Two months ago, I posted this [blog](https://codeforces.me/blog/entry/154231) on training for IOI Gold. Despite not feeling particularly good about my chances at a gold medal going into the contest, luck was on my side and I did end up winning a gold in the end! Therefore, I'm writing this blog as a follow-up to my previous blog, as well as to share about some of my experiences and opinions about IOI and training for it. ↵
↵
### Final months leading up to IOI↵
↵
In the two months between posting the previous blog and participating in IOI, I almost entirely stopped doing problems from CF and AtCoder, and instead fully focused on OI-style problems. I had previously done virtual participations in IOI 2016-2023 (and participated live in IOI 2024 and 2025), but deliberately didn't read most of the solutions so I could use them to gauge my progress as I periodically come back to them and upsolved. I worked through most of the problems that had room for improvement, and managed to pass the gold cutoff on all 10 sets. This mainly reassured me that my raw ability could comfortably place me in gold given an infinite amount of time, and it boils down more to coming up with the solutions within contest time. Outside of upsolving past IOIs, I also sporadically did some JOISC, BOI and CEOI sets. ↵
↵
A month leading up to IOI, I started to feel quite burnt out from solving problems, and lowered my practice volume quite a bit, and took some rest days too (which I did way less frequently before). I feel this also helped me gaslight myself into thinking that I don't care about IOI results that much so I wouldn't be too stressed during the actual contest. A week before the IOI, I participated in the yearly Malaysian IOI training camp consisting of 4 IOI-style virtuals, which got me back into an IOI-ready state. I did quite poorly in the virtuals, but felt my form slowly returning after the break period. ↵
↵
### Log of contest days↵
Before each contest day, I slept before 9pm (pretty much as soon as I got back to the hotel and washed up), so I was well rested even though I woke up at around 5am on both mornings due to the adrenaline. The fact that Uzbekistan was 3 hours behind Malaysia definitely helped, since it was already 12am back home by the time I went to sleep. The strategy I used for the contest was pretty similar to that of [user:E869120,2026-08-18]. I also did a caffeine detox starting a week before the IOI, and drank coffee on the mornings of contest days to maximise my performance.↵
↵
<spoiler summary="Contains mild spoilers to problems"> ↵
On Day 1, I decided my solve order would be 3-1-2 (just in case P3 was the easiest to grab easy partials on, it would be cool to temporarily be first place for a few minutes, which did end up happening!). I managed to AC P3 35 minutes into the contest, giving me a significant morale boost now that I only had to worry about two problems. I decided it would be reasonable as long as I started reading P2 by the halfway point, and I could focus on P1 for quite a while. I spent an hour coming up with the 47 point solution, and another hour getting to 71.6 points (details of my solution [here](https://codeforces.me/blog/entry/155911#comment-1385036)). I was pretty satisfied with my performance so far, as I had 1 AC and 1 almost-AC halfway into the contest. I spent more than an hour on the O(N^3) DP solution to P2, most of which I spent debugging. I had to write brute force code to stress test in the process, burning quite a bit of time. In the last hour, I made slight improvements to my P1 solution, and submitted some greedy guesses to the 5-point subtask on P2 which I didn't end up solving. I ended Day 1 in 17th place, with a score of 74+36+100=210. ↵
↵
On Day 2, I wanted to switch up the solve order a bit and decided on 3-2-1. In the first 40 minutes, I didn't make any significant progress on P3, and just implemented all the relatively straightforward subtasks for 23 points. I then moved on to P2, where I spotted a solution worth about 50 points almost immediately by forming a bunch of complete graphs. I managed to get 51.26 points about 20 minutes after reading the problem, which definitely boosted my confidence after being completely stumped by P3. I couldn't come up with any promising ideas after that, and moved on to P1. I couldn't come up with a general solution after about 20 minutes, so I started writing the C=2 solutions to subtasks 1-4 one by one. Notably, after getting the full score for Subtask 3, I went to the restroom and came up with the C=2 solution for subtask 4 then. I realised that I could extend my solution to subtask 4 if I handled the case where more than half of the students raise their hands at the same time separately, and managed to get C=4 for the general case at the 3-hour mark. At this point, I had about 150 points with 2 hours to go, which I felt was a fairly comfortable position. I went back to P2, where I came up with a solution using 13K nodes for even K, which took me quite long to implement. I was very surprised that a close-to-optimal solution only boosted my score by about 8 points. I tried coming up with a way to extend it to the odd case, but couldn't figure anything out after about 30 minutes. For the last hour of the contest, I grabbed a 10 point subtask on P3, and did K=3 and K=4 on P2 by hand for a small boost in score. I ended the contest feeling quite disappointed, with a score of 74.8+62.21+33=170.01, thinking I had dropped out of gold for sure. My suspicions were further fueled after asking the scores of the other 3 members of the Malaysian delegation as well as Jerome from the Philippines team, who all scored higher than me on Day 2. Fortunately, I ended up being in 19th place overall, and collapsed to the floor after finding out that I had won a gold medal.↵
</spoiler>↵
↵
### Some Reflections↵
I think I was really lucky that the problemset had a large amount of non-batch/constructive/ad-hoc problems, which is one of my strongest areas (for the Malaysian team in general too). Had the problemset been more DP/data structures heavy, I think my chances for a gold medal would have been much slimmer. The mindset of not caring too much about my results definitely helped a lot too, as I was much calmer during the contests compared to previous years. I believe that doing lots of practice on OI-style contests leading up to the IOI also helped me consolidate my contest strategy. ↵
↵
For anyone who may be in a similar position as I was before the IOI, I'd say it's really important to not stress yourself out before the IOI. I think taking a break was far more beneficial than any last-minute practice I could've possibly squeezed into that period. Besides that, just come up with a conservative contest strategy that works for you, and hope for the problems to be in your favour. ↵
↵
Now that IOI is over, I have finally been freed from the burden of having to grind CP for the past few years, and I can finally solve problems purely for the fun of it. Was the impact of OI on almost every other part of my life for the past few years worth it? I'm not entirely sure yet, but this was certainly the best outcome I could've hoped for given that the damage has already been done. ↵
↵
Finally, I would like to give a shoutout to [user:CSQ31,2026-08-18] and [user:gry3125,2026-08-18] for constantly supporting me throughout my CP journey, as well as [user:cadmiumky,2026-08-18] and [user:PeruvianCartel,2026-08-18] for taking the time to respond to my previous blog, your responses certainly gave me significant insights on practising. ↵
↵
Thank you, IOI 2026!↵
↵
↵
### Final months leading up to IOI↵
↵
In the two months between posting the previous blog and participating in IOI, I almost entirely stopped doing problems from CF and AtCoder, and instead fully focused on OI-style problems. I had previously done virtual participations in IOI 2016-2023 (and participated live in IOI 2024 and 2025), but deliberately didn't read most of the solutions so I could use them to gauge my progress as I periodically come back to them and upsolved. I worked through most of the problems that had room for improvement, and managed to pass the gold cutoff on all 10 sets. This mainly reassured me that my raw ability could comfortably place me in gold given an infinite amount of time, and it boils down more to coming up with the solutions within contest time. Outside of upsolving past IOIs, I also sporadically did some JOISC, BOI and CEOI sets. ↵
↵
A month leading up to IOI, I started to feel quite burnt out from solving problems, and lowered my practice volume quite a bit, and took some rest days too (which I did way less frequently before). I feel this also helped me gaslight myself into thinking that I don't care about IOI results that much so I wouldn't be too stressed during the actual contest. A week before the IOI, I participated in the yearly Malaysian IOI training camp consisting of 4 IOI-style virtuals, which got me back into an IOI-ready state. I did quite poorly in the virtuals, but felt my form slowly returning after the break period. ↵
↵
### Log of contest days↵
Before each contest day, I slept before 9pm (pretty much as soon as I got back to the hotel and washed up), so I was well rested even though I woke up at around 5am on both mornings due to the adrenaline. The fact that Uzbekistan was 3 hours behind Malaysia definitely helped, since it was already 12am back home by the time I went to sleep. The strategy I used for the contest was pretty similar to that of [user:E869120,2026-08-18]. I also did a caffeine detox starting a week before the IOI, and drank coffee on the mornings of contest days to maximise my performance.↵
↵
<spoiler summary="Contains mild spoilers to problems"> ↵
On Day 1, I decided my solve order would be 3-1-2 (just in case P3 was the easiest to grab easy partials on, it would be cool to temporarily be first place for a few minutes, which did end up happening!). I managed to AC P3 35 minutes into the contest, giving me a significant morale boost now that I only had to worry about two problems. I decided it would be reasonable as long as I started reading P2 by the halfway point, and I could focus on P1 for quite a while. I spent an hour coming up with the 47 point solution, and another hour getting to 71.6 points (details of my solution [here](https://codeforces.me/blog/entry/155911#comment-1385036)). I was pretty satisfied with my performance so far, as I had 1 AC and 1 almost-AC halfway into the contest. I spent more than an hour on the O(N^3) DP solution to P2, most of which I spent debugging. I had to write brute force code to stress test in the process, burning quite a bit of time. In the last hour, I made slight improvements to my P1 solution, and submitted some greedy guesses to the 5-point subtask on P2 which I didn't end up solving. I ended Day 1 in 17th place, with a score of 74+36+100=210. ↵
↵
On Day 2, I wanted to switch up the solve order a bit and decided on 3-2-1. In the first 40 minutes, I didn't make any significant progress on P3, and just implemented all the relatively straightforward subtasks for 23 points. I then moved on to P2, where I spotted a solution worth about 50 points almost immediately by forming a bunch of complete graphs. I managed to get 51.26 points about 20 minutes after reading the problem, which definitely boosted my confidence after being completely stumped by P3. I couldn't come up with any promising ideas after that, and moved on to P1. I couldn't come up with a general solution after about 20 minutes, so I started writing the C=2 solutions to subtasks 1-4 one by one. Notably, after getting the full score for Subtask 3, I went to the restroom and came up with the C=2 solution for subtask 4 then. I realised that I could extend my solution to subtask 4 if I handled the case where more than half of the students raise their hands at the same time separately, and managed to get C=4 for the general case at the 3-hour mark. At this point, I had about 150 points with 2 hours to go, which I felt was a fairly comfortable position. I went back to P2, where I came up with a solution using 13K nodes for even K, which took me quite long to implement. I was very surprised that a close-to-optimal solution only boosted my score by about 8 points. I tried coming up with a way to extend it to the odd case, but couldn't figure anything out after about 30 minutes. For the last hour of the contest, I grabbed a 10 point subtask on P3, and did K=3 and K=4 on P2 by hand for a small boost in score. I ended the contest feeling quite disappointed, with a score of 74.8+62.21+33=170.01, thinking I had dropped out of gold for sure. My suspicions were further fueled after asking the scores of the other 3 members of the Malaysian delegation as well as Jerome from the Philippines team, who all scored higher than me on Day 2. Fortunately, I ended up being in 19th place overall, and collapsed to the floor after finding out that I had won a gold medal.↵
</spoiler>↵
↵
### Some Reflections↵
I think I was really lucky that the problemset had a large amount of non-batch/constructive/ad-hoc problems, which is one of my strongest areas (for the Malaysian team in general too). Had the problemset been more DP/data structures heavy, I think my chances for a gold medal would have been much slimmer. The mindset of not caring too much about my results definitely helped a lot too, as I was much calmer during the contests compared to previous years. I believe that doing lots of practice on OI-style contests leading up to the IOI also helped me consolidate my contest strategy. ↵
↵
For anyone who may be in a similar position as I was before the IOI, I'd say it's really important to not stress yourself out before the IOI. I think taking a break was far more beneficial than any last-minute practice I could've possibly squeezed into that period. Besides that, just come up with a conservative contest strategy that works for you, and hope for the problems to be in your favour. ↵
↵
Now that IOI is over, I have finally been freed from the burden of having to grind CP for the past few years, and I can finally solve problems purely for the fun of it. Was the impact of OI on almost every other part of my life for the past few years worth it? I'm not entirely sure yet, but this was certainly the best outcome I could've hoped for given that the damage has already been done. ↵
↵
Finally, I would like to give a shoutout to [user:CSQ31,2026-08-18] and [user:gry3125,2026-08-18] for constantly supporting me throughout my CP journey, as well as [user:cadmiumky,2026-08-18] and [user:PeruvianCartel,2026-08-18] for taking the time to respond to my previous blog, your responses certainly gave me significant insights on practising. ↵
↵
Thank you, IOI 2026!↵
↵
