My story at ACM ICPC World Final 2016

Правка en3, от I_love_Hoang_Yen, 2016-05-20 07:05:07

My team (me + flashmt + nguyenhungtam) was probably never noticed by anyone, but we were actually aiming for medal. And I think we performed ok in the final, ending up at #15, still need some more luck & brain to win a medal. Anw, I decided to share my story. Btw if you are Vietnamese, you can also read this.

Back story

In the ACM history, a Vietnamese team has never won any WF medal. 5 years ago, team of ktuan (only Vietnamese red target on Topcoder) had high hope, but ended up at #17. 4 years ago, team of ConanKudo also had high hope, but messed up in the first 3 hours and ended up at #17.

So I really wanted to win medal. And I've been planning for this World Final for several years..

Team formation

I've chosen flashmt as teammates for several years, but were not able to choose any third member. Since I only want to select Vietnamese, we have only few good options (darknstd, infrmtcs — both IOI silver medalist). After the WF 2015, I read this comment by fhlasek about the story of his team, and decided to pick an IMO guy instead. Thus, I picked nguyenhungtam who was IMO medalist but has very few experience with programming / algorithm and I trained him from that. It was a good choice, I had a lot of fun watching/trying to understand problem solving from Math people perspective, and we did much better with better mathematical imagination.

Regionals

At regionals, nguyenhungtam has few experience, so I was really afraid of losing ticket to WF. Anw, we won in Jakarta regional which guaranteed a ticket. We were #3 at Singapore regional, but I always believe we can do much better as nguyenhungtam improved very fast towards WF.

Strategy

Our strategy is flashmt will code mostly all problems while nguyenhungtam and I solve all the problems on paper. This is mainly because I usually fail to code simple things in important competitions (anw during this WF I implemented B & K, both had 1 wrong submission, but I think it is ok considering my current standard). And we also believe that the WF problemset requires more thinking, so 2/3 team time thinking is good.

And indeed, there are many positive side on this strategy:

  • I think I had much more time solving problems.
  • The algorithm needs to be very clear before we start implementing them.
  • Since flashmt, leaving him the only one coding is definitely good enough.

My take on the WF problemset

I think the easier problems are ok, but maybe too many of them (8 easy problems..). I liked K. The judge solution for B is nice, but I think they are not very familiar with Divide & conquer DP optimization, thus this problem became copy-paste divide & conquer DP from our notebook. (I publicized our notebook here — you can use it at your own risk)

There were 5 hard problems F H I J M. I think H I J are too hard, M is hard to implement, so my team only had one choice of solving F: either we solve F and get medal or we don't solve F and don't get medal. I think the choice is very limited this year. Maybe if there was 1 medium problem instead of easy problem, there would be some choice for teams after solving easy problems.

Anw, my team strategy worked well. We always have something to code until we finished the 8th problem. And flashmt did not have too hard-to-debug bugs, so we were solving problems quite fast considering our level.

After this WF, probably I'll partially retired (probably still participate in CF rounds & big contests but mostly no training). If you have questions for me in comments, I'll gladly answer.

Теги acm, acm icpc world finals, team training

История

 
 
 
 
Правки
 
 
  Rev. Язык Кто Когда Δ Комментарий
en4 Английский I_love_Hoang_Yen 2016-05-20 10:16:10 345
en3 Английский I_love_Hoang_Yen 2016-05-20 07:05:07 284
en2 Английский I_love_Hoang_Yen 2016-05-20 07:01:41 3916 (published)
en1 Английский I_love_Hoang_Yen 2016-05-20 06:39:54 1428 Initial revision (saved to drafts)