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.
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.
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.
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.
That's fair - although I was mainly referring to the era of arcade games referred to in the original blog post, and not to console speedruns and so forth that came later and cater more towards the 'perfect run' style of play. In the late 1970s/early 1980s there really weren't a lot of games that have 'perfect' runs. You simply can play the majority of them forever once you learn to play well enough since they repeat the same levels over and over. There are some exceptions like Donkey Kong and Pac-man that have killscreens, sure, but personally I don't find that 'beating the system' to play the same levels over and over to be very enjoyable. Since I'm not spending my allowance on the game, I want the game itself to be interesting rather than just long.
I do enjoy some games of this type like Robotron that have enough variation in the random patterns of the enemies to still feel fresh. I also love games like Pole Position or Outrun that have well-defined endings and force you to play 'perfectly' like you describe. However most games of this earlier time period are unfortunately just not designed that way.
World Warrior is a bad competitive game. No doubt. Champion Edition is only slightly better. But that doesn't really take away from my point - a random person walking up to the game is not going to know how to do the CE Dictator dizzy/redizzies for example, nor know how to play lockdown defense with CE Guile to prevent them.
And it's not as simple as just 'memorizing' the combos if you never play well enough to get yourself in a position to land them. There's still a journey there both outside the game to learn and inside the game to break down the patterns in your opponent's play that will never be present in a single-player game. That's what keeps fighting games interesting and new 20+ years later for me.
If you think fighting games don't involve skill try beating an experienced Street Fighter 2 player having never played the game. When you lose, and lament that you 'simply haven't memorized the combos' go home and look them up. After you've realized that there's only one or two major combos to learn for each character in that game, come back and I promise you'll still lose. There's much more to it than that.
Most importantly, arcade games by their nature are finite. The limits of whatever storage medium used as well as the inability to rewrite or update them easily means that they exist in a vacuum. Once you've 'explored' the levels or found the 'secrets' they cease to be new. There are no more secrets. The strategy to beat single-player games never changes.
However the person you're playing against in a fighting game is always going to be different. Even if they play the same character as another person they will play them differently. They will have different skill levels and unique weak points to attack. No round of a fighting game is ever exactly the same. This is why fighting games are beautiful. They turn a static game into something timeless. There are still weekly Street Fighter 2 tournaments all over the world. When was the last time you competed with tens of other friends in say, Qbert 25+ years later?
I've used Pokertracker Stud in the past, and I recall they had an Omaha version as well. This was ~10 years ago however, not sure if either are still available.
Theoretically the best way to maximize an edge is to just wager larger amounts at all times. This has a beneficial side effect of deviating from 'standard' human play which can be further exploited, although I doubt this the intention of the AI.
I have issue with the claim that Taylor was just beating up on inferior competition historically. One way we can measure this is by the 'three-dart average' - the average score of each player's turn of three darts at the board. Taylor had the televised world record from 1991-2004 and from 2008-2012. He was leagues better than anyone on the planet. Van Gerwen is in my mind a better player now but this article really sells short how utterly dominant Taylor was. He wasn't just better than everyone else, his opponents weren't even playing the same game.
I work with Cumulus nearly every day if anyone has questions about what the experience is like. They just had a big major release recently if anyone was using the 2.5.X branch before.
In theory, an Omaha bot wouldn't have to be as advanced as a hold'em bot simply because the 'theory' of that game is not as strong as hold'em yet. Omaha still attracts a greater percentage of losing players as well.
To put it extremely simply, there are specific ranges of acceptable statistics about what winning play looks like in poker. With thousands of hands of data winning players can intuitively pick out stats that are out of line.
For example, there are two related stats: went to showdown, and won money at showdown. These calculate how often you get to the last card and reveal your hand against another player/group of players to win the pot, as well as the frequency you actually win these showdowns. Generally, the more you are showing down hands against opponents, the lower your chances of winning money, simply because it's not always possible on average to have better hands than your opponents above a certain frequency of times you're getting to showdown. If a bot/player had access to other players' hole cards (like actually happened on the Ultimate Bet network) a telling example might be that they go to showdown way more than average while also winning money at showdown far above an average rate.
There are tens (if not hundreds by now) of common poker stats, this is just a simplified look at one related pair.
For my money, in terms of accessibility and fun while having a steep but manageable difficulty curve - the 2600 game that stands up the best today is Kaboom. It can induce an almost zen-like state that is very hard to capture in gaming.