Manager
Joined: 14 Jan 2018
Posts: 80
Given Kudos: 37
Location: India
GMAT 1: 610 Q49 V24 GMAT 2: 640 Q49 V29 GMAT 3: 710 Q49 V38
GPA: 3.2
WE:Information Technology (Consulting)
GMAT algorithm's similarity to the Chess rating Elo System
[#permalink]
01 Aug 2020, 14:20
I couldn't find mentions of such a link on the web and thought I could throw some light on my observations of the GMAT scoring algorithm.
As an avid lover of chess, let me first explain the ELO system of chess.
When you start chess you are assigned the average rating of 1200.
In the beginning, the ELO is very sensitive and shifts over 150+ plus rating points in the positive or negative direction depending on a win or lose.
Slowly as you play more games the sensitivity decreases and finally adjusts to around a difference of 6 rating points per game while playing with an opponent of the same level.
Note that as each game progresses your opponent is chosen very close to your rating.
Thus, even when a Grandmaster(2500+ ELO) starts fresh with an average rating of 1200, it would only take 10-15 games for the system to grant him the Grandmaster Rating.
This is how an ELO system is perfect while judging a chess player's skillset.
The GMAT algorithm is very similar to this. Instead of getting matched with a player of your level in chess and adjusting the ELO sensitivity accordingly, it will do so with question difficulty. Hence the GMAC confidently says that it can judge the skill level of test taker up to 30 point difference with just around 30 questions.
Instead of trying to decode the algorithm, I guess its best to trust the algorithm and just focus on building a strong skill set. The magic of score improvement automatically happens with better technique and hardwork.