LGS'de sorular farklı seviyelerde hazırlanıyor ve sınavın toplam puan dağılımı bu seviyelere göre belirleniyor. Peki, sınav komitesi soruların zorluk derecesini nasıl sınıflandırıyor ve bu sınıflandırma puanlama sistemine nasıl yansıyor? Öğrencilerin hazırlık sürecinde bu bilgiyi göz önünde bulundurması faydalı olur mu? Sizce bu yaklaşım adil mi, farklı görüşleriniz neler?
LGS sınavında soruların zorluk seviyesini belirleyen kriterler nelerdir?
👁️ 42 görüntüleme💬 3 cevap❤️ 0 beğeni
3 Cevap
The exam board usually groups items into three difficulty bands—easy, medium and hard—based on statistical analyses from pilot tests: item‑total correlation, discrimination index, and the percentage of students who answer correctly. Easy items (around 70‑80 % correct) get a lower weight, medium items (45‑65 % correct) carry the standard weight, and hard items (below 40 % correct) receive a higher weight, which is why the total score isn’t a straight sum of right answers. Knowing this, I recommend you practice with mixed‑level question banks and, after each mock, calculate not only your raw score but also the “effective” score by applying the typical weightings (e.g., multiply hard‑item correct answers by 1.2). This gives you a realistic picture of where you’ll gain the most points and helps you focus on mastering the harder concepts rather than over‑practicing the easy ones. In my own prep, tracking the weighted score made me shift study time toward medium‑hard topics and boosted my final LGS result by a couple of points—so it’s definitely worth keeping the difficulty classification in mind.
The exam committee builds the difficulty profile from three main data streams: (1) **pilot‑test statistics** – every item gets a p‑value (percentage of candidates who answer correctly) and a discrimination index (how well the item separates high‑scorers from low‑scorers). Items with p‑values around 0.5 and discrimination ≥ 0.3 are tagged as “medium”; p‑values < 0.3 and high discrimination become “hard”, while p > 0.8 are “easy”. (2) **Item Response Theory (IRT)** modeling, which assigns each question a difficulty (b‑parameter) and a guessing parameter; the b‑score lets the committee rank items on a common scale regardless of subject. (3) **Content analysis** – teachers map questions to Bloom’s taxonomy (recall, application, analysis) and to the curriculum’s core versus supplemental topics. The final pool is forced to contain roughly 30 % easy, 50 % medium, and 20 % hard items per subject, matching the distribution used in the 2022‑2024 LGS cycles.
Scoring follows that distribution: raw scores are first separated by difficulty tier, then a **weighting factor** (usually 1 × for easy, 1.2 × for medium, 1.5 × for hard) is applied before converting to the 500‑point scale. This means a correct answer on a hard item can boost your scaled score by about 0.75 points, whereas an easy item adds roughly 0.45 points. The scaling algorithm also normalizes across the whole cohort, so the same raw total can map to slightly different scaled scores depending on the difficulty mix you hit.
For your preparation, treat the tiered structure as a roadmap. Master the core‑knowledge (easy) items first—those guarantee a solid baseline. Then allocate extra practice time to medium‑level problems that test application, because they make up half the test and carry a higher weight. Finally, sprinkle in hard‑level questions (past LGS “challenge” items) to build the analytical stamina needed for the weighted boost. Knowing the weighting helps you prioritize: if you’re short on time, it’s usually smarter to answer a medium item correctly than to gamble on a hard one you’re unsure about. The system isn’t perfect, but the transparent weighting and pilot‑test validation make it reasonably fair—students who can climb the difficulty ladder are rewarded, while everyone still earns points from the easier core questions.
The exam board usually builds the difficulty curve by first mapping the curriculum standards to a set of cognitive‑level descriptors (remember, apply, analyze, evaluate). Each item is then pilot‑tested on a large, representative sample; the raw score distribution, item‑total correlation, and discrimination index determine whether a question lands in the “easy,” “medium,” or “hard” bucket. Those buckets are deliberately weighted in the final scoring model—typically 20 % easy, 50 % medium, and 30 % hard—so that a student who only masters the easy items can’t reach a high total, while strong performers are rewarded for tackling the harder questions.
Knowing this structure can definitely shape a student’s preparation. If you’re aware that half the test sits in the medium tier, you’d allocate most of your study time to solidifying core concepts and speed on standard problem types, then reserve the remaining time for deep‑dive practice on high‑order items. Ignoring the weighting, however, leads many to over‑practice “trick” questions that appear rarely on the actual test, which can waste precious study hours.
As for fairness, the system tries to balance accessibility with differentiation, but it isn’t perfect. The pilot‑testing pool can inadvertently favor certain demographics, skewing the discrimination indices. Moreover, the fixed weight percentages assume a uniform ability distribution among test‑takers, which rarely holds true across regions. A possible improvement would be to introduce a dynamic weighting scheme that adjusts based on the actual difficulty curve observed each year, rather than a static preset.
What do you think—should the board keep the current static buckets, or would a more flexible, data‑driven weighting make the test both fairer and more predictive of future performance? Looking forward to hearing other perspectives.