Submission #7646: eien86, Aglar & Randil's NES Ironsword: Wizards & Warriors II in 07:09.06

(Link to video)
Console Nintendo Entertainment System Emulator EmuHawk 2.8.0 (Core: NESHawk)
Game Version USA Frame Count 25786
ROM Filename Ironsword - Wizards and Warriors 2 (U).nes Frame Rate 60.0988138974405
Branch Rerecord Count 47
PowerOn Authors eien86, Aglar, Randil
Game Ironsword: Wizards & Warriors II
Submitted by eien86 on 8/14/2022 1:43:55 PM

Submission Comments


IronSword is a charming action/adventure game that I wish I had played when I was younger. Instead, here I am, shaving 5.2 seconds out of its last stage. This movie was a challenge/consignment by Alyosha, who thought it would be a good idea to apply a bot to the infamous last boss fight in this game.
The final boss fight consists of 4 floting element heads that come and go on a 2D plane, attacking the protagonist. It is particularly difficult to optimize manually, given the butterfly effect-like every decision you make has. I ran the bot on an exceptionally large exploration database while constraining the hero to stay put (he cannot run). And it did find a better solution.
This movie contains only changes to the aforementioned fight (for a longer description of the rest of the game, see [2777] NES Ironsword: Wizards & Warriors II by Aglar, Randil, Alyosha, rchokler, Samsara in 07:14.32, and its 315 frame faster solution introduces a seemingly new concept: multi-fire. So it seems you can hack the game into firing multiple times before the previous projectiles leave the screen (if this was known but not used, disregard this paragraph).

Comparison Movie

Software + Hardware

Rom Information

  • Name: Ironsword - Wizards and Warriors 2 (U)
  • SHA1: 03130F8464B3F4418427BF124EB15FBAEB86E09D
  • MD5: CD28188CA6B0A4D1E7C34FA47285BEC9


  • EmuHawk 2.8.0 (Core: NesHawk)
Manually resynchronized from a EmuHawk 2.8.0 + QuickNES movie.

Routing Bot

  • Bot: Jaffar
  • Routing Core: QuickNES
  • Platforms:
    • AMD Ryzen Threadripper 3990X Processor (64 cores, 128 threads) + 256Gb RAM (Average Exploration Performance: 1.1M States/s)
    • 2 x AMD EPYC 7742 Processor (128 cores, 256 threads) + 512Gb RAM (Average Exploration Performance: 2.2M States/s)


Q: Why not bot the rest of the game? A: While the bot has achieved success on many 1.5D games (Castlevania, Ninja Gaiden, PoP), where the hero mainly traverses the level horizontally and ocassionaly jumps on a platform, this game is proper 2D. This means that the paths here can have dominant X, Y or diagonal directions. This makes the exploration space much bigger. Although I believe it could be still successful in saving a bunch more frames, it would be a months-long project and I'd like to take a break from juggernaut projects for now (burnout).
Q: Can the final fight be improved? A: Yes, of course. I had to cripple the hero into not being able to run in order to avoid him from pursuing one of the heads and falling to oblivion. If anyone wants to manually optimize this fight making use of the multi-shot and being able to run, then there's a real chance to improve it. But then again, in 5/10 years I believe we'll have powerful enough systems to approach a pretty much perfect solution.


Thanks Alyosha for the fun challenge, I look forward to new ones.
Thanks to the authors of the previous movie: Aglar, Randil, Alyosha, rchokler, Samsara. Yours are the giant shoulders on which this movie stands.

feos: Claiming for judging.
feos: Nice improvements! If it's true that everything up to the final boss was done by Aglar and Randil (it's non-trivial to tell since they had a bunch of WIPs, the most optimal of them being unfinished), then the authorship of this movie looks correct, because inputs for the final boss fight in this run are completely different from the predecessor. Accepting over [2777] NES Ironsword: Wizards & Warriors II by Aglar, Randil, Alyosha, rchokler, Samsara in 07:14.32.

despoa: Processing...

Last Edited by despoa 9 days ago
Page History Latest diff