Signal Grid: solve one row at a time, then check the last
Once you choose the first row of presses, the rows below are forced. Learn the method, its last-row test and the mistakes that make it seem unreliable.
Start with the row a press can still reach
A press affects its own row and the rows immediately above and below it. Once you reach the third row of a five-by-five board, none of its presses can change the first row. This limited reach suggests a method: finish a row, move down, and never disturb that finished row again.
The important qualification is that you must first choose which squares to press in the top row. You may initially choose none. That is one starting choice, not proof that no top-row press is needed. The remaining rows will tell you whether that particular choice leads to a solution.
Use the square directly below each remaining light
After making the chosen top-row presses, inspect the top lamps from left to right. Whenever one is lit, press the square directly below it in the second row. That press changes the lamp above it. Other second-row presses cannot reach that same top lamp, so each decision is forced by the lamp you see.
When you finish the scan, the whole top row is unlit. Repeat the process for the second row, using presses in the third row. Continue with the third and fourth rows in the same way. Each scan clears the row above without reaching back far enough to change any row you finished earlier.
Read the bottom row as a verdict
After using the fifth row to clear the fourth, inspect the bottom row itself. If it is also unlit, you have solved the puzzle. If lights remain, there is no sixth row from which to clear them. Pressing the fifth row again would disturb the fourth row you just completed.
That remaining pattern tells you that your initial top-row choice did not work. It does not mean the generated puzzle is impossible. Return to the starting board and try a different set of top-row presses, then repeat the same forced scans. Mixing a new top-row choice into an old half-solved board makes the comparison much harder.
Keep a small, explicit record
There are five top-row squares, each either pressed once or left alone, giving 32 possible first-row choices. You do not need to remember the whole sequence of lamp changes. Record the five yes-or-no choices and whether the final bottom row cleared, so you avoid repeating a choice that already failed.
For example, a note reading press columns 2 and 5 describes a first-row choice; it is not a suggested answer for every puzzle. Number columns from left to right and keep the same convention throughout. Restart before testing that choice, then follow the lamps instead of copying the lower-row presses from a previous attempt.
Understand what the method guarantees
Checking all first-row choices covers every possible solution under these rules: a solution has some first-row choice, and after that the need to clear the row above determines every lower-row press. Because the generated board is solvable, at least one of those choices will succeed if the scans are performed correctly.
The first successful choice is not necessarily the one with the fewest presses. This is a systematic way to find a solution, rather than a promise of an optimal one. If the method seems inconsistent, check for diagonal presses, skipped columns or a failure to restart between trials before abandoning the row-by-row reasoning.
An original guide for this site's implementation. Found an unclear step? Let us know.