[SPL] Playoffs Week One - Two Steps From Victory

Ladies and gentlemen! Welcome to my little corner of this wonderful publication that is the SPL weekly news update! Today we will be looking at why Samsung Khan is the number one team in the world, why all the current SPL maps are actually Terran-favored, and why even Bunkie has a girlfriend when I don't.

Actually, I thought I'd talk a little bit about a very interesting game from the recent playoff match between KT and CJ. This would be Leta vs Stats on Jade from the first Bo7 series.
In this game, there was some early aggression from Stats but we basically got to an even mid-game situation with both players on three bases and Stats moving to take his fourth. What happened next was very interesting. Stats starts +1 air weapons at his cybernetics core as if going carriers. Leta presumably scans this, assumes that Stats is making the carrier transition based solely on the spinning core, then pushes out with a mere 120 or so supply and only +1 attack finished on his mech army.



Stats had in fact not gone for carriers at all and easily crushed Leta's attack with a 160 supply Protoss ground army supported by storms. Now my question is, why did this happen? Leta had three comsat stations at that point in time and Stats had not gone particularly out of his way to hide his templar tech or his gateways. Everything was in the main base but the core was placed at the very bottom right. Essentially the only way for Stats' fake +1 to work was for Leta to only use a single scan on the main to check the core and not scan anywhere else on the map. Had Leta scanned anywhere else in the main, he would have seen either the templar tech or a number of gateways indicative of a bigger ground army. Had he double checked Stats' fourth just before moving out, he would have seen that Stats had in fact not made a nexus there despite walling it in, and would perhaps have been more careful. Had he scanned Stats army, he would have seen way more units than one would expect from a Protoss making a carrier switch.



Individually, I don't think I can fault Leta for any of these oversights. When you see a Protoss getting air weapons at that point in a game, it's usually safe to assume a carrier switch. When you see a Protoss moving dragoons to his fourth base and walling it in with pylons, it's usually safe to assume he's expanding there. I'll even give Leta some credit and say that he did scan around looking for the Protoss army before moving out, which is a pretty standard move, but maybe he got unlucky and only saw bits and pieces of it. The question is, did Leta just get very unlucky with his scans or was he actually being too conservative with them and trying to get away with using as few scans as possible?

My initial thoughts on this game were actually that it might bear some subtle similarities to Flash vs Dear. Flash tried to make an optimization based on an assumption he made from incomplete information, and Leta appeared to do the same. If it's really so easy to trick a Terran player like that, why is it not standard play in PvT to start +1 air weapon after your third base is up? The +1 air weapon upgrade costs 100/100, which is almost negligible on a 3-base Protoss economy. An example of this concept from another matchup would be Zerg making a fake hydralisk den along with their spire in ZvT. A hydralisk den is extremely cheap (100/50) and basically forces the Terran to double-defend against both lurkers and mutalisks until he can get a better read on what the Zerg is doing. Is something similar going to occur in PvT?

The answer to that is actually probably not, but the interesting thing is why. As I said before, there were many ways Leta could have averted disaster in that game. All a Terran really has to do when he sees the core spinning is spend 1-2 extra scans around the Protoss main and then the fake +1 does absolutely nothing because the rest of the Protoss infrastructure reveals what the real strategy is. So the question essentially becomes “is it worth spending 100/100 to force the Terran to scan a couple more times” and as negligible as that cost is, I'm sure most Protoss players would say no. On the other hand, given the number of assumptions made at the highest level of Brood War play (see Zero vs Hydra on Outlier or Soulkey vs Leta on Jade), perhaps the question is actually “is it worth spending 100/100 on an X% chance that my opponent makes too many assumptions and I get a free win?” Against Flash this is probably always a yes, but as far as 'regular' players go, I'm not so sure.

It's the second day of playoff series between CJ Entus and KT Rolster. The teams have battled it out to the sixth set with CJ leading and hoping to end it here, rather than being forced to face Flash, for a second time, in the final set.

The map is Neo Chain Reaction, a two-player map that has seen its share of 'strategic plays,' and today will be no different. Hydra spawns at three o'clock, while his Protoss foe, Wooki, spawns at nine o'clock.

Hydra opens with an overpool, while Wooki places his pylon at his natural, unaware of the early aggression unfolding on the opposite side of the map.

Five drones are pulled off the mineral line and sent on a cross-map treck. The commentators shout out in surprise, as does the crowd. The drones carefully make their way over toward the Protoss base, following a plotted route and avoiding the obvious paths by which a probe-scout might travel. They wait outside Wooki's natural until the time is right to spring the trap.



Hydra's lings spawn and race over to the Protoss base, an overlord drifts into the natural and watches a probe plant a cannon, signalling that the time is right for the drones to move in and take it down. The crowd cheers and the CJ coach raises both arms and starts clapping, already confident that victory is assured.

The nexus continues to warp in at the natural, while the cannon is cancelled. With the cannons taken care of, the drones set their sights on the lone probe, chasing it around the natural. Wooki quickly pulls nine of his own workers to deal with the unfolding crisis. Another cannon is placed and the probes take a defensive stance, creating a barrier between the drones and the warping cannon. More probes are pulled as Wooki anticipates the impending arrival of zerglings.



The lings arrive at the natural and Wooki somehow manages to interpose his probes in between the lings, scattering them and blocking most from reaching the cannon. The probes manage to hold off the lings, taking a few losses along the way. Then, they drill through the drones, allowing the cannon to complete its warp-in. The crowd shrieks in excitement. Players on the KT bench are on the edges of their seats, hearts in mouths, as Wooki holds off the assault.

A shot of Wooki's main reveals no mining activity; most of the probes have been pulled to create a small line of defense at the natural, with only two probes spared to maintain a meagre income. Hydra still has five drones mining away at his main, he starts a hatchery at his natural and continues to spawn more zerglings. He then gathers a small number of lings and attempts a runby at the natural, but Wooki’s probe micro continues to impress, stopping the lings from breaking in. All the while the cannon blasts the lings away.

Wooki warps in a gateway, further narrowing the entrance, and allowing him to start massing some much-needed zealots.

Lings continue to gather at the front of Wooki's natural. A probe scout is dispatched to survey the situation, revealing more lings en route. Six lings attempt another attack, but Wooki pulls a few more probes, successfully blocking the pathway and killing two more lings.

Hydra is dogged and dispatches a detachment of another six lings as soon as possible, this time managing to breach the probe line. Four lings squeeze into the main and, in short order, the lings are free to tear down the naked nexus.

As the nexus falls, Hydra places a third hatch at twelve o'clock. More lings are made and rallied towards Wooki. A zealot spawns from the gateway and guards the narrow entrance. The lings from the main head into the mineral line at the natural and destroy many probes before finally being wiped out. At this point, Wooki's disadvantage is insurmountable, and it's only a matter of time before Hydra over powers him.



Two zealots move out and spot a pack of roving lings just outside the cannon's range. With the zealots out of the way, the zerglings run through the undefended gap, heading straight for the half-dead cannon.

Wooki concedes, knowing there's nothing more he can do.

Overview

This season's map pool has been fairly successful The standings (as of round three in parentheses, playoffs not included) look like this:

The PvT record is 38-30, 56% (14-7, 67%)
The TvZ record is 39-28, 58% (12-11, 52%)
The ZvP record is 43-37, 54% (19-8, 70%)

There have been about 70 games in each matchup, though ZvP got up to eighty. The win-percentages are more heavily imbalanced overall than they were last year, but are (with the possible exception of TvZ) within acceptable balance ranges and are also (including TvZ) within historical norms for the game.

I have been tracking game data based on a combination of calculated win-probabilities and actual results; for the entire set of games in the first three rounds, I have generated these balance numbers:

PvT: -0.79
TvZ: -0.49
ZvP: -1.03

What do they mean? Each number represents the likelihood that an average player from one race would upset a player of another race. All of these numbers are swung towards the "weaker" race in each matchup – because that is the direction most upsets will tend to happen. Even most bad Zergs are somewhat competent when it comes to ZvP; most bad Protoss players (and some good ones) die easily to hydra busts. Notice that these facts line up well with the win-percentages; Protoss players this season are more likely to win against Zerg than Zerg players are against Terran, and TvP is in the middle.

Of course, these overall numbers were not acquired in a vacuum, each map has its own strengths and weaknesses.

---

Electric Circuit has been much the most diverse map over the course of the season, with 48 Zerg players coming out, as well as 42 Terran, and (well behind, but still reasonable) 30 Protoss. The standings are as follows:

PvT 6-2, -1.00 (1-0, +0.15)
TvZ 11-11, -0.66 (7-2, +1.39)
ZvP 5-3, +0.01 (2-1, +0.01)

The map is most friendly to Zerg and looked far more so before the changes for round three. So, what did the changes mean? The 7-2 record does not indicate an awful lot, the major swing comes from Sea – in a horrific slump – beating two competent ZvT players, great and RorO. The other five wins come from Flash, Light[aLive], Fantasy, Bogus, and the under-rated (in TvZ at least) Canata. The round was less an anomaly than a correction from the previous two rounds.

Similarly, the map is "Terran favored" in TvP, despite the record, because every Protoss to come out has been favored statistically. ZvP is right about where it "should" be.

Electric Circuit is my current candidate for map of the year; it has an interesting design, it seems viable for all races and has seen diversity of play. For the moment, I’ve attributed the lower Protoss numbers to the presence of heavily Protoss-favored maps elsewhere in the map pool.

---

Sniper Ridge is the other map in the map pool to have shown evidence of really diverse play. The map has seen 50 Terran, 34 Protoss, and 30 Zergs in season one, with the following results:

PvT 10-10, -0.94 (4-4, -0.07)
TvZ 7-5, -1.02 (1-3, -0.50)
ZvP 6-2, -0.49 (2-0, +0.87)

As you can see, Sniper Ridge is somewhat Terran-favored, but it is also reasonable for Zerg, being fairly good for ZvT, relatively speaking. Moreover, Protoss players are not too far behind. The real difference on this map between results and "balance" comes in ZvP, but is relatively easily explained; almost every Zerg has been favored, while the two Protoss wins are from Stork beating ZerO. If we saw enough games on the map, I suspect the advantage would continue to swing back to Zerg, but we can only work with the dataset we have.

Sniper Ridge is another good candidate for retention in the map pool; the games have been exciting, the design is fairly unique, and the overall results have been close to what we would expect.

---

Unfortunately, not all maps are good. Ground Zero is the last map with anything resembling equality. The numbers sit at 49 Terran, 30 Zerg, and 27 Protoss. The results of the games have given the lie to that apparent parity, though:

PvT 6-7, -0.99 (2-1, -0.12) 81%
TvZ 12-4, +0.51 (2-2, +0.18)
ZvP 4-3, -0.64 (2-1, +0.45) 44%

Ground Zero is heavily Terran favored. Games throughout the season have made this painfully clear in TvZ: most memorable, perhaps, is Canata's win against Jaedong. Meanwhile, in PvT, things aren’t quite as obvious, Protoss is favored relatively heavily. On average, Protoss players have managed a losing record, and by "losing record" I mean Light[aLive] won a TvP.

ZvP, on the other hand, has been pretty much a disaster. BeSt beat Jaedong soundly to open the season for the map; Sun beat HoeJJa, Neo.G_Soulkey retrieved the situation somewhat for Zerg, picking up three of the four wins the bugs have managed, but two of those were busts made possible by sloppy play on the part of his opponents.

Ground Zero is not a very good map right now. However, it is a fairly standard map in terms of layout layout and the only terrible results are in TvZ; it is entirely possible that, either with small changes or player adaptation, it could well be viable.

---

With Jade, we reach the maps where you look at the statistics and have to say, "...lolwut?" The map has seen 53 Protoss, 52 Terran, and ... 15 Zerg. How do those numbers break down?

PvT 13-10, -0.40 (6-2, +0.20)
TvZ 4-1, +0.80 (2-0, +0.81)
ZvP 4-6, -1.65 (2-1, +0.71)

The best thing you can say about Jade is that it's a great PvT map. ZvP is heavily Protoss-favored, and ZvT is badly in favor of Terran. True, it is a standard-ish map, reminiscent in some ways of Byzantium; it's possible (sort of) that it could be fixed. For the moment, my verdict is "not worth it".

---

To be honest, I really want to like Chain Reaction, and I am not entirely sure why it has been such an unmitigated disaster in terms of results. Regardless, here are the numbers, and they speak for themselves: 63 Protoss, 49 Zerg, 10 Terran. That leaves us with:

PvT 2-1, -1.67 (1-0, +0.83)
TvZ 3-2, -0.85 (0-1, +0.21)
ZvP 13-15 -1.34 (9-3, +0.65)

This is a Protoss map through and through – one of two maps where Protoss has a positive record against Zerg. True, with more good Zergs playing on the changed map things are swinging back the other way, so it doesn't look quite so horrific. In truth, I do not have much confidence that the map, despite its ingenious design, can be balanced properly.

---

Last in this parade of unfortunate maps comes Outlier. It’s an unholy combination of Outsider and Requiem, with a blue paint job. Play on the map so far has boiled down to 60 Zerg, 50 Protoss, and 8 Terran. And they've given us:

PvT 1-0, 0.00 (0-0)
TvZ 2-5, -2.00 (0-3, -0.59)
ZvP 11-8, -1.05 (2-2, +0.03)

This is a decent ZvP map. I am not quite sure yet that it is actually terrible for TvZ – half of the "imbalance" here comes from Light[aLive] being unable to manage his wall-in, and Mind’s inability to micro valkyries. Really though, the fact that no one has even bothered to send Terrans on the map is telling. The map is bad.

---

Conclusion

As always, the biggest problem with drawing conclusions about the balance of any particular map is a lack of data. The most games in any single matchup on any single map is 28 (that’s ZvPs on Chain Reaction). The only map with a reasonably balanced number of games across the three matchups is Ground Zero – and even there ZvP lags far behind. It is all interrelated, of course; the presence of strong Protoss maps (for instance, Jade) limits how many will play on a weaker map.

Looking over the map pool and results, taking into consideration the two maps that have done best this season in terms of balance, I would suggest (as a true amateur here) five points for mapmakers to observe to help balance.

- Avoid sunken main bases. It seems unlikely to be a coincidence that the three worst maps – Chain Reaction, Outlier, and Jade - all have sunken mains. (Whether we also need to avoid the Twilight tileset I leave to your judgment.)
- Give the natural a choke point. It doesn’t have to be the interminable overhanging cliff of Lost Temple, but you don't want zerglings waltzing through at random intervals in the early game because there is simply too much ground to cover (or should I say, too much ground to zero?).
- Allow the third.
- Allow the third to be contested. By these two points I mean that the third base should be connected to the natural; but there should be some way for an opponent to attack. The textbook example is Fighting Spirit; as long as there is no contain on the natural bridge, the third can be taken freely, but it can be contested from outside on the ramp. On Electric Circuit, the "third" is accessible from the main, but is easily dropable and the temples can be destroyed. In contrast, on Ground Zero the third can be easily denied by a Terran without even attacking the natural; the shorter ranged Zerg and Protoss have more trouble.
- Break up the middle. This is the most difficult principle to successfully implement, I suppose. You have to have a middle region, but it needs to have features. Sniper Ridge has ridges; Electric Circuit has a relatively small middle and larger outside paths. Jade's middle is virtually featureless (the violation on one side); Chain Reaction doesn't have a "middle", just bridges (the violation on the other).