Todd's 5.51 Dragster score

[Tempest]

I'm not even suggesting that a "super rare" rom exists. I am suggesting that only ONE ROM has ever been analyzed. A few random cartridge roms could be dumped and the code checked to see if it matches exactly. Try an early release of the game, maybe one that contained the sponge dust cover from the really early days of Activision. Try a very late release too. There could be a difference. Maybe since Activision determined that the best possible score had been reached on the game, they created a subtly different rom to create more buzz about the title among the avid gamers in the community of the day.

 

I have a hard time believing such a rom variation could exist so long without being found. Given all the rom collections and comparisons (there's a huge thread on AA) someone would have noticed a different Dragster rom and analyzed it by now. Is it *possible*? Of course. Is it *probable*? Not likely.

 

If you believe in Todd's integrity, then we have all the evidence that we need. For those that do not believe in Todd's integrity, will they ever have enough evidence?

 

Can you explain that one to me? If we think Todd is being honest then we have all the evidence we need even though people have basically proven that his time is not possible with any known version of the game? If we think Todd is lying then we'll never have enough evidence to prove it?

[D.Yancey]

 

1. I have a hard time believing such a rom variation could exist so long without being found. Given all the rom collections and comparisons (there's a huge thread on AA) someone would have noticed a different Dragster rom and analyzed it by now. Is it *possible*? Of course. Is it *probable*? Not likely.

 

 

2. Can you explain that one to me? If we think Todd is being honest then we have all the evidence we need even though people have basically proven that his time is not possible with any known version of the game? If we think Todd is lying then we'll never have enough evidence to prove it? Really?

1. I disagree. My statement was that if the rom differed that it would be so subtle as to be undetectable by observation alone.

 

2. That's not what I said. I said if you think Todd is lying, you'll likely never have enough evidence to change your mind.

 

Please don't make me sorry that I tried to add to the discussion. I've actually been playing Dragster since to 80's and I have a TG verified second place score on Dragster of 5.61. I also know Todd personally and find him to be trustworthy. I thought maybe I was minimally qualified to give an opinion. Peace.

Edited August 30, 2017 by D.Yancey

[Tempest]

1. I disagree. My statement was that if the rom differed that it would be so subtle as to be undetectable by observation alone.

 

2. That's not what I said. I said if you think Todd is lying, you'll likely never have enough evidence to change your mind.

 

1. But that's why we have code comparisons done by computers (clonespy and whatnot). They find the tiniest changes which are then examined by people to figure out what they do.

 

2. Ah ok. But I disagree with that since when Thomas posted his original (now flawed) findings that said the score WAS possible I was convinced and even said so.

[+stephena]

May I suggest that history may be repeating itself here. Activision originally created a computer simulation of a perfect run in Dragster to be 5.54. Along came Todd and broke that time. Activision scrutinized his performance. They even required him to repeat his performance. Activision found Todd's 5.51 to be legitimate. Activision even contacted Guinness and had the score submitted and verified by Guinness. Now 35 years later, it's happening all over again. May I ask exactly what is it that makes the data of today's analysts more reliable than that of the original programmers of the game at Activision? Why should any of us believe that they somehow know more than David Crane and the Activision team?

 

My own POV, for those who care. First of all, I have no stake in this at all, as I rarely play many games anymore, and concentrate on development of Stella, its debugger, etc, and above all its accuracy. So what some person got as a high score, and bragging rights, etc, are irrelevant to me. What I'm really interested in is whether the ROM that's being analyzed is the same as the one in the cart that Todd used for the alleged high score. If the ROM is different, then we need to find and analyze it. But if it's the same ROM, I wonder if there is something happening that allows the 5.51 to occur that can only happen on real hardware, and not in emulators.

 

So obviously my POV is wrt emulation, and I would be very interested to see if some undiscovered (up to this point) aspect of the real console could be influencing the results. Because from my POV, that would be something that definitely needs to be fixed in emulation.

[Tempest]

 

But if it's the same ROM, I wonder if there is something happening that allows the 5.51 to occur that can only happen on real hardware, and not in emulators.

That could very well be. Maybe something gets randomly set at power up? I think the analysis of the rom took that into account though (I seem to recall reading that all the important variables got reset when you started).

[+stephena]

That could very well be. Maybe something gets randomly set at power up? I think the analysis of the rom took that into account though (I seem to recall reading that all the important variables got reset when you started).

 

Based on my experience over the years, strange things can still happen even when all important variables are reset. The 2600 is a strange beast, and we're still finding weird and unexpected behaviour in the hardware 20 years after Stella was started, and 40 years after the original console was created. At this point, I rule nothing out. The only true test would be if someone can get 5.51 again, on real hardware. And a plus would be if they could do it in an emulator too.

[D.Yancey]

It doesn't help that in many people's eyes he already has sort of a black mark as some of his other scores seem to be impossible as well.

I don't think any of his scores have been said to be lies. Any of them that have been previously disqualified as world records in the TG database have been rom variations of early non-public releases by Activision or maybe even typos by Activision -- like Sky Jinks, Stampede, Barnstorming, and Starmaster game 4. I was the one who challenged Todd's scores on Stampede, Starmaster game 4, and Sky Jinks. (Starmaster game 4 has a bug in the scoring.) I even doubted some of his scores on Decathlon until I saw him play in person. I never doubted those Decathlon scores any longer! Todd has proven to me many times that he is quite honest. We should be looking for ways to support his score rather than ways to pile on more alleged evidence against his score.

Edited August 30, 2017 by D.Yancey

[D.Yancey]

 

Based on my experience over the years, strange things can still happen even when all important variables are reset. The 2600 is a strange beast, and we're still finding weird and unexpected behaviour in the hardware 20 years after Stella was started, and 40 years after the original console was created. At this point, I rule nothing out. The only true test would be if someone can get 5.51 again, on real hardware. And a plus would be if they could do it in an emulator too.

 

Thank you for posting this. I also rule nothing out. I was trying to point out too that only one rom has ever been analyzed on the emulator. And we cannot discern if there are subtle differences in roms just by playing the game as you well know. It doesn't even have to be a rare rom. It could actually be quite common.

[+stephena]

To be clear, based on the fact that Activision and David Crane verified this, I do believe that the highscore is accurate. But after reading the analysis by Omnigamer and further test by Thomas J. (and knowing his skill in this area), I also believe those analyses to be correct. So we have a conundrum: two equally opposing viewpoints from qualified parties. If we treat this as a simple boolean expression, there are two ways to accept this:

1. Either (or both) are incorrect. As I said, I don't personally think this is the case.
2. Both are correct.

The only way for both to be correct is if the ROM is different, or the analysis is proceeding from false assumptions. IOW, it is possible to have 100% impeccable logic and still arrive at the wrong conclusion, if your premises are incorrect. And that could be happening because something is happening on real hardware that isn't part of the analysis.

 

Anyway, I don't have much else to say on this. If someone finally figures this out, it may lead to a better understanding of the hardware, and better emulation of it. And that's all I really care about.

[Tempest]

Any of them that have been previously disqualified as world records in the TG database have been rom variations of early non-public releases by Activision or maybe even typos by Activision -- like Sky Jinks, Stampede, Barnstorming, and Starmaster game 4. I was the one who challenged Todd's scores on Stampede, Starmaster game 4, and Sky Jinks.

 

Once again, is there evidence that all of these were either early roms or typos? Maybe I'm being too hard on Todd, but it all seems too convenient that all his questionable scores can be explained away by typos and pre-production roms.

[Tempest]

Anyway, I don't have much else to say on this. If someone finally figures this out, it may lead to a better understanding of the hardware, and better emulation of it. And that's all I really care about.

 

I agree. I'm in this as much for the mystery as for the technical explanation as to why it is or isn't possible. I find it fascinating.

[Goochman]

To be clear, based on the fact that Activision and David Crane verified this, I do believe that the highscore is accurate. But after reading the analysis by Omnigamer and further test by Thomas J. (and knowing his skill in this area), I also believe those analyses to be correct. So we have a conundrum: two equally opposing viewpoints from qualified parties. If we treat this as a simple boolean expression, there are two ways to accept this:

1. Either (or both) are incorrect. As I said, I don't personally think this is the case.
2. Both are correct.

The only way for both to be correct is if the ROM is different, or the analysis is proceeding from false assumptions. IOW, it is possible to have 100% impeccable logic and still arrive at the wrong conclusion, if your premises are incorrect. And that could be happening because something is happening on real hardware that isn't part of the analysis.

 

Anyway, I don't have much else to say on this. If someone finally figures this out, it may lead to a better understanding of the hardware, and better emulation of it. And that's all I really care about.

 

I could be reading all the evidence wrong but I believe Thomas has shown that Omnigamers evidence is not complete. THomas verified that if he followed Omnigamers execution path to a tee that he got the same result. This would make perfect sense since Omnigamers dissection is of the actual code. We have to assume to Omnigamers path is the absolute perfect path.

 

Didnt Thomas find a path that got him 5.54 - matching Activisions theoretical best score? If this is the case then how does this align with Omnigamers theoretical best of 5.57?

 

I believe it was also stated that if one could start in 2nd gear that better then 5.54 is a theoretical possibility also?

 

Real hardware is going to behave differently than emulation - look at the Concerto problems on the 7800 - depending on when it was manufactured certain games work and others will not. Im with Yancey that in this particular case Activision validated it - not some TG Ref or gaming magazine, but the actual developer of said game.

[D.Yancey]

 

Once again, is there evidence that all of these were either early roms or typos? Maybe I'm being too hard on Todd, but it all seems too convenient that all his questionable scores can be explained away by typos and pre-production roms.

 

I have photographic evidence of the Stampede rom difference which I am not willing to share. The Starmaster game 4 was proven to have a scoring bug, so if you can get a high enough score, you can see for yourself. I do not have 1st hand evidence of the Sky Jinks rom difference, but I do believe it to be true. That's the best I can give you, for what it's worth.

Edited August 30, 2017 by D.Yancey

[Goochman]

Demon Attack has different ROMs also

[+stephena]

Real hardware is going to behave differently than emulation - look at the Concerto problems on the 7800 - depending on when it was manufactured certain games work and others will not. Im with Yancey that in this particular case Activision validated it - not some TG Ref or gaming magazine, but the actual developer of said game.

 

Yes, but I want to know how the real hardware is differing, so that I can properly emulate it. The actual highscore is quite irrelevant to me, only that it seems to have happened yet defies logic of ROM analysis. From my POV the why is important.

[Omnigamer]

To be clear, based on the fact that Activision and David Crane verified this, I do believe that the highscore is accurate.

 

Just to clarify, only Todd Rogers' statements indicate that David Crane personally checked out the score. David has no recollections of such a thing now, nor did he actually create an AI with any real confidence it was correct or complete. Those are direct statements from him in June of this year.

 

The core mechanics are well-enough understood at this point that the opportunity for other things to allow a better time are pretty slim. Outside of stuck-at-faults in the memory, I don't see much opportunity for affecting game state from the wider system. The other possibility is some oddities with instructions in the 6507 that aren't documented anywhere, but I think it's pretty unlikely they would have gone unnoticed for so long. The last option is that a particularly obtuse bug exists that allows Dragster to start a race with a "dirty" memory state, but so far I can see no way this can be accomplished just based on human inputs, even to system buttons like reset and select. There is a routine in the ROM that wipes nearly all gameplay addresses before starting a race.

[Tempest]

I have photographic evidence of the Stampede rom difference which I am not willing to share.

Any reason why? What was the difference in the rom that allowed the score?

 

 

The Starmaster game 4 was proven to have a scoring bug, so if you can get a high enough score, you can see for yourself. I do not have 1st hand evidence of the Sky Jinks rom difference, but I do believe it to be true. That's the best I can give you, for what it's worth.

Right, the Starmaster one is well known so that one has a legitimate explanation. What about Barnstorming? Any evidence for that one? I believe someone proved that even if you don't have to move the plane the score wasn't possible (unless my memory is wrong on that).

[Goochman]

 

Just to clarify, only Todd Rogers' statements indicate that David Crane personally checked out the score. David has no recollections of such a thing now, nor did he actually create an AI with any real confidence it was correct or complete. Those are direct statements from him in June of this year.

 

The core mechanics are well-enough understood at this point that the opportunity for other things to allow a better time are pretty slim. Outside of stuck-at-faults in the memory, I don't see much opportunity for affecting game state from the wider system. The other possibility is some oddities with instructions in the 6507 that aren't documented anywhere, but I think it's pretty unlikely they would have gone unnoticed for so long. The last option is that a particularly obtuse bug exists that allows Dragster to start a race with a "dirty" memory state, but so far I can see no way this can be accomplished just based on human inputs, even to system buttons like reset and select. There is a routine in the ROM that wipes nearly all gameplay addresses before starting a race.

 

There are ways of 'frying' a 2600 with repeated on/off attempts - not sure if you were aware of this. You can get some pretty cool game variations on some ROMs. This is effectively cheating if one were to get an advantage this way, but on real hardware you can impact ROM execution in some limited ways.

[Tempest]

Demon Attack has different ROMs also

Which has been well known for several decades now. This is why I have a hard time believing that a rom variation like this could hide for so long (although I concede it is possible).

 

 

 

Just to clarify, only Todd Rogers' statements indicate that David Crane personally checked out the score. David has no recollections of such a thing now, nor did he actually create an AI with any real confidence it was correct or complete. Those are direct statements from him in June of this year.

 

He could have just forgotten as it was over 35 years ago. Still, that is a bit troubling.

[Omnigamer]

 

I could be reading all the evidence wrong but I believe Thomas has shown that Omnigamers evidence is not complete. THomas verified that if he followed Omnigamers execution path to a tee that he got the same result. This would make perfect sense since Omnigamers dissection is of the actual code. We have to assume to Omnigamers path is the absolute perfect path.

 

Didnt Thomas find a path that got him 5.54 - matching Activisions theoretical best score? If this is the case then how does this align with Omnigamers theoretical best of 5.57?

 

I believe it was also stated that if one could start in 2nd gear that better then 5.54 is a theoretical possibility also?

 

Real hardware is going to behave differently than emulation - look at the Concerto problems on the 7800 - depending on when it was manufactured certain games work and others will not. Im with Yancey that in this particular case Activision validated it - not some TG Ref or gaming magazine, but the actual developer of said game.

 

Thomas was engaging in an idealization of Dragster, not the game itself. He simply placed shifts at the speed ceiling caps and observed what time fell out. I run through the same exercise in this video, however I misquote the speed cap for gear 3:

 

Getting that pattern in actual gameplay is not possible due to decays on the tachometer value and other issues which will inhibit acceleration.

[+stephena]

The core mechanics are well-enough understood at this point that the opportunity for other things to allow a better time are pretty slim. Outside of stuck-at-faults in the memory, I don't see much opportunity for affecting game state from the wider system. The other possibility is some oddities with instructions in the 6507 that aren't documented anywhere, but I think it's pretty unlikely they would have gone unnoticed for so long. The last option is that a particularly obtuse bug exists that allows Dragster to start a race with a "dirty" memory state, but so far I can see no way this can be accomplished just based on human inputs, even to system buttons like reset and select. There is a routine in the ROM that wipes nearly all gameplay addresses before starting a race.

 

I'm not willing to get into any deeper analysis, since I don't have the time (Stella takes pretty much all of it). But based on my somewhat unique experience as an emulator developer, I will say that we are finding out new things about the system every day, even 40 years later. So just because it's been a long time since the console was introduced, that doesn't necessarily imply that we won't find something completely new tomorrow.

 

In implementing the new TIA core for Stella 5, we discovered things that were never documented anywhere or emulated by any other emulator. And as recently as just a few days ago, we had a discussion on how 'FE' bankswitching works, and how it differs from all known documentation and implementations in other emulators. So I can't be confident that we know everything about the console. And as we've discussed on the Stella team, we won't know unless/until someone does a gate-level emulation of the hardware. And even then, there were multiple variations in the original hardware, with different tolerance levels, etc.

 

I'm just being devils advocate here. We are all still learning about the quirks of the system, and a year from now will probably still be learning new things. Maybe this has nothing to do with this ROM and highscore, but it can't just be discounted just because we haven't discovered it so far.

[D.Yancey]

1. Any reason why? What was the difference in the rom that allowed the score?

 

 

2. Right, the Starmaster one is well known so that one has a legitimate explanation. What about Barnstorming? Any evidence for that one? I believe someone proved that even if you don't have to move the plane the score wasn't possible (unless my memory is wrong on that).

1. If Todd had wanted the explanation public, he would have explained it. I cannot give you further explanation. You will have to ask Todd in private about that.

 

2. I do not have an explanation for Barnstorming. My opinion is that is was a typo or simply remembered incorrectly. I am glad it was cleared up and so is Todd. The score of 32.04 was not possible. It was changed to 32.74, which I totally believe.

[+stephena]

Thomas was engaging in an idealization of Dragster, not the game itself. He simply placed shifts at the speed ceiling caps and observed what time fell out. I run through the same exercise in this video, however I misquote the speed cap for gear 3:

 

This is what I was referring to earlier. It is possible to do an in-depth, impeccable analysis of a problem, and still arrive at a result that is incorrect. I agree with your logic (inasmuch as I've looked at your spreadsheet). I believe it makes perfect sense. But it also kind of implies that the hardware is working a certain way. If the hardware is working in the way you imagine, then your analysis is absolutely correct. But based on my experience, we can't always be sure that the hardware works as we expect. That is the assumption that may be incorrect.

 

Anyway, enough discussion; I need to get back to working on Stella.

[Tempest]

1. If Todd had wanted the explanation public, he would have explained it. I cannot give you further explanation. You will have to ask Todd in private about that.

It's things like this that just keep churning up the controversy. You seem to indicate that Todd had access to some pre-production or early rom variant that allowed a higher than normal score, but he won't make any information on it public. That seems really fishy to me. Why keep something like that secret?

 

Don't misunderstand me, I'm not blaming you for keeping what you know private if Todd told you to do so. I'm just saying that the whole 'secret explanation' thing only makes Todd look like he has something to hide.

 

Am I really the only one that thinks all of this sounds really odd? Non-reproducible scores, mystery rom variations, typos, swap gas reflecting off Venus? I want to believe and all, but it's really REALLY getting hard to.

 

I'll let it drop, as I've said all I really can and I don't want to seem like I'm trying to stir up a witch hunt.

[Omnigamer]

 

This is what I was referring to earlier. It is possible to do an in-depth, impeccable analysis of a problem, and still arrive at a result that is incorrect. I agree with your logic (inasmuch as I've looked at your spreadsheet). I believe it makes perfect sense. But it also kind of implies that the hardware is working a certain way. If the hardware is working in the way you imagine, then your analysis is absolutely correct. But based on my experience, we can't always be sure that the hardware works as we expect. That is the assumption that may be incorrect.

 

Anyway, enough discussion; I need to get back to working on Stella.

I agree with you that there could always be something else. That's part of being a proper scientist - open to new evidence and theories wherever they pop up.

 

For the time being, there's been no empirical evidence that what we've modeled of gameplay is wrong. Every past video that I've checked conforms to the model. Unless and until there's some actual evidence of something that the model doesn't support, it's the best knowledge we have of how the game and its ecosystem function. I look forward to anything new on that front so that it can be further refined.