TwinGalaxies had quite the extensive writeup on Todd's various cheating [1].
There was a staggering amount of evidence, including TAS (Tool Assisted Speedruns) proving that Todd's time is physically impossible. That blew up in a 274-page thread, but ultimately I believe it took Robert T Mruczek's writeup to finally ban Todd.[1] Not surprising as Mcruzek is part of the "old guard".
I think this reflected very poorly on TG and I have no intention of using them in the future.
Wait what? I was with you until the last line and now I'm confused. Why does this reflect poorly on TG if they removed it when there was good evidence it was false?
The evidence had existed for games like centipede many years, and the dragster game had been taken apart and proven fake via the actual machine code for 6+ months. TG didn't act on it until one of the inner circle got involved, that is the shameful bit.
They refused to remove it after having the evidence of falsehood for a very long time... until they got the personal arguments of one of their 'old guard' members. In other words, they don't care about objective reality and care only about the emotional investment their friends have.
> Todd's "Dragster" performance is one that I have qualified as being one of the top five video game performances of all time within multiple forum posts and in several interviews.
I'd be curious to know what the other 4 performances were.
Because they accepted beyond ridiculous scores without any real proof, or even skepticism? In like every game this dude has a record in his score is impossibly higher than 2nd place. And these aren't even games or records which have been around for a couple weeks so that you can argue there's room for improvement.
I think it's at the very least equally likely that the PR beating they were taking as a result of the Apollo video led them to make the decision rather than Robert's guidance.
Since watching King of Kong, I haven't been surprised about any of the shady stuff I hear about TG. Also, I'd recommend that anybody who hasn't watched that movie to watch it, it's hilarious.
>Rogers’s record on the classic Atari arcade game Centipede, for example, had been listed on the Twin Galaxies site as being 65,000,000. The second-highest score, 58,078, barely comes close
Yet try to submit legitimate high scores as an _unknown_ and you'll find them repeatedly dismissed on "technicalities".
Many of their arcade "high scores" for lesser known titles are easily broken by someone practicing for less than a week, but happen to be long held by TG regulars, with hundreds of "high scores".
To be fair, many of those games are rather uncommon and may just not accessible for public play by anyone. TG arcade scores must be done on original hardware. So some scores are a testament to rarity more than anything.
Particularly on the MAME (emulator) side of arcade scoring - on a previous incarnation of the TG website, the rules for submission to a game were not public until a score had been submitted for that game. Some folks with many high scores were simply submitting any score at all so that others could even play along.
Most, if not all, of Todd's scores were entered by a single referee. At the time, each 'platform' (console, computer, arcade etc.) had its own referee that would verify scores. That referee clearly was not trustworthy. However, Todd's scores were as equally valid as any other score on the site - all verified in secret by a single other human.
There has been years of doubt about Todd's scores, among those by other gamers. TG was recently purchased by a new owner. That new owner launched a 'dispute' system whereby any score could be challenged and examined by the community. There was never a formal process until now that would have allowed for this, regardless of how many doubters there were.
> Beyond the software analysis evidence, which speaks directly to Todd Rogers’s Dragster 5.51 score time
I didn't understand this sentence: it means that, looking at the code and calculating all possibilities of gear changing moments, it would be impossible to achieve 5.51 ? If so, why would any other evidence be necessary ? Or, if it wasn't such analysis (and I think that this one would be definitive), why wasn't this one done ?
Exactly, they calculated the best possible time and found out that 5.51 is not even possible.
Todd Rogers however has multiple impossible scores which are still considered "records" even though they can't possibly be obtained, some of them are stupidly obvious, like games in which the score increases in steps of five, yet he has a number not evenly dividable by five and such stuff.
I guess nobody really knows why he still holds these records.
Yes, the analysis proves 5.51 is impossible, and it's been confirmed that 5.57 is the best possible time through the use of an emulator with preprogrammed inputs, meaning that it plays the game perfectly. Todd Roger's response is that the analysis failed to consider the "human element" and how the console would respond to that.
Ah, yes, the "human element", the same thing that leads people to believe they can beat a chess AI by "confusing it" by moving randomly. Still a surprisingly popular belief despite how easy it is to test the theory.
You're probably thinking of Kasparov vs Deep Blue, the rematch. Kasparov was freaked out by a move he couldn't understand the motive for, bit assumed the computer did it for a reason. Wired magazine claims it may have been a bug.
If nothing else, an interesting side effect of this whole discussion is the realization that 5.57 is even possible. Previously the highest scores besides Todd's were multiple 5.61s. The way the game works is that 5.57 is only achievable on certain starting frames. Any one of those 5.61s could've been 5.57s but we never knew until now.
> Yes, the analysis proves 5.51 is impossible, and it's been confirmed that 5.57 is the best possible time through the use of an emulator with preprogrammed inputs, meaning that it plays the game perfectly.
An emulator with preprogrammed inputs doesn't mean it plays perfectly. It just means that it plays reproducibly. Whether or not it plays perfectly depends on the preprogrammed inputs.
How did they choose the inputs?
Looking at the game manual, it comes down to deciding when to shift, and deciding how to use the gas pedal.
Shifting looks like it is easy to deal with in emulated testing. You have neutral, 1, 2, 3, and 4. You start in neutral. So, an optimal run will start with you shifting once at t=0. You cannot down shift, so your only choice when shifting is up, and once you reach 4 you can no longer shift.
From another comment, it seems the game works in ticks of 3/100th second. That means that for a 6 second race, there are around 1.3 million possible shifting timings, assuming that you cannot skip gears.
For using the gas pedal, it looks like if you are not red lined, there are no circumstances where it would be better to not give it the gas. When red lined but at a time when your shifting plan calls for staying in the current gear, I can't tell from the manual what you should do. It says if you go past that you risk blowing the engine. What it doesn't say is if you get more power when going past the red line, so that it may be worth it.
Assuming that red line is max power, then when at red line but not shifting optimum would be to modulated the gas so as to keep it right at the red line.
For each shift pattern, it should be relatively easy to figure out when the gas has to switch between steady and modulated. Putting it all together that would give 1.3 million candidate inputs, one of which should be optimum.
If going past the red line does give more speed (at increased risk of engine blow out), it is more difficult. If engine blow out is probabilistic, you have to worry about the possibility that someone could get insanely lucky.
At that point, you probably need to analyze how the random number generator works, and how it is used, and how user inputs affect it. One particularly worrisome possibility would be that the timing of user input affects it. For example, if whenever the user presses or releases the gas it uses a random number to pick from several sounds associated with that action, that could be bad news. It would mean that the earlier analysis that the gas would optimally in each gear start out as constant and then switch modulated at the red line is no longer applicable. It could be that there is a run where if someone releases briefly the gas during the "constant" phase, that happens to get the random number generator to a position where they will get very lucky on the engine blow out rolls later.
If that meant we had to try every possible gas pattern, instead of just "constant followed by modulated", we'd have a combinatorial explosion.
Probably the best to do in this case (the "past the red line gives more speed, and the random number generator has a lot of user provided entropy" case) would be to tweak the emulator so that the engine cannot blow out, change the gas pedal input to simply be full gas all the time, and see what the results are. That would at least give us a lower bound on what the luckiest person in the universe could conceivably achieve.
My clouded recall from when I played this game was that the best times were obtained by quickly shifting to 4th gear and then waiting for it to reach the finish line. So "optimum shifting" may not be the way to get the best score.
You're exactly correct about the kind of analysis that was done. The code was analyzed, and the game was played frame-by-frame in an emulator to demonstrate perfect play. I believe some members of the speedrunning community also ran brute-force searches of the input space to make sure there wasn't some kind of obscure glitch that could trigger that wasn't obvious from the code.
The reason that Twin Galaxies didn't accept that analysis is because they're run by members of a clique of old-school gamers who stayed behind when the speedrunning community abandoned TG, and Todd Rogers is part of that clique. The analysis was largely done by Omnigamer, who is a well-respected speedrunner outside of TG but isn't part of the clique. TG's staff chose to believe their own in-group rather than a evidence-based challenge from an outsider until their hand was forced by a viral Youtube video[1] calling them out.
This kind of behavior isn't new for Twin Galaxies. They have a long history of mismanagement and corruption. Most of the times/scores posted on their site lack any publicly-viewable proof because until very recently, proof only had to be shown to TG staff "referees" so that people could keep their special techniques secret from the competition. Modern speedrunning eschews secrecy and requires publicly-viewable video proof for all runs.
"I seem to recall Walter Day had sold the rights to TG a couple years back. I used to work for TG some years ago, but hated it. My brother I and were constantly having to fight with what I call "the old guard" comprised of record holders from the early 80's. First, they resisted every one of our innovative new ideas like speed-gaming records for consoles, and then later we discovered a lot of the records from the '80s were bogus because video proof was not required back then. The worst was "Mr. Activision" (aka Todd Rogers), where I and several others were able to prepare a case and conclusively prove Todd had been completely fabricating world records for various Atari games. To name a few:
Barnstorming (2600): Todd's record, which stood for many years, was proven to be impossible once we broke down the game code and stripped the stage of any obstacles. With the stage completely blank, flying a straight line to the finish was slower than Todd's record. When we presented this evidence, we were attacked by fans and supporters of Todd, and eventually an excuse was cooked up that I lovingly refer to as "the coffee stain excuse". Yes, after being attacked and told we were clueless about how good Todd was, one of the referees covered for him and claimed the 'document' detailing his record had a coffee stain on the part where the record time was listed. Instead of throwing the record out and forcing Todd to do a legit one on video tape, they just simply adjusted the record to be MAYBE possible by adding a half-second to the time."
That is one hilarious and fascinating documentary. Considering when it was made, I found it incredible how there was a niche of society that had managed to come so close in capturing time in a bottle.
"It's unethical!" cried the group so obsessed with protecting those in their inner circle that they regularly find loopholes to keep from recognizing legit records.
There are some good mods at TG, but I find that many, if not most of them, treat it like an elite social club instead of the supposed impartial record authority that they promote themselves as.
For those who can't watch. Some games were scored in increments of 5 and he somehow got a score finishing with an 8. Many scores were ridiculous like 650000 for a game that goes in increments of 100 and the second highest video verified score was in the 90k range. Then the fact that all his records were referee verified instead of video verified. The referee is serving a sentence for child rape. Then there is the lie about the videos being taken by the police department and never returned which was proven false by a call to the police department. Then the fact that the record holder had direct access to the Twin Galaxies database and ent3ered his own records though he says he only did it once.
So, here's a thought experiment: consider "Carl H". Carl H ran the site computerscienceforeveryone.com. He was an incredibly passionate individual who spent his time creating free video tutorials on C programming. Along with these videos, he created supplementary materials, quizzes, and engaged with anyone who had questions. He even provided live one-on-one instruction to people over Skype, all for free. By all accounts, he was an excellent Computer Science teacher.
Carl was arrested for molesting his nine year-old son. He committed suicide in jail awaiting trial.
With that said, do we disregard all of the learning materials he created? Is their value inherently reduced to zero now that we know they were the product of a sick individual?
This isn't just an thought exercise. Carl was a real person. He had a Reddit board dedicated to his lessons and instructing people. His website has since lost its host and pretty much all the videos are gone from the web, save for p2p torrenting, because no one will be caught dead hosting videos from a child molestor, not matter how innocuous the videos themselves were.
Creating learning material isn't something that hinges entirely on trustworthyness. Sure authors of learning materials may be biased towards certain practices or styles, but it's trivial to verify that the content itself is legitimate - any coding tutorial that says 1 byte chars can increment up as far as 4 byte ints is going to get called out quickly.
But for a referee putting down high scores from a video game, the referee's value stems entirely from the idea that the referee is a trustworthy observer.
As the other commenters have pointed out, publishing knowledge is not the same as acting as an impartial referee or judge.
In the latter case, his inability to make fair decisions in life is directly relevant to his qualification for a job that involves making fair decisions affecting other people.
Unfortunately I don't think this thought experiment is very relevant.
Criminals can certainly have integrity in the absence of a moral compass but for this sort of crime, integrity is in fact lost.
Child crimes are crimes that involve the abuse of trust in most cases. Like embezzlement, someone trusts you not to steal money or molest children given a situation with ample opportunity for you to do so, but you do it anyway. You demonstrate a lack of integrity when you commit a crime of this nature.
So why we should also trust him to be a living authority of a particular record? He has proven that trust in him is misplaced.
People commit horrible crimes. That's one of those things that happen. It doesn't actually mean that everything they did was a lie or a crime, even if they are a horrible, detestable human in many senses of the word.
I'd nearly argue that it would be in his worst interest to participate in defrauding something since it might have revealed his horrible offenses.
The issue is that "child rapist" implies "poor judgment of character". This implication cannot be made because there are plenty of accounts where "non-child rapist" have "poor judgement of character".
No one is arguing a child rapist is okay; the issue other posters are having is that it is a incorrect argument.
Any further "but what if's..." are simply assumptions that do not have proof.
1. The issue is that "child rapist" implies "poor judgment of character". This implication cannot be made because there are plenty of accounts where "child rapist" have "good judgement of character".
or
2. The issue is that "poor judgment of character" implies "child rapist". This implication cannot be made because there are plenty of accounts where "non-child rapist" have "poor judgement of character".
being at least a logical (regardless of correctness) statement
My apologies, you are correct. I suppose what I was attempting to say was that the connection between video game referee and child rapist was a weak connection due to the relation of "judgement of character".
The statements "child rapist" implies "poor judgement of character" and "poor judgement of character" implies "poor ability to be a referee" can be negated if the person turned out to be a qualified referee. The gray area is what people agree to be what constitutes good/poor "judgement of character". This is where I believe what AdmiralAsshat is attempting to argue in his post "While that's deplorable, it does not logically follow that a child molester is necessarily an unreliable referee."
I wouldn't want to listen to anything a child molester has to say, but that doesn't mean their knowledge of something completely unrelated is suspect at all. Just that I really don't care to know what a child molester has to say, on any topic. Although, I agree with the video in that I would question his honesty and character... So maybe you're right.
Ironically, in the case of the Dragster record, his exact time does match the nature of Dragster's clock being only precise to 3/100ths of a second:
> Despite Omnigamer’s claim, as previously mentioned, Activision supposedly had created a machine-assisted run that determined 5.54 a perfect score, which is .03 better than Koziel’s model, thus making it impossible in his determinations. Crane explained to me that the “Dragster timer is .03 seconds per tick, so the difference that is being discussed is one tick of the game clock in a record set 35 years ago.”
This video stands by itself for all you need to know about the situation. It is pretty clear cut that this Rogers guy 'scum bagged' his way to his prominence when gaming was in a quite less developed state.
One thing that has maybe not been considered is that the Atari 2600 is susceptible to what's known in the online Atari community as 'frying', whereby cycling the power off and then back on very quickly while the console is running can lead to weird changes in the program state.
Not saying this is what occurred, but it could be a possibility, depending on the code...
Reminds me of a controversial technique discovered in the TAS community. Start saving your savefile, then cut the power mid-write. Usually creates a savefile with a bad checksum, but with certain timing on certain games, can result in playable corrupt savefiles which can be used for a faster win.
One of the reasons this idea is so controversial is the TAS community always assumes games should be played frame by frame, but this technique would open sub-frame "gameplay" (timing when you cut the power with much higher precision)
There are some games where you can modify the program state to glitch to the end, like Super Mario World: https://youtu.be/vAHXK2wut_I The explanation for the glitch is really interesting.
But something like this has to be given its own category. There's a Super Mario World category that allows for that "Credits Warp" glitch. The other categories have different rules for what you can and can't do.
So it's possible that Todd did something funky, but that would still make his scores illegitimate.
While that could be a possibility, that would still invalidate his score. It needs to be a full run and the game resets its memory at the start of each attempt so there'd be no way to start a new attempt with a "fryed" console.
So, is the opinion that he just fabricated the photo, or that he had a knack for figuring these glitches out? Because as far as I know, glitching things is considered all well and good with speed runners, and lots of his other high scores imply some kind of glitching or another.
Glitching is not universally considered "well and good" by speedrunners. There are multiple categories. Claiming you have a glitchless run that uses glitches is cheating. You may also run in a category that permits certain glitches, in the no-holds-barred category that must be run on real hardware, or the tool-assisted category, or any other game-specific category. Same deal.
This is the current state of the world. 10 years or so ago and before, there was a much stronger presumption of glitchlessness. Though you'd still encounter the problem that rigorously defining "glitch" can be quite challenging.
Twin Galaxies is notorious for banning glitches completely and it's one of the many reasons the speedrunning community abandoned TG. If he glitched, his runs would be invalid by TG's own rules.
As far as glitching in speedruns nowadays, it depends. You play by the rules of the category you choose to run, which may allow no glitches, some specific glitches or all glitches. To give an example, let's take a look at a popular speedgame, The Legend of Zelda: Ocarina of Time[1]. The "Any%" record, meaning beating the game with all glitches allowed, is 17m09s. The "Glitchless" completion record is 3hr48m38s. Categories like "No IM/WW" and "MST" are glitch limit categories where specific glitches are banned and the times are intermediate between any% and glitchless. Less-popular games will have fewer categories, but if there's interesting glitches, there'll usually be at least "no/few glitches" and "more/all glitches" categories.
In the original Super Mario Brothers, you could dash and then crouch and then slide through narrow tunnels while crouching. It wasn't documented in any official manual or anything, but everyone did it. Is that a glitch?
Different speedrun communities define "glitches" and decide which to allow/ban differently. There's always going to be some subjectivity.
In the community I'm most familiar with (Legend of Zelda: A Link to the Past), we assume that the devs did not intend for you to wiggle through single-pixel diagonal holes in sloped walls to get out-of-bounds or trigger the flute you may not even have by using a potion on the same frame you go through an east-facing door, and because those techniques allow you to play the game very differently than designed, we consider them major glitches and put runs using them in different categories.
There are several glitches that we classify as "minor" and allow in every category like item dashing and spin speed. Some people argue whether bomb jumps are glitches or not, but they're also allowed in every category.
At the far glitchy end, there's something called the "exploration glitch" that allows you to change Link's sprite's layer to one that allows you to walk semi-freely between map tiles. Exploration glitch is banned in all but a few very specific categories because it completely trivializes the game and lets you walk to the Triforce room within 90-120 seconds of starting the game.
The point of all this isn't to determine the single valid way to play the game (hi Twin Galaxies!) but to allow for sets of rules that make the speedrun more interesting. The same goal (reach the Triforce room) takes 1m31s with no restrictions, 26m50s with the "worst" game-breaking glitches banned and 1hr23m19s with all major glitches banned, and all three are completely different experiences.
Unless we're talking about something different than I'm imagining (like sliding through a wall or something), that seems obviously not a glitch. Mario has momentum that continues to apply when he changes to a crouching state and it would look really weird if it didn't. (I recognize that this is just proving the point that "glitch or not" is hard to define... and probably harder to define than one would expect.)
Ultimately the definition of what is or isn't a glitch isn't the most important question because it's often such a fuzzy question.
Like how some competitive gaming communities often have informal rules to keep competitive play interesting by banning specific overpowered mechanics, the more important reason for allowing/banning specific glitches or types of glitches is to keep the competition interesting for speedrunners and viewers.
Obviously some games are more interesting than others to speedrun, but like in the example of Zelda: Ocarina of Time, by ruling out specific glitches you can make sure that speedrunners continue to explore the surface area of the game rather than only trying to optimize a specific glitch with outsized impact on run-time.
Interesting article. For as long as I've been playing games there have been people claiming they did things which seemed unlikely if not impossible. However, given today's cycle accurate simulators and computation power it seems like you could certainly answer the question of whether or not the software would ever get to a certain point.
The game runs at NTSC rates, which we'll round up to 60Hz. We'll also say that 6s of inputs need to be considered, and scores which are above 6s will be discarded. We'll finally say that the possible input states consist of the joystick in either left ("shifting") or neutral ("popping") position, and the gas button either up or down. This generates four possible input states.
The number of possible input states in total will be 60 * 6 = 360, and so the total input space is 4 * * 360 = 2 * * 720, which is a little big for brute-force exploration. Many of these states suck; after all, the total number of shifts is not ~300, but ~5. So a guided search is likelier to pay off.
Edit: Thanks HN for eating asterisks and having no math mode.
There's brute force and then there's brute force. You know that the joystick inputs have to be applied in a certain order, and you basically want the throttle on unless you're above a certain RPM, so the search space is more like 60^4 with a couple of parameters, which is plenty tractable.
That depends if the best result is similar to a bunch of slightly less-good results, or if the best is a wild leap that doesn't look anything like other good results.
Contains some interesting insights about the Activision's spotty history of archiving game records. The most surprising thing to me about all of this is how Dragster's creator, David Crane, feels so sure that the record is sound: "I have no doubts, then or now".
The King of Kong film makes him look bad. He clearly fills a kind villain role.
In the movie extras he's seen delivering a Q-Bert machine to a excited senior citizen so she can practice for a tournament. Seemed oddly out of character from the way he was portrayed in the movie.
I don't know him personally, but he keeps showing up at video game events and clearly is enthusiastic about them.
The movie definitely makes him look a bit mustache-twirling Bond cartoonish but I've seen him at things like the Southern-Fried Gaming Expo and he does come across as pretty conceited.
There was a staggering amount of evidence, including TAS (Tool Assisted Speedruns) proving that Todd's time is physically impossible. That blew up in a 274-page thread, but ultimately I believe it took Robert T Mruczek's writeup to finally ban Todd.[1] Not surprising as Mcruzek is part of the "old guard".
I think this reflected very poorly on TG and I have no intention of using them in the future.
[1]: http://www.twingalaxies.com/showthread.php/175364-Dispute-Di...