Why Solvable DILR Sets Still Take Far Longer Than Expected

DILR

You picked the set correctly. You read it, you understood it, you could see how it worked, and you solved it. It took eighteen minutes. That is nearly half the section for one set, and the worst part is that every individual decision along the way looked reasonable.

This is the most common way a DILR section goes wrong, and it is distinct from picking a bad set. A solvable set that overruns is not a selection failure in the usual sense, because the thing you were assessing at selection time was whether it could be done. The variable you failed to assess was what it would cost, and that is a separate property with its own markers. This piece names the four drivers of overrun, gives you the checkpoints that catch them, and explains the one abandonment rule that saves sections.

Solvable and Cheap Are Different Properties

The framing error sits right at the start, in the moment of selection.

When candidates assess a set, they are asking whether they can do it. That is a reasonable first filter and it screens out the sets that are genuinely beyond reach. What it does not tell you is how long doing it will take, and those two properties are only loosely related.

A set can be conceptually simple and computationally enormous. A set can be immediately clear and then require you to test fourteen cases. The clarity you felt at selection was about the first of those and told you almost nothing about the second, which is why the overrun feels like a betrayal: the set kept its promise, you just misread what the promise was.

The Four Things That Make a Set Expensive

Each of these is visible early if you know to look, and each behaves differently once you are inside the set.

Branching Is the Largest Cost

If one of the four dominates, it is this, and it is the hardest to see at selection.

A well-behaved set has constraints that lock things down quickly. You place one element, that forces another, and within a few steps the structure is largely determined. A branching set gives you constraints that permit several configurations, so instead of building one grid you are maintaining three, and every subsequent constraint has to be tested against all of them.

The cost is not linear. Two open branches is manageable. Four is usually fatal inside a 40 minute section, because the bookkeeping starts consuming more attention than the reasoning does, and errors in the bookkeeping are invisible until they have propagated.

Why Each Extra Minute Feels Justified

Here is the mechanism that turns a ten minute overrun into an eighteen minute one, and it is not stupidity.

At minute eight you have real progress: a partial grid, several placements, work you can see. Abandoning means losing it. At minute ten you have more, so abandoning costs more. At minute fourteen the investment is large enough that leaving feels absurd, and the set now has you regardless of whether continuing is rational.

The error is treating spent minutes as an argument for spending more. Those minutes are gone whichever way you decide, and the only live question is what the next five minutes buy compared with what they would buy in a different set. Framed that way the decision is usually obvious, which is exactly why the framing has to be built in advance rather than found at minute fourteen.

Checkpoints, Not Willpower

Because judgement degrades inside a dragging set, the answer is a rule that runs on the clock rather than on how you feel.

Checkpoint	What to ask	Action if it fails
2 minutes	How many open branches am I holding?	More than two with no closing constraint, leave
5 minutes	Is the structure mostly determined?	Still fully open, leave
8 minutes	Can I answer at least one question now?	No, leave and do not return
12 minutes	Is the remaining work bounded and visible?	Not visible, take the answers you have and go

Take the Partial Answers and Leave

The checkpoint at twelve minutes contains the most valuable habit in the section, and it is one candidates resist because it feels like accepting defeat.

Many sets let you answer one or two questions from an incomplete picture. A partially solved set that yields two correct answers is a good outcome, not a failed one, and the instinct to hold out for the complete solution is what converts a decent return into a section-destroying one.

The marking supports this directly. Each question stands alone: a correct answer is plus three, an incorrect MCQ is minus one, an unattempted question is zero, and Type In The Answer questions carry no negative marking. Nothing rewards completing a set, so there is no bonus waiting at the end that justifies the last six minutes.

Adding Cost to Your Selection Read

Since the root cause is a selection read that only checked solvability, the upstream fix is to add a second question to it.

After you have decided you understand a set, spend ten more seconds asking what shape the work has. Is there a constraint that locks something immediately, or do the conditions all permit multiple placements? Are the questions asking for the same structure or for different ones? Is there a table to build once, or a computation to repeat?

That ten seconds is the cheapest thing in the section. It converts selection from "can I do this" into "can I do this, and roughly what will it cost", and the second half is what protects the 40 minutes.

How to Train the Cost Estimate

This is a recognition skill built from volume plus a specific review habit, and the review habit is the part that is usually missing.

Record time taken on every set you practise, including the ones you solve correctly. A set solved in six minutes and a set solved in fifteen look identical in an accuracy log and are completely different events. Until your data separates them you have no information about cost at all, which means no basis for estimating it.

Then, for every set that overran, identify which of the four drivers caused it and whether the marker was visible at the start. Most candidates find the same driver appearing repeatedly, which is useful, because it means the fix is narrow. Someone who consistently loses to branching needs to practise counting branches early, and that is a specific drill rather than a general resolution to be faster.

Practise Reading Without Solving

There is a drill that targets this directly and takes very little time.

Take ten sets and, for each, read it and write down two things only: whether it is solvable for you, and your estimate of how many minutes it would take. Then solve three of them and compare your estimates against reality. You are training the estimate itself rather than the solving, and ten sets treated this way cost a fraction of what solving them all would.

After a few rounds the estimates stop being guesses. That is the whole skill, and it is the one that separates candidates who select well from candidates who merely solve well.

The Summary

A solvable set that overruns is not a selection failure in the usual sense. It is a failure to assess cost, because the selection read only checked whether the set could be done.

Four things drive the overrun. Branching, where the constraints permit several configurations at once and the bookkeeping starts costing more than the reasoning. Late constraints, where the decisive condition arrives after the early work is done. Recomputation, where different questions need different structures from the same base. And partial resolution, where the set never fully closes but keeps you working towards a completeness that is not available.

Use checkpoints rather than willpower, because the judgement degrades as the set drags and spent minutes cannot be valued at zero by the person who spent them. Take partial answers and leave, since nothing in the marking rewards completing a set. And add cost to your selection read, because ten seconds spent asking what shape the work has is the cheapest protection available for a 40 minute section.

If your DILR review only records right and wrong, you have no cost data and no way to build the estimate that protects the section. A CAT preparation strategy review will identify which of the four drivers keeps costing you, and a personalised CAT preparation plan schedules the estimation drill alongside the solving.