mizzao / bot.land Goto Github PK

if equipped with regen.

Higher the distance is from the centroid, the more likelihood we would move toward it.

Might help with #14.

If melee is not equipped, it will keep going around at 2x DPS (flash zapper).

But if it is, it should bash adjacent units a bit, cloak and all.

Will help with both side sneaks, as well as dealing damage that can't be reflected while tanks keep enemies still.

On defense:

• waypoint movement code
• do not wait for sensors to move (but activate them if available)
• micro at a smaller evade distance (min of 5)

Like spartans... band together to reduce surface area. Could be useful against uncoordinated/mass melee enemies.

Phase 1: dedicated reflect 3 / thrusters 3 miners lured out the tanks.

Phase 2: 2 of the same, plus 2 reflect 3 / melee 2 teleporters. Once rangers engaged the miners were pretty useless.

Phase 3: laser 3 / reflect 3. These guys were dumb and shot chips, but most of the damage was self-inflicted via reflect.

Ideas:

• lasers on defense?
• if tanks are damaged and beyond halfway point, pull back.

Not sure why this happened, but everyone just stopped moving or shooting.

Happened in 1 out of 50 games...

Tanks should have reflect 3, no shield.
Behind units should cast shield on tanks, as big as possible.

Need to figure out a heuristic to get shields cast on tanks.

Is canceling the call to zap.

List of things to test:
#18
#21

It is a lot more resistant to reflection (DOT instead of up front), and can shoot while backing up

Just like missile micro, it should run more the closer the enemy gets, not wait around to get pummeled.

Also, try running it on speed 3 instead of 1.

• Reflect / shields (especially team activations) are correct
• Movement / patrol is correct
• Artillery sneaks are dealt with
• Seeing new units and following is correct
• Mining team
• Lure and sneak

If missile kiters outnumber enemies (3>1, 4>1, 5>2) they should just fire instead of running, and taking a lot of damage in the corners.

Has repair, as well as a strong melee attack to break up anyone who gets near.

Requires code to coordinate movement.

Reflect 3 + cloak 2 + melee 2.

https://puu.sh/ErLzo.gif

Regen 3, missile 2, thruster 2? or maybe TP

Gets around the cloak debuff by setting people on fire. Could be part of laser squad (#33).

Currently the n>5 check approximately achieves this, but we should be able to go straight for lasers without going around when reflection is active.

include movement code for when no enemies are in sight

Ideas:

• light them on fire with inferno zap or laser
• always attack to an enemy location that is within 3 columns of the right
• defenders should bump cloak locations (require getting on the space of enemy target): this also is unaffected by the fire zappers dying.
• Missile micro should not back up when they uncloak

It doesn't use thrusters that much with sensors, so no need to have thrusters 3.

And it doesn't micro that much anyway.

It will run across a whole column and back.

If there is an enemy in the row we're moving to, don't move there. Just move the other way, or back, or shoot. (This helps with offense as well).

This code may inadvertently help with artillery dodging, but that is only at 5 range, so we can factor it out to a separate section.

It seems that when one of the miners goes ahead and leeeroy jenkins everyone, there is not a good distribution of mines when enemies charge. We should help them to coordinate better.

Mine layers also don't lay enough mines when enemies are further out but not that close.

Ideas

• advancing in a diagonal should stay that way
• random up/down movements actually don't help for miners...coordinate?

Would be interesting if miners could keep formation, at least for the initial assault.

Melee: 300 / 450 / 550
Shield: 150 / 300 / 400

Only worth it if we have shield 3 and they have melee 1. Otherwise it just loses a turn.

Use state for this.

By the time attacking artillery that start from the back make it up to the front, defenders have backed up to the CPU, and just let the chips die.

Maybe this is OK though. The chips buy us a turn.

When:

• no enemies in sight, but allies are in sight, or
• enemies are in sight but it's already shielded.

Of course, generally try to self-cast shield first.

Loadout:

• zapper 3
• thrusters 2/3
• reflect 1/2

Attempts to keep the closest enemy at (1,1). On a diagonal so avoids lasers. Moves too quickly for artillery. Can only be melee'd with charge available. Reflects missiles.

Could complement missile kiters well. Worth giving a shot.

It jumbles around generally being useless.

https://stackoverflow.com/q/58278709/586086

With evadeEnemy. This will allow more mines to explode!

tryFireArtillery() will shoot a chip or CPU even if we are getting bashed by a bot.

• Initial coordinated mine laying "carpet" (can prototype #44) until reaching back wall
• Configurable VS_LASERS to either stay on row/col or on a diagonal
• Configurable EVADE_DISTANCE: how close to stay to enemies while laying mines
• Support/try various defenses:
• reflect 3 / cloak 3 (shadow minelayer)
• reflect 3 / cloak 1 / thrusters 2
• thrusters 3 + shield/reflect (fast minelayer)
• teleport + shield/reflect (evasive minelayer)
• co-shielding of other minelayers
• regen 2 / shield 2 / thrusters 2 (roach minelayer)
• can also use regen in any combination above.

Generally very flexible because 6 slots are open!!

• Do not shoot chips when out of EWMA range; this has caused many groups of bots to get ded.
• Bots shouldn't be forced to move toward EWMA, the ones that are actively microing should "pull" the ones that are not occupied. So move the code toward the end rather than the beginning.

• Allow sneaktillery to kill chips from the side t(hen, it may be able to continue along undetected.)
• Allow it to move around blocking units in the way
• If teleport is enabled, jump forward when trying to advance.
• better support for 2 sneaktillery at once

Handle other equipment:

• cloak
• reflect
• teleport

Used on attack.

• looks for friendly bots to repair. broadcasts its location

• if a bot sees repair bot nearby and no enemies, it waits to get repaired

• mainly for supporting micro or artillery units.

Regen 3, TP 2, thrusters 2

The whole point of this bot is to keep the enemy team occupied for as long as possible. For hard to crack, coordinated defenses that need a distraction so that a crack team can take out the CPU.

Happens going up/down. Not sure why.

The failure mode here is just like the CPU, with artillery shooting.

Although, chips have less health and are more useful to kill. But in this case the bot was about to die and we were close to winning :-/

Check that we aren't waiting on sensors too much in movement code.

On side walls:

• When evading along bottom wall randomly step upward. This is because enemies follow horizontally and won't step on the mines, etc. This allows us to get out and around again.

On rear wall:

• Bots get stuck when there are more than 2 enemies, but should just run forward and try to go around.

Update the zap kite bot: note that when using thrusters with zapper, damage is dealt every time the bot moves. So effective DPS can be twice as high (!!) as with normal zapper. But enemy melee units can only hit you half the time.

Why melee a single unit for 550 when you can splash in all directions for 700, and diagonally for 350? Then just cloak when zap runs out, and do it again!

test in sandbox.

Not sure which one of these is authoritative, but the ones that aren't should be deleted.

• .vscode/settings.json
• .editorconfig
• .prettierrc.json
• tslint.json

They get stuck when there are more than 2 enemies, but should just run forward.

Too stringent and we risk getting a run-around and the CPU killed. We can probably clear if either condition is satisfied:

• <= 3 spaces away with sensors on
• <= 1 space away with no sensors on

Also, bots should not follow enemies more than halfway across the map (just delete enemy target location). This will both cause them to be less sensitive to mines, and get back to the CPU faster in case of lures. Because it will return and not just sit at the edge, it won't be vulnerable to artillery micro.

Via @freaktechnik: https://gist.github.com/freaktechnik/e1f42c309555a48c8e0e93f3298c6e84