STATCraft w/ Playhem Dailies: Pros vs Scrubs

It's easy to see here that Pro vs Pro games simply last longer. Whether this is because Pros don't all-in other Pros as often, or Pros are better at surviving all-ins, or because Pros are less likely to engage in premature egrackulation, is unclear. But the games do last longer. Also while ZvT is the slowest matchup based on the full data set, Pro TvT does have the largest number of extremely long games. Meanwhile, Pro vs Scrub games are the shortest. Again, we don't know if this is because Pros all in and win, or Scrubs all in and lose, or just that Pros are better at closing out games against Scrubs once they get ahead. But it is a very real phenomenon.

Just to bring this home, here is a chart showing the cumulative game lengths in TvP based on the Pro/Scrub matchup (click for a full-sized image):


Playhem TvP Cumulative Game Lengths: Pros & Scrubs by niq77174, on Flickr

The gap here is largest in the 10-15 minute window. While 57.8% of Pro vs Scrub games end before 15 minutes, only 42.6% of Pro vs Pro games end that quickly.

In PvZ Pro games, Zerg players maximize their advantage in the early game, and Pro games ending in 5-10 minutes are more common than in Scrub games. This suggests that pro Zergs are recognizing more opportunities for early aggression and/or are more effective at repelling Protoss all-ins. If the aggression fails and the game continues, or both players play a macro game, the Pro Protoss has much larger advantage during midgame and "midlate" game--between 15 minutes and 25 minutes--than in Scrub PvZs. In the extreme late game (30+ minutes), the Pro Protoss advantage appears to erode, though the sample size here is very small; Protoss players are 29-35 in games lasting longer than 25 minutes.

The Pro vs Scrub matchup results are unusual. Because there are more Zerg pros than Protoss pros, and because Pro vs Scrub games end quickly, my hypothesis was that Pro vs Scrub would show a very high win rate for Zerg in the early game. In reality, the Zerg win rate in the early game is worst for Pro vs Scrub. Instead, Pro vs Scrub PvZ is characterized by less advantage for Protoss in the midgame and "midlate" game. I'll have to unpack the Pro vs Scrub games further to see what's going on here.

The Scrub v Scrub matchup shows Zergs dying to the 2rax, but heavily favored in games lasting 5-10 minutes. This suggests baneling busts, Roach/ling/baneling all-ins, and other forms of early Zerg aggression are quite powerful. Interestingly, in Pro v Pro, Zergs either decline to engage in early aggression, or are unable to force Terrans to give up in the early game. The Zerg advantage in this time window, combined with a strong advantage in the late game than in pro matches, means that at the Scrub level, the matchup tilts a hair in Zerg's favor.

In the Pro vs Scrub matchup, both races are seeking to maximize their advantages. Terrans press their advantage in the early game and early-midgame, while Zerg's play for the late game and go for a big 300 food push.

Pro vs Pro matchups show Terrans as modestly favored. Again, we see slightly more aggression than in Scrub matches; 15.7% of Pro ZvTs end before the 10 minute mark, but only 13.3% of Scrub ZvTs. Considering how little evidence we can find of successful Pro Zerg aggression, this means there is likely substantially more Terran aggression in Pro matches; though build order analysis is required to confirm this hypothesis. Overall, the pro ZvT matchup is characterized by less advantage for Zerg in all phases of the game than in Scrub ZvTs, except for the 10-15 minute.

The TvP lategame continues to confound even the pros. Combining all games longer than 25 minutes, Terran Pros are a staggeringly bad 9-23 against Protoss Pros (!!). While this is a small sample size, the probability of such a lopsided margin suggests that Protoss is at least favored, if not heavily favored, in the lategame. Terran Pros keep the matchup close by facing fewer all-ins, and by holding their own in the 20-25 minute bucket.

Likewise, the Scrub TvP games also show abysmal win rates for Terran as the game goes on. The situation here is worse than in pro matches; by the 20-25 minute bucket, Protoss is heavily favored and never looks back. This is all strong evidence that the current metagame strongly favors Protoss in the late game. If the players are of roughly even skill, we should expect the Protoss to win as the game drags on.

Is there any hope for Terran? Possibly. The Pro vs Scrub matchup shows a significant swing back into an even match in super long games. In addition, in the five 35+ minute Pro games, Terran players went 3-2. So perhaps it is possible for Terran to split the map and grind their way to victory if they are unable to end the game early. Also, I hear the factory can be used to produce units; maybe someone should look into that...

For the curious, those 5 games are:

• mousesports (P) vs DMApo (T)
  
• yooin (T) vs PswLuxury (P)
  
• RGNArtist (T) vs PswBoSs (P)
  
• coLrsvp (P) vs NoNge (T)
  
• GoSuPokebuni (T) vs Russano (P)