Wednesday, September 13, 2006

More research

Yesterday I found that I was familiar with almost all problems that I had right at CTS. Today I investigated how familiar the wrong answers were. I had a look at 32 failures at CTS.
Here is the distribution of how often I had seen the problems before I did them wrong:


#times seen before
#problems
familiarity
0
6
new
18vague
211familiar
34familiar
4
2
familiar
5
0
familiar
6
1
familiar
1.75 avg
32


I had seen 6 problems 0 times before, 8 problems once, etc..
The problems that I had seen never or only once, looked unfamiliar or vague to me while the rest looked familiar.
The familiar problems I had wrong because I fell for "cheapos".
That is basically what happens when you try to speed up familiar problems: you fall for cheapo's while gambling. So the school that goes for accuracy at CTS is definitely right.

But neither the school of accuracy nor the school that aims for speed at CTS has found a method to improve the speed of solving familiar problems at CTS.

So the question arises: how often have I seen the problems before that I solve right and fast?
This is the distribution of the problems I solved correctly within 3.0 seconds:


#times seen before
#problems
0
2
12
24
32
4
2
5
2
6
2
15
1
3.5 avg
17


The relation between more repetitions and fast solving is not as clear cut as I would like. Maybe due to the wee problemset involved? The fact that unknown problems can be solved in 3 seconds without repetition is encouraging though.

So the question remains: is massive repetition the only method to faster solving or is there a better way?

1 comment:

  1. I wonder if median is a better measure because of the skew of the second distribution. The medians are 1.5 for incorrect and 2.5 for correct.

    ReplyDelete