This page will explain the most basic principles of Prismata strategy and tactics. Once you read and absorb this information, I'd say that you officially know "how to play Prismata". Do note, however, that I'm assuming you have some minimal level of experience with actual play - you should at least play a game or two of base set only against Adept Bot before you come here. Finally, to see the panels for the units I mention, go to prismata.net/units and use your browser's text search functionality. Depending on the browser, you can do this by pressing Ctrl+F and then typing text.
This is the most important tip on the list. Normal defenders, such as Wall, will heal back to full health immediately after the defense phase if they take partial damage. This ability is called absorb and is extremely important to take advantage of. To see why, consider the following: if you defend for 5 and your opponent attacks for 5, you lose all of your defenders and nothing else. If you buy 1 more point of defense and defend for 6 instead, you can lose all of your defenders except the Wall, put the last 2 damage there, and at the start of your next turn it has 3 health again. Thus by buying just 1 more hp of defense you gained yourself 3, because you were able to waste ("absorb") 2 enemy damage on your Wall. For this reason, you should be willing to go to almost any extent to avoid being breached, even for 0, because getting to absorb enemy damage is just so efficient. If you have to block with 3 of your Drones to save your Wall, then that is what you must do. Always absorb. (307th, a top player, has more to say about why absorb is so important here.)
An extension of the above-principle. Prismata is a very well-balanced game, so most of a unit's upsides and downsides are factored into its cost. Most of the time, therefore, you don't just have to look at 'which unit is better / has more advantages', you have to look at 'which unit's advantages are worth extra in this situation'. However, one thing that is deliberately not factored into defenders' costs is the power of absorb. Doomed Wall is of a similar cost-to-HP ratio to Wall (about 3% better), but also allows you to absorb 1 more damage for free each turn. For this reason it is very important to keep at least one Doomed Wall around to supplement your defense so that you are maximizing your absorb.
Sometimes it is possible to effectively obtain free damage by attacking for a number that prevents the opponent from getting their full absorb. For example, against an opponent with nothing but 3 Walls, attacking for 3 or 6 causes them to lose the same amount of defense as if you attacked for 5 or 8. In this case, you are exploiting for 2 by denying 2 absorb to your opponent. Attacking for 4 or 7 is an exploit for 1. If you have offensive finesse (the ability to control exactly how much damage you deal), look for opportunities to exploit your opponent's defense.
But what if instead of three Walls you had two Walls and three Engineers for the same total health? Against that defense, the exploit doesn't work anymore - if they attack for 3 you just lose one Engineer instead of an entire Wall. Having a couple of low-health defenders mixed in, called defensive granularity, prevents your opponent from exploiting your defense. The best way to achieve this is by making sure you always have a couple of Engineers or other 1-HP defenders available, and replacing them ahead of time if you can tell you're going to lose them. Defensive granularity is important because it stops your opponent from effectively obtaining a few free points of damage through exploits. If you find yourself without any Engineers on the board and you realize you're going to be exploited, what you can do is hold back a Drone to block. If this allows you to absorb one more damage in addition to blocking one itself, it's a strict improvement over spending green on a Forcefield, and a massive improvement over spending 5 entire gold and a red on a Rhino.
Even if you cannot exploit the opponent's defense yet, offensive finesse can still be important. For example, if you have 5 damage against a defense consisting of 2 Walls and 2 Engineers, you should consider clicking a Militia so that you kill both Engineers instead of one Wall. That way you will be able to exploit them next turn.
Since we've established that absorb is extremely powerful and you want to get as much of it as you can, it should follow that you also want to stop your opponent from getting absorb value. There are a couple ways to do this. One is to use long-buildtime units like Tarsier as your first attackers - the longer you don't actually deal any damage for, the longer it will be before the opponent can start absorbing. Another way is with units that cost attack or have abilities that cost attack. Take Lancetooth for example. If you have 2 Tarsiers and the opponent buys a Wall to absorb the damage, Lancetooth becomes an insanely good purchase because they already have their Wall ready and waiting to absorb, so the two damage it costs is worth literally 0. That makes it a lot cheaper. Lancetooth is actually balanced around this idea - it's a pretty bad purchase if you're paying 2 real attack for it.
Units that pay a cost for their attack (like Rhino which loses stamina each time) are much better off not attacking if they would not actually do any damage. For example, if your opponent has a Wall and you have no attack except two Rhinos, it would be a very foolish move to attack with the Rhinos because this effectively wastes 2 damage that could have been dealt later. Don't attack if your attack would be wasted. In fact even if you have 3 Rhinos you shouldn't attack into a Wall. This is especially relevant against blockers that only have 1 lifespan left, since they will die at the start of the next turn whether they take damage or not. For example, if your opponent has a dying Chieftan on defense as well as an Energy Matrix, then the first 11 damage goes to waste, so you'd better wait until next turn to fire your Scorchillas (unless you have close to 11 attack even without them).
You should almost never buy prompt defense before the turn you actually need it. You might be worried that if you don't start building defense ahead of time you'll get taken by surprise and find yourself unable to defend, but this fear is almost never justified. You'd be amazed how much defense you can pull out of nowhere if you really need to in Prismata. So next time you're thinking of buying a Wall that you don't need yet, try buying a Steelsplitter instead. The Steelsplitter will start destroying enemy defenders immediately, forcing them to buy additional defenders instead of more attack, leaving you free to do the opposite. You always want to buy units so that they get put to use as soon as they finish construction. To be clear, this doesn't mean defenders without prompt are bad. Non-prompt defenders have other advantages over prompt ones to compensate for needing to be bought one-turn in advance, like more hp (Infusion Grid vs Wall) or a lower cost (Perforator vs Rhino), and actually tend to be really good; but you still don't want to buy them so that they start blocking before you need defense.
Ramping means buying permanent attackers like Tarsier. Pressure means forcing your opponent to buy defense, but not ramping (like Rhino). These units have a tradeoff - one is permanent, but the other attacks sooner. How do you decide which one to do? The key realization is that when we have to buy defense, we exhaust the most efficient source (often Walls) first and only once we run out of blue do we turn to less efficient things like Rhinos for defense or holding back Drones. Pressure units are balanced so that they're weak when they force out efficient defense but strong when they force out inefficient defense. Therefore, ramp if you're nowhere near your opponent's breaking point, and once you get close, start making pressure plays.
Be careful of buying too many tech buildings if you lack the Drones to make use of them. If you buy two Blastforges, two Animuses, and two Conduits with just 15 Drones, you won't have nearly enough gold to spend all of your tech each turn, and you will end up wasting lots of resources as a result. Wasting resources is the path to defeat. 307th came up with a rule of thumb that I think is pretty accurate: in general, one Conduit requires 3 Drones to spend efficiently, one Blastforge requires 5, and one Animus 8. (The Conduit rule is because while Forcefield only costs 1 gold upfront, it also consumes a Drone, so you need more than 1 Drone per Conduit if you want to continue buying Forcefields.) Of course, these numbers are subject to change depending on the random set. In a set with Immolite for example, an Animus only requires 6 Drones to support.
The random set units in Prismata have been deliberately made more powerful than the base set units to ensure that each game of Prismata is significantly different from the last. Therefore, base your strategy on the random set units. Don't make an army of Tarsiers in a set with a strong random set attacker such as Hellhound or Cynestra. If you're faced with a random set unit you've never seen before and you don't feel you have a good idea of how strong it is, it's a pretty safe assumption that it's better than base set units that fill the same role.
In most games of Prismata, after you pick a strategy, you should get the amount of Drones you'll need for it before you start buying attackers, and then stop buying Drones once you start buying attackers. The reason is that attackers are straight-up more efficient than Drones in this game (For a proof, see my article on the math behind Prismata). The only reason you buy Drones at first is because the first couple points of damage can be absorbed, meaning that attack doesn't actually start killing enemy units until you have a good bit of it. But the max amount of damage your enemy can absorb is a constant (2 if they're using Wall), not a percentage of your total. Therefore you should get a decent economy first so that having your first couple of attackers nullified hurts less. To illustrate this, if you go for a turn 1 Blastforge into Steelsplitters, you can make 1 Steelsplitter per turn, meaning you lose two entire turns of production to absorb. If you waited until turn 3 and then got 2 Blastforges for 2 Steelsplitters a turn, you only lose one turn of production to absorb. The advantage of waiting to buy attackers is clear. But you don't want to wait too long, because once you're over that absorb barrier and you've started to kill enemy units, attackers are way better purchases than Drones, so you need to find the point where the increased efficiency of attackers starts to just barely outweigh the drawback of being absorbed at first. Even weak attackers are better than Drones if they don't get absorbed. Don't buy two Drones in the late-game if you could buy a Gauss Cannon instead. If you only have 3 gold, consider spending it on an Engineer instead of a Drone. The Engineer is cheaper and will block next turn, allowing you to devote more resources to buying attack instead of defense, not to mention the benefits of granularity.
If absorb is the main reason to buy Drones before attackers, it should follow that the size of the biggest absorber in the set largely determines how many drones you want to get before you start making attackers. In a set where Wall is the biggest absorber, 12-15 Drones is usually enough. In a set with a big absorber like Energy Matrix, you'll normally want around 20 Drones. Try getting a third Engineer on turn 2 in sets with big absorbers so that you can make Drones faster.
Attacker-defenders that have lifespan are almost always better off blocking then attacking on their last turn. The reason is that since they almost always have more HP than attack, you are choosing between a certain amount of attack or a greater amount of defense. For example, Doomed Mech can either give you 2 attack or 5 defense on its last turn. Of course, you should still check before doing this if the Doomed Mech's attack would breach or something.
Overdefending is generally a very bad thing. But if you have attackers that are blocking (Borehole Patroller is the best example), you should buy enough defense so that you don't have to lose them if possible. Borehole Patroller is worth more much as an attacker than as 2 defense. (Another way to look at it is: buying 2 more defense to save it is cheaper than replacing the attack it provides.) So if you have 5 Borehole Patrollers and 10 damage is incoming, you need 21 defense, because you need to be able to not use the Patrollers and still defend. Remember that not every unit that says 'Blocker' on the top is actually meant for defending. Oh, one exception: Xeno Guardian has so much health for its attack that it's usually cheaper to just replace the attack it provides than to buy 4 more defense to save it.
Chill is extremely effective against blockers with lifespan, because if you freeze a unit that only has 1 lifespan left, the unit dies without ever being able to block, meaning you essentially cheated your opponent out of all that free defense. It's such a hard counter that it's often correct to just attack with lifespan units on their last turn if your opponent could freeze them, or sacrifice them early so that they never make it to 1 lifespan (which doesn't work if the chill is from a unit like Shiver Yeti, as they can just freeze your Doomed Mech continuously).
You will often find yourself overdefending if you always defend against the indicated attack from your opponent. Sometimes your opponent is defending with attackers and they are going to have to lose some this turn, which the attack indicator does not account for. For example, if your opponent is defending with a Wall and two Rhinos against your 6 damage, they have to lose at least one Rhino no matter what, so you need 1 less defense than you think. There's also the issue that the attack indicator assumes the opponent has enough resources to click all their stuff, so if they have a Zemora Voidbringer but not enough green to fire it, you can defend for 8 less than the indicator says. (It also assumes that all their chill will be put to use without any overflow.)
Alright, that's all I consider to be the basics. If you absorbed all of this information (like my sicku pun?), you should be able to beat at least Adept Bot every time now.