1.1 Spend Resources
1.2 Develop Reconnaissance
1.3 Know the Map
1.4 Leverage the Map
1.5 Check the Minimap
1.6 Attack, Attack, Attack!
1.7 Use Hotkeys
1.8 Use Patrol
1.9 Understand the AI
1.10 Expand Well
1.11 Stage Attacks
1.12 Own the Initiaitive
1.13 Kill Expansions
If one's resource reserves just keep rising, then one is not playing well. More to the point, one is not investing resources in structures or units that will win the game. Saving for the future often yields a future in which the spending is pointless; i.e., by the time one needs to spend resources, the enemy will be destroying the structures through which resources may be spent. At that point, throwing any amount of minerals or gas at the problem is not going to help.
If one is in the enviable position of harvesting faster than it is possible to spend, then build more structures. As a Terran, for example, having four or more barracks, three factories, three starports, etc. can allow one to utterly overwhelm a more resource-cautious enemy. Similarly, the Protoss may warp in more gateways and stargates, and the Zerg may create more hatcheries—or perhaps invest in more colonies, both spore and sunken.
The one caveat to this general rule is to keep enough of a mineral balance in the bank to support another expansion in case of trouble. This isn't as important for Terrans because they can simply fly an old command center to a new location, but it is vital to Zerg and Protoss survival in desperate situations. And of course, a Terran command center flies pretty slowly; a mineral reserve might be easier to maintain than a garrison of forces to protect it en route. But that is a decision best left to individual Magistrates.
I learned the value of good reconnaissance the hard way. In one particular game, I had an expansion base just south of the river on the "Bridge too Near" map, and my opponent parked a siege tank across the river and started shelling my base. I knew he had some missile turrets, and I was not entirely ready for this, but I improvised. I quickly brought my four guardians to bear on the tank, escorted by a squad of six mutalisks. A mere couple of seconds later, two battle cruisers showed up and used their Yamato cannons on two of my guardians. A third guardian was cut down by a pack of wraiths, and the fourth died while trying to take out the opposing missile towers attacking my mutalisks, which were trying to defend the guardians in the first place. In that situation, an utter lack of reconnaissance sealed my doom. I had no idea he had battle cruisers, and I certainly didn't think he had two of them. Their Yamato cannons were devastating in this encounter. My response to his attack was utterly hopeless due to insufficient reconnaissance.
Knowledge of the map is an extraordinary advantage—particularly in the early game. It allows safer, faster expansion, suggests which tactics might be more successful, dictates the starting positions of one's enemies, etc. Island maps (i.e., maps that feature player start locations separated by water or similar impassible terrain), for example, change completely the approach one must take to be successful. Terrans must concentrate on building wraiths, dropships, battlecruisers, goliaths and missile turrets; Protoss must concentrate on building scouts, shuttles, carriers and photon cannons; and of course, Zerg must concentrate on building mutalisks, upgrading overlord speed and transport ability, building guardians, etc., unless of course, one wishes to rely entirely on rather fragile Zerg nydus canals.
There are several maps for which it is utterly crucial to exploit certain features. For example, the "Bridge too Near" map can be dominated by any player who holds the center of the bridge. The computer AI is so stupid that any units told to attack will cross a few at a time, which will allow them to be eliminated comparitively easily. At the very least, it often requires a ridiculous expenditure of forces to cross a firmly fortified bridge on the ground. Knowing about the bridge gives one a fighting chance to cross it early or plan to pay later. Similarly, arranging one's forces such that opponents must attack from lower ground, traverse narrow paths, cross open "killing fields", etc. can be very useful.
I suspect the majority of players use the minimap only to gain a bird's-eye overview of where things are, but it can be far more useful. Get in the habit of checking it every few seconds. Many players stage their attacks, which is a good strategy if implemented properly; fortunately (or unfortunately as the case may be), they frequently err in estimating the sighting distance of their enemy. On more than one occasion, the minimap has revealed troops at the periphery of my base, allowing me to deal with them on my terms, not theirs.
Of the various Starcraft "sins", I commit that of defensive gluttony far too often. Attacking in the early game is not merely useful, it is imperative! Leaving a player alone—especially a Terran player—for too long allows all manner of devious schemes and devices to be brought to bear. Such early attacks do not need to be overwhelming in force; they need merely to be effective in destroying some of his stuff and distracting his mind. The point is not necessarily to win immediately, although this sometimes happens, but rather to disrupt his plans and force him to overbuild defenses. This allows the clever player to cut back on his own defenses somewhat and pursue offense, a tactic that will frequently result in victory.
With regard to computers, 'mouse' refers to a wonderful interface tool, but using such a mouse is orders of magnitude slower than using the keyboard. For example, when using a Protoss templar, one must click the templar to select him, click the psionic storm button, then click the destination for the storm, a process which can take several seconds, during which the templar may well be killed. Using hotkeys, however, one may press a number key to select the templar, press 'T' to initiate the storm, and use the mouse to select a destination, all of which may be accomplished in less than half a second. With practice, one should be able to manipulate mouse and keyboard at the same time, shaving that down even further to a quarter of a second or less. This kind of fast response time can make the difference between one's forces remaining largely intact or winning a sadly Pyrrhic victory, if emerging victorious at all.
The following chart illustrates these two principles. For each race, I have listed the types of units that I bind (generally) to the various keys. This is not intended to provide the "best" arrangement, but it has worked pretty well for me.
| Key | Group | Terran | Zerg | Protoss |
| 1 | Light Ground | Marines, Firebats | Zerglings | Zealots |
| 2 | Light Ground and/or Light Ground Support | Marines, Firebats, Medics | Zerglings | Zealots, Dark Templar |
| 3 | Heavy Ground | Goliaths, Vultures, Siege Tanks | Hydralisks, Ultralisks | Dragoons, Reavers |
| 4 | Heavy Ground | Siege Tanks | Hydralisks, Lurkers, Ultralisks | Reavers, Archons |
| 5 | Special Units | Ghosts, Science Vessels | Queens, Defilers | Templar, Dark Archons |
| 6 | Light Air | Wraiths | Mutalisks | Scouts |
| 7 | Light Air and/or Light Air Support | Wraiths, Valkyries | Mutalisks, Devourers | Scouts, Corsairs |
| 8 | Heavy Air | Battlecruisers | Guardians | Carriers |
| 9 | Transport | Dropships | Overlords | Shuttles |
| 0 | Quick Access | Comsat(s) | (depends) | (depends) |
The purpose of the quick access key (viz., 'Ctrl+0') is to select whatever unit or building I need frequently. For Terrans I always leave it bound to a comsat for quick scanning. With Zerg or Protoss, I generally bind the key to whatever unit is momentarily important. For example, I often use a lone Zerg queen to scout the map and parasite various enemy units; before leaving base, I bind the quick access key to this unit. With Protoss, I will frequently bind this key to a shuttle carrying archons/reavers to allow fast drops and lifts. These key bindings are somewhat more effective with the Protoss because the Protoss generally have fewer units to send into battle in the first place. Two squads of zealots are, after all, very deadly. When playing Zerg, however, I will frequently have forty or more zerglings to act as cannon fodder. In such a case, I bind the keys as shown above, then use the ctrl + click shortcut to select additional groups of twelve zerglings as needed to bring them to the front lines. As the foremost squads of zergling die, I reassign hotkeys as appropriate.
When preparing for an all-out assault, I often leave my troops in tightly packed groups in my home base, but this is a very bad habit. On more than one occasion, I have had entire squads of marines, wraiths, goliaths, etc. rendered at least temporarily useless, if not slaughtered outright, by a defiler's plague, a templar's psionic storm, a ghost's nuke, etc. Rather than mass troops in tight groups, put them on patrol. Even if their patrol is confined to a small radius around the periphery of the base, it will likely save many of them from such attacks and may significantly improve the level of reconnaissance for avoiding such stealthy assaults in the first place—especially if one checks the minimap frequently.
There is no polite way to say what needs to be said: the computer AI in Starcraft is dumb as a rock. As a programmer, I certainly hope the problems being addressed are more difficult than I understand them to be; otherwise, Blizzard should be embarrassed at the stupidity of the AI. More to the point, there are four serious difficulties with the AI that must be grasped in order to work around them and/or leverage them effectively against one's opponent.
A clever player can, of course, turn these limitations into advantages. Placing siege tanks on the high ground near a Protoss base can make it practically impossible for the Executor to prevent his zealots from running blithely off to certain death. This too is a lesson I learned the hard way. Similarly, a small force at a choke point can easily destroy a larger force as the enemy units pile up or wander around stupidly due to lack of room to move.
To Blizzard's credit, some of the problems involved in such AI algorithms are very difficult. But I have played quite a few other games (e.g., X-Com Apocalypse, Total Annihilation, Battle Zone, etc.) that do not have such glaring problems. If there is a Starcraft II, I hope that they address the problems with the AI. For the time being, however, one must be aware of these issues and learn to work around them.
Although I elsewhere treat this issue specifically for the various races, there are some general comments relevant across the board. Expanding successfully involves at least three factors: (1) when to expand, (2) where to expand, and (3) how to expand. There are other considerations, of course, but these seem to be the three key factors in roughly the order, I suspect, in which they occur to us. I believe, however, that these three factors should be considered in a different order. More specifically, they should be treated in the order suggested by the "good-making properties" of a successful act of expansion.
Since the second consideration seems more primal, I shall deal with it first. The where-question seems more primal insofar as one cannot know the how or the when until one knows the where. For example, expanding to an island will add transports to the how and likely push back the when until the tech-tree can be climbed. Knowing where to expand is a matter of knowing where the resources are in relation to the positions of one's allies and enemies. Knowing the location of resources is merely a matter of having good reconnaissance, so scout the map as soon as possible. Regarding the selection of a location, it generally simplifies matters to expand near one's own base; e.g., providing forces is easier and it is often possible to share them between bases. It also simplifies matters to expand near one's allies (if any) or at least along the sum of the vectors from the enemy bases toward one's own; i.e., behind one's base where 'behind' is evaluated in terms of the enemy position(s). With maps that have few resources or only resources in the starting locations, I suggest choosing a site as close to home and as far out of harm's way (i.e., as far from the enemy) as possible. I know this is a rather general suggestion, but it is nevertheless both valid and easily overlooked in the heat of the moment.
There still remain certain issues regarding where to expand, however, not the least of which is defensibility. It may make more sense, for example, to choose a site with fewer mineral patches that is more easily defended than another site with many mineral patches surrounded by easily accessible high ground. Personally, I find this balance depends upon which race I'm playing. With Terrans, for example, being able to place siege tanks on high ground sans ramps is a big plus; similarly, being able to close off a choke point with bunkers is very effective. Generally, I think it best to choose a site with as few lines of access as possible; it is usually difficult to defend one's expansions, and any features of the site itself which may ameliorate the problem should be taken into consideration.
Moving on to the third consideration, the question of how to expand will be influenced greatly by where and when one chooses to expand. When playing Zerg against Terran or Protoss, for example, I will often expand first with nothing more than a drone, and the reason is simple: the Zerg can expand before those two races can project any serious force across the map. And by the time they gain the ability, my expansion will have many drones and will already be producing offensive units. As Protoss or Terran, however—particularly against Zerg—it is utterly crucial to take and hold one's expansions. This is fundamentally opposed, of course, to the desire to expand earlier rather than later, but loss of a crucial expansion can determine the course of the game.
Thus, in deciding what troops to send, there are effectively four options: (1) send none, (2) send minimal forces, (3) send moderate forces or (4) send heavy forces. Option (1) is plausible if and only if one can guarantee that the enemy will be too busy to attack, which is a guarantee almost never forthcoming. Option (2) is worthless; if the enemy attacks with utterly minimal forces, then the harvesters should be able to protect themselves (particularly if they are supported by sunken colonies or photon cannons). This leaves us with options (3) or (4), sending moderate to heavy forces. Selecting between the two depends upon the where and the when of expansion. If the where is located close enough to one's own base, then moderate forces will usually suffice. If the where is located between between the enemy base and one's own base, then send all available forces and start building more at home. Try to avoid a first expansion far from home; if unavoidable, however, then send at least moderate forces (or more if expanding close to the enemy) and hope for good luck. Regarding the when of expansion, the longer one waits, the more forces one had better send; otherwise, holding the expansion will be more difficult when the inevitable attack comes.
Finally, backing up to the first consideration, it is always advantageous to expand earlier rather than later, but this must be tempered in light of the prevailing strategy as well as the other two factors. For example, if one focuses too heavily on expanding in the early game, one may easily lose if one's enemies focus more upon building forces.
There is little more frustrating than watching the perfect battle group ripped to shreds because of a lack of cohesion, yet it happens all the time. Zerglings or zealots will die by the hundreds, for example, when arriving at Terran defenses in single file. Be sure to move troops to a staging area before attacking, then attack all at once. The staging area should be far enough from the objective to be safely out of sight yet near enough to allow a massed attack.
There are two postures, essentially, that one may adopt regarding the events that unfold during a game. One may (1) direct the course of the game or (2) react to it. The former posture, save for (hopefully rare) fatal miscalculations, wins, whereas the latter loses. To be more specific, the person who takes the initiative and makes things happen is always in a better position than those who are forced to react.
This does not mean that one must always be throwing one's troops wantonly at the enemy. Occasional attacks are generally sufficient. At the very least, they force one's enemy to overbuild defenses, repair damage and perhaps most importantly, they make one's enemy react rather than initiate. Harassment by small squads of mutalisks, EMP shock waves followed by a quick nuke, scout hit-and-run raids, etc. all throw one's opponent off-guard, forcing him to react rather than direct the game. And ultimately, he who controls the events of the game most often wins it.
The tactic I have found most useful in winning games is, sadly in my estimation, strangling the enemy economically. Too often, players will leave an expansion unharmed in favor of going after their opponent's main base. This has the dubious advantage of allowing one to destroy some key structures, but this is useful only to a limited degree.
It is far more useful to kill the opponent's expansion. Why? Because it will (1) impede his harvesting, (2) likely force him to minimize his defenses at his main base and (3) build units to replace those being destroyed at the same time his "income" is being reduced. Further, because the enemy is otherwise occupied, one can expand with relative impunity during the attack. In short, pursuing this tactic generally results in crippling the enemy's economy while one's own economy grows safely—a sure recipe for victory.