Average Match Time Calculations from tournaments

[Keitaro]

This is a placeholder for data from tournaments regarding average play time for Smash 4 within a set number of tournament sets. Data will likely be taken directly from stream footage as data from every match of the entire tournament isn't so necessary.

*updated 9/30/2014*

Link to Tourney Locator's 3 stock tournament with kill percents and average time per stock: http://smashboards.com/threads/blas...m-tourney-locators-invitation-tourney.368643/

Crossfire Smash 4 Times:


Peach vs Donkey Kong - 3:10
Little Mac vs Donkey Kong - 2:06
Little Mac vs Yoshi - 3:03
Little Mac vs Yoshi - 2:28
Little Mac vs Yoshi - 2:22
Little Mac vs Yoshi - 2:28
ROB vs Sonic - 3:10
ROB vs Sonic - 3:44
ROB vs Sonic - 3:46
Little Mac vs Diddy Kong - 2:17
Little Mac vs Diddy Kong - 3:01
ROB vs Diddy Kong - 3:32
ROB vs Diddy Kong - 4:09
ROB vs Sonic - 2:35
ROB vs Sonic - 3:42
ROB vs Sonic - 2:56
ROB vs Sonic - 3:40
ROB vs Sonic - 4:18
Sheik vs Sonic - 4:06
Sheik vs Sonic - 3:43
Sheik vs Sheik - 1:36

Average time for a 2-stock match: 3:08.4

 

[Overswarm]

Don't use stream footage, as that will be bunk. Just wait until you can host an actual tournament or get a lot of people playing the japanese 3DSes for impromptu tournaments.

An 8 man round robin (if everyone is playing a different character) would actually be a pretty decent start when looking for things like consistency. Would you be able to do something like that? Have 8 people over all play each other?

http://smashboards.com/threads/data-compilation-for-smash-4-3ds-wii-u.367949/

You can use that^ for even more data. :B

 

[Keitaro]

I don't see how it would be bunk. It's similar to taking a group of 1000 people and surveying them out of millions for a decently accurate representation of the rest of the population's opinions. The stream footage contains both lower level and higher level matches too. With more of an emphasis on the higher level ones.

 

[Overswarm]

I don't see how it would be bunk. It's similar to taking a group of 1000 people and surveying them out of millions for a decently accurate representation of the rest of the population's opinions. The stream footage contains both lower level and higher level matches too. With more of an emphasis on the higher level ones.

I'm not gonna go into the intricacies of sampling, but trust me when I say you will get bad data from this exercise. No matter what the results say, they won't be worth anything on their own.

There's three things to do when sampling:

Find the problem you want to define:
"Do competitive matches take too long?" / "How long do competitive matches actually take?"

Define the population from which the sample is drawn:
This would be the players, characters, and stages you choose. For our case it would be "all competitive players" (we don't care about goofball friendlies), "all legal characters", and "all legal stages" combined.

Find a representative sample or subset of that population:

That last part is where you trip up, as the data will become muddy. If you watch Zero's stream and False's stream, you may find that matches end very quickly for False and take forever for Zero across the board because of play style. Or maybe False only plays Mario and Marth and his games end quickly, but Zero switches characters over and over. Tons of issues pop up and what you'll end up with is "here's how long Zero and False spend on their matches, on average" and even then it'll be wonky.

Survey sampling is very difficult to do properly. Even professionally done surveys have issues with this.

Given the wide amount of variables (stage, character, player, playstyle, etc.), the best way to get a representative sample is to just limit the variables.

Pick one stage
Pick one character per player (no duplicates)
Pick a select amount of players

Then have those players with their set-in-stone characters all play on the same stage against everyone in a round robin, best of one is fine.

The more stock you have (and tracking how long it took for each stock to be taken), the clearer your data will be. I'm personally going to be using 4 stock and a larger time limit on Oct. 11th so I can garner as much data as possible.

With this setup, your information should then be more focused. You'll lack certain information (if no one plays Megaman, you won't know Megaman's effect on time limits, and since you only play on one stage it only really applies to "that stage"), but you can typically extrapolate a decent guess from this information because it will be so pure.

Watching random streams to record information won't give you much clarity at all, so the conclusions would be shaky without a huge, huge data set.

 

[Keitaro]

Updated with data from Tourney Locator's tournament and the recent Crossfire Smash 4 2 stock tournament.

 

[Ishiey]

Not a lot of data, but so far this still supports 3 stock being fine, which is nice.

Do we have a tally of how many tournaments are running 2 vs 3 stock?

 

[Zigsta]

@ Keitaro , would it be beneficial for us to talk with our respective TOs to have players write down their match times and characters? Figured I'd check since we're super close to launch and I imagine the majority of us will be going to tournaments this weekend.

 

[Overswarm]

Really the only variable that needs to be tracked is "how often do timeouts occur" (and possibly "are those timeouts deliberate"). If every match takes 6 minutes but timeouts never occur, it's not really an issue.

That said, more data is almost always helpful.