HayDay Bot Python
I've create this for two reasons:
- The two other packages on GitHub for HayDay are viruses (bad ones at that). Please reset your computer if you've ran these programs at all.
- It's a game that preys on your patience, so it's fun to automate.
Finally, I will not provide the full automation script as it's against terms of service. I will however, show the full work to get there. Also thanks to @claritycoders for inspiration with the fishing bot.
As per any automation problem we should break down the exercise:
- Automatically detect the soil
- Plant Wheat (most profitable)
- Sell during growth (downtime)
- Harvest
- Repeat
As we determined above, the most important will be to detect the area the bot can use to grow crops.
To do this we can do a couple of things:
-
Color
How to do it:
class SoilDetector
def __init__(self):
template_path = os.path.join(os.path.dirname(__file__), 'templates', 'soil.JPG')
self.template = cv2.imread(template_path)
self.template_color = np.mean(self.template, axis=(0,1))
self.color_threshold = 15
def detect(self, screen):
diff = np.abs(screen - self.template_color)
current_mask = (np.mean(diff, axis=2) < self.color_threshold).astype(np.uint8) * 255
-
Thresholding/Masking/Countours
contours, _ = cv2.findContours(mask, cv2.RETR_EXTERNAL, cv2.CHAIN_APPROX_SIMPLE)
largest = max(contours, key=cv2.contourArea)
epsilon = 0.02 * cv2.arcLength(largest, True)
- Shape matching
Then we can use cv2 built-in approx = cv2.approxPolyDP(largest, epsilon, True)
This is a complex algorithm but essentially will adapt the rectangle to fit the perspective of the game.
Having a uneven amount of tiles is essential (see screenshot above, 21 tiles works because the center will be accurate, while 24 tiles will not as the center will be ambiguous) but could be corrected easily by using square closest to viewport.
But I'm lazy:
M = cv2.moments(contour)
if M["m00"] != 0:
cx = int(M["m10"] / M["m00"])
cy = int(M["m01"] / M["m00"])
M00 represents the total area M10 represents the sum of all x-coordinates M01 represents the sum of all y-coordinates
Dividing first order moments by zero order moment gives the average (center) position
Well now that we have a reliable middle point you know what time it is... Offsets... This is horrible code example but hey, if it works...
# Planting phase
while True:
if not self.paused:
pag.click(cx, cy)
while self.paused: # Wait if paused
time.sleep(0.1)
time.sleep(3)
if not self.paused:
wheat_x_offset=-100
wheat_y_offset=-150
pag.click((cx+(wheat_x_offset)), (cy+(wheat_y_offset))) # Wheat
while self.paused:
time.sleep(0.1)
time.sleep(1)
if not self.paused:
pag.mouseDown() # START DRAGGING OVER COUNTOURS
contour_points = contour.reshape(-1, 2)
for point in contour_points:
if self.paused:
pag.mouseUp() # Release mouse if paused
while self.paused:
time.sleep(0.1)
pag.mouseDown() # Resume mouse down when unpaused
x, y = point
pag.moveTo(cx+35,cy-20, duration=1) # Make sure we drag low enough for planting to begin
steps = 50
for i in range(steps):
if self.paused:
break
ix = cx + (x - cx) * (i/steps)
iy = cy + (y - cy) * (i/steps)
jiggle_x, jiggle_y = self.add_jiggle(int(ix), int(iy))
pag.moveTo(jiggle_x, jiggle_y, duration=0.1)
if not self.paused:
pag.moveTo(int(x), int(y), duration=3)
time.sleep(0.5)
pag.moveTo(cx,cy) # RELEASE AT MIDDLE SAFE POINT
pag.mouseUp()
Well we just reverse the sequence above and use a slightly different offset for the tool:
print("Time to harvest!")
if not self.paused:
pag.click(cx, cy) # Click middle of field again
time.sleep(1) # Wait for tool to appear
pag.moveTo(cx-110, cy-10, duration= 1) # Move mouse to harvest tool offset
# Then same as planting movements...
def add_jiggle(self, x, y, amplitude=45):
jiggle_x = x + np.random.randint(-amplitude, amplitude)
jiggle_y = y + np.random.randint(-amplitude, amplitude)*2 #MORE HORIZONTAL MOVEMENTS THAN VERTICAL
return jiggle_x, jiggle_y
This helps get the whole field 98% of the time.
print("Starting 2 minute timer until harvest...")
start_time = time.time()
elapsed_time = 0
wheat_harvest = 110 # Seconds - Time to plant
has_sold = False
while elapsed_time < wheat_harvest: # 2 minutes WHEAT --offset 10 secs
if not self.paused:
elapsed_time = time.time() - start_time
if not has_sold and not self.paused:
go_sell()
sell()
close()
has_sold=True
pag.moveTo(cx, cy)
# GO SELL DURING GROW
time.sleep(0.1)
Then the go_sell, sell, close are simple automation scripts that use the shop to unload storage.
The main thing was to make a loop that would replant directly (so we still have wheat, then sell as much as possible, or what's left)
Here is a preview in x16 speed.
Video.sans.titre.Realisee.avec.Clipchamp.3.1.mp4
Hourly estimate coins: 2 000
pip install pytautogui mouseinfo opencv-python mss
Mouseinfo is super useful to determine exact coords create file: mpos.py:
import mouseinfo
mouseinfo.MouseInfoWindow()
MSS for the screen capture loop.
import mss
import os, time
from datetime import datetime
def capture_loop(x=100, y=100, w=500, h=300):
if not os.path.exists("shots"): os.makedirs("shots")
with mss.mss() as sct:
monitor = {"top": y, "left": x, "width": w, "height": h}
try:
while True:
sct.shot(output=f"shots/shot_{datetime.now().strftime('%H%M%S')}.png", **monitor)
files = sorted([f for f in os.listdir("shots") if f.endswith('.png')])
if len(files) > 5:
os.remove(f"shots/{files[0]}")
time.sleep(5)
except KeyboardInterrupt:
print("Stopped")
if __name__ == "__main__":
capture_loop()
And PyAutoGui for automations.
import pyautogui as pag
pag.click(x,y)
You can apply the same logic to remotely any game, triggerbot is a good example, while you shouldn't do it in multi-player is a fun project to code in a day.
Also can make neural network implementations based on visuals + controller module (possible actions) Your main issues will be 1. Working with PyTorch or TF 2. Finding a way to reward based on game state.
Neural Networks Introduction: Snake AI, How machines outperform us consistently
I also plan to release a couple of code pieces for Minecraft which is especially fun as you can interface with it directly in JavaScript using existing libraries. And play with your monster creation!
Minecraft Simple Neural Network
Just for fun :D This software is provided for educational and research purposes only. The use of this software to automate or modify behavior in third-party applications may violate their terms of service. The authors assume no responsibility for any misuse or legal consequences. It is intended purely for educational purposes and should only be done in secluded envrinoments.