The source needs to go directly to +12V. Connect that 10K to the source of Q2
Look carefully at the data sheet of your FET and your drawing. You have your FET upside-down with the Source connected to +LOAD. The first clue should be the polarity of the body diode. If it's connected as you have it shown, the FET will always be "ON" with current conducted by the body diode.
allready explained to him before but I think the schematic comes from an ftp and the drawing was updated.
allready explained to him before but I think the schematic comes from an ftp and the drawing was updated.
The Updated March 04 drawing is still wrong.
I hope he will update his drawing.
Thanks for the replies guys! So this is the situation now..
Changed the fets to correspond ones im planning to use. Added ACS712 for current and capacitors for it.
Also added a diode to protect from coils. Is it in the right place?
Will be adding a jumper to bypass ACS712 if not needed. Traces for external ACS712 missing at this point.
If reset is pressed while load has power, it may continue running due to charge at gate. Add another resistor, Q2 gate to gnd.
It is not going to be an issue. This arduino will only shut down when no one is aboard the boat. Basically main switch for the boat is the main switch for the arduino.
And if it does get resetted somehow, it doesnt matter if the fet stays on. It will be on most of the time.
I was more worried about the fet going OFF when arduino goes to sleep, but luckily arduino's outputs stay on the state they are in(?)
Feel free to comment on the schematics and of course on the board.
the source and drain are reversed on Q2.
the +12v input must be connected to the source of the IRF4905 (pin 3)
and beware of PTC. they have high internal resitance compared to fuses and they can self heat with high current like 5 Amps.
The copper traces must be large for D1 and starting directly (without vias) between the ground and output tabs to damp high EMF and avoid ringing loop with inductive loads .
Hmm, arey ou sure about the reversed pins?
PTC should be rated Rmin = 0,02ohm (initial)
Thats P = R*I = 0.02 * 5 * 5 = 0.5W, right?
Some details about the load side.
Device AVG MAX
Fridge 0.4A 3.3A
Lights-out 2A
Lights-in 1A
Map/Gps 0.2A
Sonar 0.2A
Autopilot 1A
So if everything is on, which is a rare case, it will be 5A. But Im counting on peaks and stuff
And at this point thats all the load. So as you can see, basically fridge and autopilot are the ones with some sort of motor/compressor in it.
it's a PMOS so the +12v must be connected to it's source.
the source from the datasheet is the pin 3
you can check the figure 12a page 6 of the datasheet 
The PTC self heating with 0.5W can rise it's internal resistance -> more power -> more heat-> thermal trip.
so carefully choose the right PTC model and take the ambiant temperature in the current derating
Latest additions:
Need to fix the 3pin ACS712. Somewhy its not routing it in the board. Its routed in schematic.
But anyhoo. how does it look like? Are those Input and Output polygons a total mess, or good for the current?
Or would it be the same to just route them as FETs Drains is? (2mm wide trace)
Or in general what would you do differently?
The PTC self heating with 0.5W can rise it's internal resistance -> more power -> more heat-> thermal trip.
so carefully choose the right PTC model and take the ambiant temperature in the current derating
Any other suggestions to protect from too much current, without taking much footprint?
it's a PMOS so the +12v must be connected to it's source.
the source from the datasheet is the pin 3
Hmm.. Ill have to believe you and change it 
neepie:
Any other suggestions to protect from too much current, without taking much footprint?
yes: a fuse. it's cheap, fast or slow (like you want) and reliable
about the polygons they are ok but you will have some troubles to weld the tabs due to the copper area. (heatsink effect)
try to enable the thermal relief on pads.
check if "isolate" is not too low because i find the isolation space very thin.
yes: a fuse. it's cheap, fast or slow (like you want) and reliable
Fuse takes too much space 
try to enable the thermal relief on pads.
I actually disabled them for the best effect on current flow.
Isolation between FET and output is about 0.5mm. Is that not enough?
fuses are available in very small sizes.
little fuses
nano fuses
even with SMD holders
about the "isolate" i'm talking about the gap between two signals when you have a pad (ground polygon options -> isolate)
fuses are available in very small sizes.
little fuses
nano fuses
even with SMD holders
You're so right. I did take smd socket with 500mA fuse to cover the arduino input. But i think ill save space and use fuse inline with cable to save pcb space!
about the "isolate" i'm talking about the gap between two signals when you have a pad (ground polygon options -> isolate)
I was talkin about the same isolate
Got some parts today. Will be trying stuff soon!
so the isolate value seems to be too low. the gap between the pads and the gnd seems to be very very thin