so, as soon as someone takes a look, we will know that for sure...
team 1 is Marines, and team 2 is Aliens I assume?
*Edit*
Quick/ dirty look at the time of games (I assume in mins) for each team/ each map.
a few of the matches in the last 100 ended in 2 or 3 seconds, so they might have been people leaving matches.
<a href="https://picasaweb.google.com/lh/photo/lP8e9_I3yDKv21yZq2aMTg" target="_blank">https://picasaweb.google.com/lh/photo/lP8e9_I3yDKv21yZq2aMTg</a>
How about caching them for half an hour when a filtered request is made?
Or even easier, generate the JSON with a cronjob, save it as a static file and make the static file available to us. We can parse it and filter it ourselves.
a few of the matches in the last 100 ended in 2 or 3 seconds, so they might have been people leaving matches.
<a href="https://picasaweb.google.com/lh/photo/lP8e9_I3yDKv21yZq2aMTg" target="_blank">https://picasaweb.google.com/lh/photo/lP8e9_I3yDKv21yZq2aMTg</a><!--QuoteEnd--></div><!--QuoteEEnd-->
I think this would be a bit more useful as a line graph, king like this:
<a href="http://infosthetics.com/archives/people_time_spend.jpg" target="_blank">http://infosthetics.com/archives/people_time_spend.jpg</a>
but with only the 2 options.
Also, would like the option to specify both minimum and maximum game time, to make it easy to filter out very short games.
It was quiet a fun match until it was ruined.
...
But most times the skulks just rush and win the game.<!--QuoteEnd--></div><!--QuoteEEnd-->
The skulks are only winning because the turrets are prohibitively expensive. Marine start is most vulnerable torwards the beginning of the round, and the result is the quantity of stats regarding "long" games is minimal which could make them statistically insignificant.
<!--quoteo(post=1829844:date=Feb 5 2011, 06:28 AM:name=Price)--><div class='quotetop'>QUOTE (Price @ Feb 5 2011, 06:28 AM) <a href="index.php?act=findpost&pid=1829844"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec-->Every game i play the marines win if they got the flamethrower.<!--QuoteEnd--></div><!--QuoteEEnd-->
Sure, because a team that has the FT also has many rts, some/all L2 armor + damage upgrades, and possibly 2+ CC's locked down or equivalent hives knocked down, etc. The FT itself is exclusively a skulk/lerk killer. The other day i had to reload 3 times to kill an unbuilt whip which still managed to fully grow! It just plain sucks.
The majority of marine victories in v162 are the result of marines armed with shotguns.
<!--quoteo(post=1829844:date=Feb 5 2011, 06:28 AM:name=Price)--><div class='quotetop'>QUOTE (Price @ Feb 5 2011, 06:28 AM) <a href="index.php?act=findpost&pid=1829844"><{POST_SNAPBACK}></a></div><div class='quotemain'><!--quotec--><b>Lerks are very easy with a grenade launcher</b><!--QuoteEnd--></div><!--QuoteEEnd-->
Ya good luck with that tactic lmao.
<a href="http://tenfiftyfour.com/axperiment/experiments/05_ns2/" target="_blank">http://tenfiftyfour.com/axperiment/experiments/05_ns2/</a>
Pretty simple for the time being, but it's nice to have some data to play with! I'll keep working on it in my spare time!
- Size of circle indicates length of the game
- Green = aliens win
- Blue = marines win
- Mouseover for some details (date, map etc)
Toggle between arrange by winner / version down the bottom
(edit : I just noticed the dates are totally wrong! I must be parsing them incorrectly, let me see what I can do about that)
(edit : Dates fixed!)