Donkey Kong Forum
General Donkey Kong Discussion => General Donkey Kong Discussion => Topic started by: krehztim on January 02, 2018, 11:13:47 am
-
In pondering the answers to my previous thread, I wonder how high the original developers thought players would go? And, no, this isn't a drug-related question.
I wonder if during development and testing, if they anticipated million point games, and I wonder if there was any analysis done back in the stone age about maximum scores (not likely). If Activision could accept and TAS 5.54s, maybe Nintendo could envision 1.5 million. Or, maybe they didn't care.
-
I don't think the developers knew about the killscreen. Having said that, most developers underestimated the players big time. Given that the game doesn't count the millions digit, I'm going with somewhere between 500k, and one million.
-
Having said that, most developers underestimated the players big time.
Yeah I think I read an interview with Jarvis when he said in testing they thought about 100k would be about the ceiling for scoring in Defender and he said within a few months of its release people were getting millions.
-
Having said that, most developers underestimated the players big time.
Yeah I think I read an interview with Jarvis when he said in testing they thought about 100k would be about the ceiling for scoring in Defender and he said within a few months of its release people were getting millions.
Yeah, and I'm still stuck at 100k
-
I don't think the developers knew about the killscreen.
I think this is an interesting thought, and a question for any game with a kill screen. Did the developers not know about it, or did they think it would never serve as a limitation?
-
This could be totally wrong, but I have a feeling the 3rd elevator stage and beyond was intended to create a 50/50 survival situation to shorten play times. I doubt they thought players would find a way to master it.
-
And the steering barrels on 5+. To a new player it’s certain death since they’re being hunted by barrels.
-
I'm pretty certain they didn't expect players to get beyond L=5. The L=4 elevators would have been considered almost certain death. The L=5 wild barrels were (intended) to home-in-and-kill, plus the high steering to kill Jumpman approaching ladders.
-
I often wonder what the HSL would look like if levels 5 through 21 were L4 wilds and L1 steering. Game would be absolute hell
Thank the devs for these oversights
-
I often wonder what the HSL would look like if levels 5 through 21 were L4 wilds and L1 steering. Game would be absolute hell
Thank the devs for these oversights
This forum likely wouldn't exist, since the game would be like playing the lottery.
-
The nail in the coffin would be the increased speed up to L5 speed level, for L5-L21, but with L3 wilds and unsteerableness.
Hmm...mild increased difficulty hack/variation anyone? Sock Master? ;D <popcorn>
-
A really brutal variation could have seen the last 4 levels being a "finish", as opposed to the start now. L21 on WR pace and you get a couple of murder barrels to the head. I don't think DK would be a very popular game ;D
-
Remember that 40 years ago this was a "maximize the extraction of quarters" business, not a "maximize the depth of gameplay" business.
I'm with marky and Sock on this. It seems pretty clear that the third elevator board was designed to be as deadly as possible without being impossible, and I don't think it's a coincidence that they plateaued the internal difficult two boards later. Reaching level 5 would take about 10 minutes (without doing the goofy stuff that we do), and that would be about the limit for how long they'd want you on the machine.
Arcade games were designed to collect more quarters by making the player want to get better, but also to restrict play time to a reasonable maximum. So not too long, but not too short. It was a tough balance, but they definitely took that problem very seriously. Atari would field-test all of their games to make sure the average coin wouldn't last too long (I think they were shooting for 3 minutes or something like that), and if play times were too short or too long, they'd optimize until they hit their goldilocks target.
If long play times became a problem, manufacturers would release "speedup kits". "Speedup" as in, speed up the rate of quarter-turnover. The patch for the barrel cheat that Nintendo released for DK in December '81 was created exactly for that purpose.
So I can pretty much guarantee that the DK designers were not envisioning scores in the several hundred thousands, and certainly not in the millions (that's why there's no 7th digit), because that scenario was actually their worst nightmare. It would mean that they had failed at their jobs. They were doing whatever they could to PREVENT long play times, short of making the game cheat outright.
(Whereas the Crazy Climber designers just went ahead and designed the game to cheat outright.)
-
This could be totally wrong, but I have a feeling the 3rd elevator stage and beyond was intended to create a 50/50 survival situation to shorten play times. I doubt they thought players would find a way to master it.
Judging by the code, I'd say the developers thought they had a fix for a killscreen potentially happening. In general though, the features that seem to be carried over from Radar Scope are really half-assed. The game is programmed to handle level display up to 99 without glitching (and as far as I recall, any level above will just display as 99 without issues, too), but that's probably just leftover code from Radar Scope.
Either way, I think the general impression that the game was expected to end at L=4 (or 5) is false, rather I think the apparently "impossible" springs on this level might be their mistake here. It is by far the worst designed challenge in Donkey Kong, and attempts at increasing the difficulty just doesn't result in the same depth as pretty much any other challenge in the game.
I think a more correct assumption is that it's attempting to be more of a random casaulty than an impossible barrier, allowing the player to sometimes get through via sheer luck.
The game code does nothing to intentionally place springs close enough to make the gap impassable, I think it's safe to say they just input some variables and tweaked them until they reached the difficulty they wanted. So I think a more realistic situation was that the stage was tested by just having someone attempt to make it through sporadically, and when they reached a low enough succes rate, they deemed the difficulty balance "correct".
In fact, I bet that's how every stage in the game would be tested, and I doubt they ever aimed for a 0% success rate - however as an arcade developer, even in 1981, I think they knew they could expect that even if they had a 1% success rate, crazy players out there would find a way to conquer the challenge. If they really wanted the game to end, why not just have a genuine ending? Instead, if the intention is to give the players an idea that L=5 or L=6 can be reached, it only makes sense that they would occasionally allow the player to successfully beat the stage.
-
Brute force is time constricted. Human play is restricted by collective human knowledge, emotions, and physical limitations. Long term, machine learning will yield the highest donkey kong score.
I'm pretty hopeful everyone agrees here.
All that's required for the machine learning to beat the humans is a little bit of economic incentive.
But I agree that we are getting high enough now that we're going to quickly venture into machine learning land to advance the score.
I hope that I can participate in that somehow assuming that the human players don't take offense to it!
a reference:
https://www.techrepublic.com/article/google-ai-beats-humans-at-more-classic-arcade-games-than-ever-before (https://www.techrepublic.com/article/google-ai-beats-humans-at-more-classic-arcade-games-than-ever-before)
I would very much like to see someone work out a similar system as MarI/O (https://www.youtube.com/watch?v=qv6UVOQ0F44), but for Donkey Kong. I think that video says the program is coded in Lua, and MAME has Lua support now.