Although I think everybody agrees by now that SC2 is suffering from small maps and other factors which are detrimental to the spectating experience I felt like looking at some numbers. So I looked for some replay packs from the good old times and also from more recent events and went through the painstaking process of manually copying all the games' length in order to analyze them. I chose random events from different years with total replay numbers between 78 and 192, namely the WCG Grand Final 2004, the Team Areola Nipple League of 2005, WCG 2006 and TSL#2 2010 for Broodwar and Blizzcon2010 and the Kaspersky Invitational for Starcraft 2.

Here are the results in one picture, ordered chronologically:
[image loading]

Starcraft 2 games are both significantly shorter than BW games and their game length also has a much lower variation, which might be a nice thing for tv channel's programming staff (oh, wait, there are no SC2 tv channels) but sucks for me.

I'd loved to have statistics for pro gaming matches, but since there are never replays, I didn't know a comfortable way to get data.

Fa1nT

United States3423 Posts

January 15 2011 19:53 GMT

It will deviate as people get better and less people all-in off 1 base

Whiladan

United States463 Posts

January 15 2011 19:55 GMT

You need more samples from multiple regions. You can't conclude anything from this sampling.

EtherealDeath

United States8366 Posts

January 15 2011 19:55 GMT

I don't know if there is a convenient way to get the length of a flash video from the file, but if so, perhaps you could parse the links in TLPD, and then assume that because there is some padding in some games (maybe some intro or something, which hopefully over the long run is mostly averaged out), the mean might be say, 30 seconds lower than it looks, although the variance I suppose would be harder to examine. Maybe you could then individually look at the longer games and ensure there is not a lot of padding?

Definitely not as convenient as replays though..

edit - how about using the www.sc2rep.com database of replays? Use those for further sc2 data. And then maybe use the ygosu replays for bw data.

Holgerius

Sweden16951 Posts

January 15 2011 19:59 GMT

I remember once when I was watching Sc2 and BW simultaniously, and the entire Bo7 (4-0, might've been Jinro vs MC) in SC2 took roughtly as long time as one single TvT in BW.

True story.

Numy

South Africa35471 Posts

January 15 2011 20:01 GMT

Your sample size is tiny. There's enough data out there to have a bigger size. Really strange to start a topic, have an opening discussion but terrible data set.

HollowLord

United States3862 Posts

January 15 2011 20:01 GMT

4 BW tournaments and 1 SC2 tournament, not a large enough sample size.

lyAsakura

United States1414 Posts

January 15 2011 20:01 GMT

On January 16 2011 04:59 Holgerius wrote:
I remember once when I was watching Sc2 and BW simultaniously, and the entire Bo7 (4-0, might've been Jinro vs MC) in SC2 took roughtly as long time as one single TvT in BW.

True story.

Sometimes happens in BW too when it was bo5 format.
+ Show Spoiler +

holy shit that guy has 3.3k posts in 6 months.

InvalidID

United States1050 Posts

January 15 2011 20:02 GMT

Blizzcon is pretty old. You should look at GSL4, if anything, for your sc2 data.

Kaasflipje

Netherlands198 Posts

January 15 2011 20:04 GMT

#10

You should take some proleague/osl/msl stats and compare it with GSL stats. It's much more accurate.

Asta

Germany3491 Posts

January 15 2011 20:05 GMT

#11

On January 16 2011 04:55 EtherealDeath wrote:
I don't know if there is a convenient way to get the length of a flash video from the file, but if so, perhaps you could parse the links in TLPD, and then assume that because there is some padding in some games (maybe some intro or something, which hopefully over the long run is mostly averaged out), the mean might be say, 30 seconds lower than it looks, although the variance I suppose would be harder to examine. Maybe you could then individually look at the longer games and ensure there is not a lot of padding?

Definitely not as convenient as replays though..

edit - how about using the www.sc2rep.com database of replays? Use those for further sc2 data. And then maybe use the ygosu replays for bw data.

Yeah, I didn't want to use video file length's because I thought that the overhead would depend on what event the videos were from. But also I don't know how to automatically retrieve the links from TLPD and I suspect that you need to load the whole video to find out the length, unless the host has already done that and provides the time somehow. Maybe somebody with more knowledge and more time could do that if they are interested.

I consciously chose to pick specific events for the analysis because it makes sure that the players are of high caliber and more importantly of similar skill level. Random games might either involve short games when one player is too strong or very long games because those are always favored by replay collectors.

And to the guy who claimed the sample size was too small: I don't think so.

Tippany

United States765 Posts

January 15 2011 20:05 GMT

#12

I'm not as disappointed by the average game length. I think that is more or less dictated by the maps. But as much as I love SC2 (I don't play BW anymore) I still prefer to watch BW pro games. There's constantly action taking place all over the map whereas SC2 is almost always one big control group vs another.

blackbrrd

Norway477 Posts

January 15 2011 20:24 GMT

#13

After GSL4 was split in Code A and Code S matches, I think you can see that the Code S matches are longer and the Code A matches shorter. In other words, the more even the competitors are, the longer the games will be.

GSL4's Code S is probably the first properly seeded tournament so far in SC2 history and already we are seeing longer and better games. In the above data, not a single SC2 game was over 30 minutes, while we have already had two TvT that are longer than that in GSL4.
+ Show Spoiler +

Slayers_Boxer vs oGsHyperdub that lasted 32:38 and LiquidJinro vs oGsEnsnare lasted 50:27, and it only ended because oGsEnsnare had mined out his half of the map. I believe about 50 minutes is the maximum length SC2 will currently have for games, due to map size and mining out.

Trap

United States395 Posts

January 15 2011 20:26 GMT

#14

True story.

Not a very fair comparison... empires rise and fall during a Canata TvT.

gillon

Sweden1578 Posts

January 15 2011 20:27 GMT

#15

I'm sure SC2 games will be longer when it's been out for 5+ years aswell.

DuckS

United States845 Posts

January 15 2011 20:29 GMT

#16

Not only is your sample size bad, but look at the tournament dates. Brood war also had more time to evolve, SC2 hasn't even been out for a year yet..

Piski

Finland3461 Posts

January 15 2011 20:30 GMT

#17

Interesting stats but they don't really say anything conclusive

Holgerius

Sweden16951 Posts

January 15 2011 20:40 GMT

#18

On January 16 2011 05:26 Trap wrote:

Show nested quote +

Not a very fair comparison... empires rise and fall during a Canata TvT.

Ya, it wasn't a serious argument regarding the issue.

I just found it sort of amusing.

Shikyo

Finland33997 Posts

January 15 2011 20:43 GMT

#19

You're not using the in-game time for SC2, right? It's different.

By the way, early SCBW had mostly 1base allins and cheese as well and the skill levels were so lopsided that the players beat each other very quickly.

Holgerius

Sweden16951 Posts

January 15 2011 20:49 GMT

#20

On January 16 2011 05:43 Shikyo wrote:
You're not using the in-game time for SC2, right? It's different.

By the way, early SCBW had mostly 1base allins and cheese as well and the skill levels were so lopsided that the players beat each other very quickly.

The importance of macro is already well-known now though, through BW. Sc2 shouldn't need to go through the same evolution.