500 problem set

[Vandivier]

The Neetcode 250 is the Neetcode 150 plus 100 more problems. It's the most beginner friendly list, intended for people completely new to algorithms.

ok let's do that then get to 500 to maximize gains according to interviewing.io

let's also combine insights from the problem coverage data...basically skip bit manipulation and add hard problems in other areas instead

we can retain the "Ladderly Apocrypha" which are those neetcode questions I don't think are high value for overachievers. so that set would be like 510 to 523 depending on how many of the math+geo I trim

[Vandivier]

related #38 and #38 (comment)

[Vandivier]

ope! don't forget the grind 75
https://www.techinterviewhandbook.org/grind75/
https://www.techinterviewhandbook.org/grind75/faq

the grind 75 is actually a list of 169 questions: 37 easy, 107 medium, and 30 hard.

https://www.algoexpert.io/questions
is another good reference, although we can't legally use their full list as it's not an open list, but we can informally compare

it's a len(200) list where 31 are easy, 73 are medium, 58 are hard, and 38 are very hard. Leetcode doesn't describe a very hard tag, so this is roughly equivalent to 96 hard problems.

grind 75 guy says:
Practically speaking the return of investment (ROI) on studying and practicing for DP questions is very low. Hence DP questions are less important/optional and you should only do them if you have time to spare and you're very keen to have all bases covered (and interviewing with Google).
https://www.techinterviewhandbook.org/best-practice-questions/#week-5---dynamic-programming

grind 75 guy also says that his 169 problem set should be enough by itself and he may well be right - the interviewing.io study didn't correct for pattern coverage and structured learning. The 169 problem set covers most classes of problems well and it may do a better job than a relatively random group of 300-500 problems, but the Ladderly 300 and 500 have all of these advantages and more for the risk averse individual and data-driven consumer: pattern coverage, structure, ease of onboarding, and volume. If Yangshun is right, and I think he is, then the Ladderly 300 and 500 problem sets should be even more effective than the math from the interviewing.io article indicates.

It's important to note that if Ladderly problem sets are better than the interviewing.io article indicates, then there is even less room for further question answering to matter. The diminishing returns to further problems are even smaller, that is, extremely small. This agrees with our recommendation to drill these problems, improving time, accuracy, and consistency, rather than growing the problem count beyond 300 or 500.

so we break out a 300 and 500 where 300 trims low value algos like DP, bitwise algorithms, and special-algorithm problems in favor of flexible-solution hard problems as advised by interviewing.io

the 500 has the full union of all the things including exotic algorithms and bit manipulation

[Vandivier]

[Vandivier]

this is partially closed and creates four sets in the kata blog article:

• 300
• 420
• 500
• 690

TODO is: create actual leetcode.com custom problem set lists

nice-to-have follow ups:

1. random kata problem #276
2. enhance solutions to include these lists...

...in fact, let's leave this open to include the solutions bc i don't want more than 145 issues in the backlog rn lol