Coding-decoding is the most mechanical family in the reasoning section and one of the fastest to clear once the alphabet positions are automatic. It is also the topic where a single untested assumption costs an entire set, because the same first letter can be consistent with three different rules and only the second word separates them.
A is 1, E is 5, J is 10, O is 15, T is 20, Y is 25. Anchoring on those six lets you find any letter's position in under a second by counting from the nearest anchor. Every method below assumes this is instant; if it is not, the topic will always feel slow.
Put the original word above its code, letter under letter, and write the numeric difference beneath each column. The differences make the rule visible at a glance — all the same means a constant shift, alternating signs mean a two-step rule, and a decreasing run usually means positions are being used.
If the differences look random, read the code backwards against the word. A large share of these questions are a simple shift applied to the reversed word, and testing that one hypothesis takes five seconds. Do it before hunting for a more complicated pattern.
A rule that explains one word is a guess. Papers deliberately construct sets where two rules fit the first example and diverge on the second, so a candidate who applies immediately gets a plausible wrong answer. If only one example is given, test your rule by re-encoding it and confirming you land back on the code.
| What the differences look like | The rule |
|---|---|
| Same number in every column | Constant forward or backward shift |
| +1, −1, +1, −1 | Alternating shift |
| Differences grow 1, 2, 3, 4 | Positional shift by letter index |
| Random until read backwards | Word reversed, then shifted |
| Letter maps to its 27-minus position | Opposite-letter (A↔Z) coding |
| Digits appear in the code | Letter replaced by its alphabet position |
Every question below was generated by the same engine that mints the Deduce daily puzzle, and machine-proven to have exactly one correct answer before it reached this page. The walkthrough is the engine's own, step for step.
If "ROAD" is written as "JGSV" in a certain code, how is "BIRD" written in that code?
Determine the coding pattern from the example and apply it to the target word.
If "FIRE" is written as "NQZM" in a certain code, how is "MILK" written in that code?
Determine the coding pattern from the example and apply it to the target word.
If "BIRD" is written as "WDMY" in a certain code, how is "MILK" written in that code?
Determine the coding pattern from the example and apply it to the target word.
The first letter of a word is consistent with many rules. Confirm across the whole word and, where possible, a second word, before you encode the target. Almost every wrong option in this family is the result of a rule that was right about one column.
Shifting Z forward by two gives B, not a letter past the alphabet. Treat the alphabet as a circle of 26 and take the remainder. Wrap errors cluster at the end of the alphabet, which is exactly where questions place them.
Some codes use position from Z rather than from A. If the numbers look mirrored, try 27 minus the position — that single check resolves a family of questions that otherwise look patternless.
One puzzle a day, timed on the server, with a national and state leaderboard and a streak that survives as long as you do. Practising a topic once is worth little; the aspirants who move are the ones who show up daily. See what Deduce does or read the rules first.
Start your streak free →Write the word and its code one above the other and subtract the alphabet positions column by column. The pattern in those differences is the rule. This takes about fifteen seconds and replaces the guess-and-check approach that makes the topic feel slow.
You need six anchors, not twenty-six: A=1, E=5, J=10, O=15, T=20, Y=25. Counting a step or two from the nearest anchor is fast enough for exam conditions and far more reliable than rote-learning the full list.
Then the rule is wrong, and finding that out is the point of testing it. Go back to the difference row and look for the pattern that explains both. Two words constrain the rule far more tightly than one, which is why sets usually give two.
The numbers are almost always alphabet positions or a simple function of them, such as position plus one or position squared's last digit. Convert every number to the letter it would represent and the mixed code usually collapses into an ordinary letter rule.
Deduce is not affiliated with IBPS, SBI, SSC or RBI.