Hey guys, between spare moments I did some quick math based on the the last barwars stream footage to give an idea how how stock count affects event length in our community.
I looked at the 2nd half of the previous event, which spanned from winner's semis through grand finals. (Actually a bit more than half the tourney: 5 bracket-stages/sets spanning 17 games, taking a little over an hour.)
The average game length was 2:24, or 1:12 a stock. This is including both outliers: The one 18-second game, and both of Fujin's very stally games. It would be 1:09 per stock with these omitted, but I think that's unfair; both these sort of things do happen.
Based on empirical data from other events in the smash community, we know that 3-stocks plays faster per-stock on average, though only slightly--about 5 seconds. (There are a few explanations for this, including rage and a slight tilt towards aggressiveness.)
We can estimate that if last week's tourney were 3-stock, those 5 stages would have taken around 18 minutes longer. The entire bracket would have taken around 25 minutes longer.
This is not counting any early round "Game 3s" (or 4/5) that additional stocks would prevent. 3-stocks minimize flukes, which makes slightly fewer sets go to Game 3. In the previous two tourneys, my first round went to game 3 due to a 0% SD. In both cases I played it out, and barely lost 1-vs-2. It is more likely I could have won these games 2-vs-3, in which case a 3-stock set would have actually taken *less* time than a full Bo3 2-stock. Situations like this are not included in the previous estimate.
For context, let's compare this to other decisions we make that affect time:
Because we were ahead of schedule, we decided to delay WF to run a loser's bracket match (Manny vs. Fujin) on stream. (This was my suggestion) This took an extra 8:06, and is included in the previous 3-stock estimate. Assuming we take this away, 3-stock would add only be a 15 minute difference.
Running WF, LF, and GF Bo5 instead of Bo3 added 18:23 additional time. If we went to GF round 2, it would have been around 22-25 minutes total. I'm not saying we shouldn't do Bo5 (spectators seem to like it), but just giving context for the time cost of that decision.
Eric [the TO] had over 66% setup-time efficiency for the second half of the event, which is remarkable and goes under-appreciated. He does a great job getting people to playing their bloody games, like all TOs should. What if he did even a slightly worse job? If setup efficiency was reduced merely to 50%, the event would have taken a full additional 60 minutes.
tl;dr - The bottom line impact of stock count is surprisingly low, and for an event like last week would add to 12-15 additional minutes.
Finally, a quick note about match time limit. It might be counter-intuitive, but lowering the time limit *INCREASES* average match length. (For reasonable bounds, obviously) This is because the lower the time limit is, the more viable stalling is. (This sort of statistical phenomenon is known as a "Laffer Curve".)
We have witnessed this first hand in our community. As I said above, our average stock length for the last event was 1:12, even including Fujin. We did not have a single game go to time; except Fujin, the longest game was 3:39, and the second longest was 2:57. (The last two matches of GF)
At this rate, our 3-stock matches should take around 3-minutes average, and around 4:30 usual worst case. Yet when we held a 3-stock 5-minute tourney, a very large percentage of games went to time, including 5 of mine! This is because stalling became a dominant strategy.
The other night at Nicole's the main TV got accidentally set to 3-stock 5-minute for the entire night, except that nobody noticed and assumed it was 8-minutes. We were surprised when we realized it--not a single 3-stock game the entire night had hit that time limit, even when Fujin was playing.
tl;dr - Higher time limit actually results in shorter matches on average.
