Great article. The dilemma of fairness versus playability seems to be of increasing importance as the casual or semi-casual gamer garners more and more attention. One thing that I would point out however is that a lot of the "fun" that comes from games derives from the feeling of learning and then mastering a new skill. This is widely accepted. Anytime we as game designers effectively change the rules of the game in response to how the player is doing, we risk losing that feeling of progression toward mastering that skill.

The system laid out in this article is certainly well thought out and the idea of a script for a race is interesting. I wonder if any other scripts were used other than the one mentioned in the article? It would seem that having every race play out in a fairly predictable way (assuming the player is consistant) is not necessarily a good thing.

I do give the author and his team praise for having obviously put a lot of thought into the problem. We all remember how frustrating it was to hit your rival with a beautifully timed shell in the original Mario Kart, only to have them be back on your tail in three seconds.

Excellent design article and kudos to the author and their team for developing an system that's responsive and non-rule breaking. It makes me want to play Pure to see the system in action so I can study it in more detail.

In some ways, it seems similar to the AI Director in Left 4 Dead in that rather than enemies and opponents being basically mobile obstacles, the enemies are thought of as part of the gameplay. Similarly, the experience of playing is designed around the player rather than around an arbitrary difficulty curve - someone playing on the highest difficulty but not very good relatively can still get a very challenging experience without feeling cheated. And it remains challenging as they get better with their skills coming from learning tricks, behaviors, and game systems rather than necessarily quirks in the difficulty system.

Great article. I like the idea of groups of racers that players have to progress through, which them gives some good peaks to the racing action.

This is an excellent article describing an ingenious alternative to vanilla rubber-banding AI. This method is great for maintaining excitement and general entertainment for the casual to hobbyist gamer.

As a hardcore gamer, though, I feel that any sort of rubber-banding AI which artificially impedes progress and marginalizes mistakes. Rubberbanding is so hated because you can only be so skilled at a game which implements it until no additional skill will make an applicable difference (with the exeception of time trials).

Now I forewarded this comment with my thoughts that this is an excellent strategy, because I really do believe it is. For lower difficulty levels or semi-causal gaming, ideas like this are what makes good games into GREAT games. In my opinion, though, if you wanted to either market the game to the hard-core niche or include an extremely difficult difficulty level, take a note from rogue-likes and treat the player like a masochist.

Long story short: Great article, great algorithm for what you were aiming for and I'd love to read some more of your ideas.

*******************
Note that for reference I worked for close to 10 years in a senior design, lead design and creative direction capacity for 8 versions of Road Rash, I developed all the vehicle physics and mechanics systems for Scarface and I spent the last few years at THQ working with various internal and external racing developers in a Creative Direction capacity so suffice is to say I have some racing development insight.
*******************

Bravo, Eduardo!

Kudos for calling out the outdated and archaic rubber banding methods of the 1990s and striving to evolve racing AI beyond those methods that are still frightfully used by 99% of racing developers (though to varying levels of success).
I was most pleasantly surprised and encouraged with your description of leveraging your Dynamic Skills System to control both Difficulty and race Pacing ("fighting for the first position the last 20-30% of the race"). Sadly Pacing structure appears to be so rarely considered in racing games even in this day and age nearly 20 years into console gaming and it is always an area that I frequently advise racing developers to improve upon.

I am continually surprised every year with the newest racing games that still use the heavy handed and one dimensional rubber banding methods and I have been even more surprised when multiple racing developers I have worked with the past few years have had little interest in correcting the rubber banding limitations that I already knew how to solve. On both Road Rash 3D and Road Rash: Jailbreak (of which I was Creative Director) we did away with the lame one-dimensional rubber banding techniques in favor of Dynamic Speed Management and the results were the best delivery of racing competition in any of the 9 versions of RR or IMOHO of any of the other games on the market since then (sadly we still got beat bad on graphics once the original, market-changing Gran Tourismo came out two months before we shipped).

For RR: Jailbreak the lead AI programmer and I evolved the Dynamic Speed Management system further to treat the pack as a whole like one long accordion which can be expanded and contracted by adjusting target speeds of the first and last rider ahead of the player and first/last behind the player (interpolated for speeds in between). We also heavily limited the best case speed advantage of the opponent bikes to 2% so they will never blow past the player who is at or near top speed.

All the speed management parameter settings would change based on race progression (discreetly for each 1/4 of the race) and on player performance (race place). When play balancing these systems I consciously formulated the Pacing I wanted - one that spread the pack out in front of the player at the start, made it easily to pass each rider one at a time with at least one more opponent in view for the 2nd and 3rd quartiles, then later in the race contracted the pack so there were often 3-5 bikes jockying for position near the finish line when the player is doing well (in the top few places). Because the settings were tied to player place rank we could adjust the settings for various performance and ability levels to increase the sense of excitement and competition towards the finish line but without making it too easy for a player who did really, really bad; my criteria for tuning was that a player who crashes twice in the first 70% or so of the game should still have the opportunity to finish in 1st place, but certainly not every time. The tuning trends of my Pacing structure was maintained from level to level in the racing mode but the competition was increased sooner and more fiercely at higher levels and player mistakes were made more costly. Each of the other modes (Five-O, etc.) had their own unique pacing structure and supporting tuning trends.

These systems proved to deliver exceptional competition game play in the single player racing mode IMO because they ensured the player always saw other opponents during a race (which is harder to do when there are no rear-view mirrors). It also kept the opponents from bunching up into large swarming groups around the player (which will prohibit passing) and it prevented the annoying slingshot of opponents that all of a sudden blow past the player when they are already at top speed (which would otherwise expose an unfair speed advantage). One of our key philosophies was to never make it appear the opponents had any unfair advantages so in addition to capping their top speed targets they would also slide out around turns or crash into cars at roughly the same rate as the player.

I believe strongly that Dynamic Difficulty Systems in all genres are the wave of the future because they allow designers to make a larger sweet spot of fun for a wider range of ability and performance levels AND because they can be leveraged to adjust the Pacing of a game. So despite what you read from Ernest Adams (who can have interesting things to say but the bulk of his experience lies in programming and not so much in design), Dynamic Difficulty has a huge future in racing games and racing developers would be wise to follow Eduardo's lead on Dynamic AI systems in order to strive to improve the quality of the single player experience through the fun of competition in the years to come. I believe the racing genre is far behind the shooter and platform genre in terms of AI development so I look forwards to a time where racing developers start to implement and iterate new types of racing AI. It would be cool to see new AI systems innovations coming from the racing genre instead of only from the shooters of tomorrow.

@Jake Romigh

I believe your logic of Dynamic Difficulty not supporting hard core gamers is flawed as it is really the execution and tuning/balancing that determines the challenge rate. It might be the case that it does not support those who prefer technical, heavy-sim driving like that in Gran Tourismo, but my argument is sales trend in that sub-genre have been dramatically decreasing the past 3-4 years.

See my previous comment on Dynamic AI Speed tuning but the key for us was in tuning the game to be challenging for a player who is doing exceptionally well early on in a race and I think we delivered that as well. There were hoards of rabid and hard core Road Rash fans who were delighted with the competition experiences in RR: 3D and RR: Jailbreak. [Agreeably RR was not a heavy, technical sim though our games did always utilize a sophisticated physics simulation under the hood.]

The very advantage of a Dynamic System (as opposed to one dimensional rubber banding) is that it can increase the sweet spot of success, challenge, and excitement for a much wider range of player ability and player performance. The main limitation with heavy technical simulations is that they are tuned to be exceptionally hard and so only support a very narrow range of the hardest core racing gamers. With Dynamic Difficulty you can support all types of players since you know what their performance is at any time throughout the race and you could for instance make it significantly harder for the best player at the same time you make it a bit easier for the worst player.

Hi,

First of all, I want to thank you for the interest showed in the article. I'm really glad that you liked it and find it interesting and, maybe, useful.

Next I'd like to say that the most important part of the success (I want to think it was such) of the AI in Pure was due to the involvement of the team. We had a very talented team with a clear focus and that showed clearly in the AI (as well as many other aspects of the game). From the top to the bottom, the people involved in the AI understood what the necessities were and proposed clever ideas that helped shaping the system the way it is. The team didn't think it was enough to have a rubber band around the player, understood the importance of a good, challenging, fair AI, wanted to deliver the best possible and allocated the time and resources necessary to do so. It was really a team work what led us to develop the system explained in the article. And I must say I'm very happy to have been part of it.

Regarding the lack of dynamic balancing in games, I must agree it's sometimes socking how little effort some games put in challenging the player no matter what and no matter how good he/she is. A few of us spent a night playing a few games together against the AI to try to pick up what they were doing and trying to take some references for our own project. We were trying what then we thought were the reference racing games and we couldn't find many useful tips. The dynamic adjustments were either too subtle (mostly in simulators) or too exaggerated and unfair (in the most arcadey games). We really wanted to appeal to anyone, from the very hardcore gamers to the most casual players. And we thought we had to get to something more sophisticated than what we were seeing there.

I agree that the racing genre hasn't developed in the AI field as much as some others and I think there's still a lot to do here and a lot to learn from other areas.

I finally would like to say that this system still allows for very hardcore difficulty settings. We decided not to offer that possibility for a number of reasons, but not because the system didn't support it. If you set the initial skill for all the players to a huge value (2.0 for instance), since you are restricted to the 0..1 range you will be forcing all the players to be as good as they can be. That means they won't be forming groups and they will ignore the race script, but you will have a very challenging and very unforgiving AI. Whether or not that's what you want for your highest difficulty setting in your game is to be discussed somehwere else.

@ Mike:
I agree that my proposed "hardcore difficulty" is not for the general gamer. The hardcore niche is exactly that, and moreso a vocal minority. If you made a racing game without Dynamic difficulty and expected it to sell well, of course you'd be wrong; the market just does not want that sort of thing. I agree wholeheartedly.

My point was that if you WERE crazy enough to cater solely to the hardcore player and include the best possible AI drivers the game could support, any sort of AI which makes it easier for the player or for itself is a frustrating prospect. Furthermore, it is simply unfair when the same increase and decrease is applied to player cars in head-to-head mode, just for the sake of assuming it would be more fun that way.

Now, I admit, for the huge amount of the market, no one wants to play games where one or two mistakes forces them to finish 3 or 5 laps miles away from any competitor. That's a ridiculous notion to have if you're trying to sell a video game; in the end, that's what we're all trying to do, right? I just want to say that if you were trying to make a mode where the playing field is level for head-to-head or AI competitors to market to the hardcore player (who takes solace in fact that their lead will not be artifically shortened or whose lost ground is only made up by skill, not a charity boost of speed/control), you'd want to watch what you do with this. After reading Eduardo's great article and insightful response, I fully believe Pure would be able to replicate this, but for most likely good reasons (judging by the great reviews and sales that Pure has been getting, congratulations guys!), it was not included.

That is just my opinion though, as I'm mostly a indie developer, so I don't have the vault of experience the people I'm having a discussion with. Anyone who's reading this and think I'm full of crock, well, that just may be.

One last update, because I doubt anyone else is going to be actively looking at the comments:

A kind of reference is the February 2, 2009 Patch for TF2 released from Valve. If you know anything of how TF2 calculates damage from guns, you know it's a random amount in a range determined by what gun you are using and the distance from the target. Also, you'll have a random chance to crit, doing 3x the damage to your opponent. The community made many suggestions, from the way crit chance was determined and how to implement a mode for the "competitive TF2 player", which was meant to standardize the encounters so that it was mostly based on player skill and not semi-random damage. Valve read these comments and made non-default options so that players could create their servers to be fined tuned for this vocal minority through a revamped "tournament mode".

So after three posts of lengthy discussion and well-thought out rebuttals from actual industry brain tanks, my last few statement are simple:

There will always be a minority of players who want their experience to be solely based on their skill and the skill of their competitors, and artificially changing their competition will be unwelcome. The main player base (and the majority of your sales) will not care about these squabbles, but including the option (certainly not default) to have these changes in gameplay will show your interest in the community and, in my opinion, therefore increase the longevity of your product.

@ Jake

Well said.

I would go further to state that the heavily technical simulation is an even smaller niche of overall hard core racing than in can be said for the frag master in shooters. I would also argue that technical sim style of racing is much more about perfection of vehicle operation vs. the track and in that case I actually see most AI as mostly irrelevant to the bulk of that experience except in passing and final rank. I have always designed a more mainstream racing games to cater to those who want to master the vehicle mechanics and also to feel they outsmarted and outmaneuvered the AI opponents in a series of micro events that vary from race to race.

Shooters have a larger base of hard core players than racing does and it is of course a larger genre. Making a heavy technical sim that is not Gran Tourismo in this age is suicide IMO (not that you are proposing that); witness the disappointing performance of Baja where I fought the developer unsuccessfully to make the gameplay more mainstream and accessible in learning curve, AI and sliding recovery. Despite a mostly excellent physics model the shortcomings and the ultra-niche appeal, accessibility and usability of a technical sim off-road game outweighed the pros.

Great, incitefull article.

That said, any sort of technique that evens up races like this frustrates me so much. I would much rather win a race on my skill than because the AI has been downgrade or my abilities upgrade. Perhaps the most frustrating aspect of this is that when you're okay at the game, you'll be winning many races. When you're great at the game, you'll still be winning the same amount of races. I expect that if I win when I'm okay, I should kill the computer when I'm great. There is a certain amount of satisfaction of being WAY out infront that any sort of technique like this prevents.

Great Article,

A example of how AI it's not just about programming skills, we can see a lot of game design here too!

I believe that this dynamic evolution system cannot be perceived and work well to casual/semi-casual gamers, who like a casual/non-so casual race game (like Super Mario Kart, Micro Machines, ...), but probably will be perceived by a hardcore gamer who want a more "simulation race" style of game (like Gran Turismo, Need for Speed, ...) and will be frustated.

One game that probably use a similar mechanism is Super Mario Kart, i can remember me playing the game with friends in SNES and in Nintendo 64 and commenting that all the time you are doing a bad race you get rewarded with a very good bonus (like thunderbolt and star), when you are in the first position you just get bad bonus (like bananas), and the cars who are in bad positions runs fast.

@ Fernando

Need For Speed is far from a hard core simulation title. On a scale of 1 to 10 with 1 being the extreme technical simulation and 10 being the extreme Arcade experience I would give NFS a 7. Now days I would give Gran Tourismo about an 2.5 or 3 (the original was closer to a 1) and Project Gotham maybe a 3.5.

Also not all hard core racing fans are technical simulation fans. Typically the hard core description is not attributed directly to the play style preference but to the experience and time dedication of the gamer and while there is clearly a correlation between hard core and technical sim fan they are not equal. In conducting multiple Usability and Playability Test sessions on Baja I experienced a mix of preferences in hard core racing fans. Although the preferences of hard core were weighted a bit towards technical sim there were a reasonable amount of hard core racing fans who play racing games 15-30 hours per week and have mastered 4-10 racing games over the previous year or two but who preferred Need For Speed and Burnout hands down to Gran Tourismo and Project Gotham. These players rarely cited the technical sims as positive examples of gameplay in the discussions and they tend to thrive more on the competition of opponents in both single player and multiplayer modes than the technical sim fans who thrive on achieving perfect track mastery and lap times.

Note also that a racing game can have a strong physics simulation but non-technical gameplay that does not focus heavily on the perfect line and lap times. These racing games are typically poorly described as arcade racers which shows the limitation of a simple 2-term category descriptor; the same single scale category limitation exists with Dirt where the physics are very arcade-y but the gameplay is very technical (perfect line and time focused).

I have actually been formulating the outline for an article on the need for better marketing categorization of racing games necessary to clarify the play style to consumers and to better educate the audience/industry; I content that physics style (arcade vs. sim), race focus (technical vs. competition) and track style (loop vs. point to point) are three separate scales that should be used throughout the industry to accurately categorize any racing game and to educate the public on the gameplay focus.

Comment on racers... simulation vs arcade

Games are supposed to be entertaining, the reason arcade racers sales have grown and sims have declines is simply because the numbers of gamers who prefer and know what FUN and ENTERTAINMENT is outnumber the "simulation racers" (note I use that term loosely to refer to people who prefer hardcore perfectionism of track times/perfect lines, etc, over other aspects of the game that add way more to the entertainment value - like burnout and it's spectacular crashes and stunts).

Simulation can be fun if it is done well and containing the game killing simulation elements. Need for speed porsche Unleashed is the best and perhaps ONLY example of a game that bridges simulation and arcade in a perfect way, which I'm guessing did not sell as well as hoped due to Need for speed moving into total more arcade territory.

Almost all movies have elements of unreality or complete myth and illusion, what is entertaining is drawing out and focusing on the experience of what is fun and not merely trying to spoonfeed the player what amounts to a joyless experience.

Need for speed series for the most part nailed the fun part, even the bad games in the Need for speed franchise are still OK racers even when they are bad by measures of previous games in the series.

Racing as a genre is quite limited, I found the article on AI intersting. But the reason why mario kart is so fun is the fact that it keeps the tension between players alive through the use of powerups. Regular racing games have to focus on something else to keep the players engagement and interest up or focus on the aspect that driving is itself enjoyable and providing the best gameplay mechanics to have a fun experience with.

I seem to recall playing Need for Speed (I think it was "Underground"), the rubber banding was attrocious in that game. It was actually an advantage to be significantly behind the field at the start—even up to a whole lap depending on the race—then you could use the rubber banding to "slingshot" past the other cars at the end and by the time your opponents speed had a chance to correct itself, the race was over! Excellent article, clearly the traditional one dimensional rubber banding isn't acceptable, it's good to see racing games exploring more sensible strategies!

I would probably be considered a hard-core racer. I prefer games like Forza and RacePro although I do pop in NFS or burnout once in a while to spice things up. Rubber banding should be outlawed in any racing game. I would more easily accept losing because of my mistakes than being assisted by the rubber banding effect. On the other hand I do feel cheated when I race a perfect race and the AI are still able to win because of a small mistake in the last part. If I'm better than the AI, then I should be allowed to put some distance between myself and the AI to protect my lead in the event of a mistake.

I think the implementation described here seems very reasonable although it seems as if it will create a similar experience regardless of player skill. This seems a bit odd given all the development and I'm sure that the higher difficulty levels were tweaked to compensate for that. Good article and definitely peaked my interest to buy the game and experience the implementation for myself.

i think the basic idea of rubber banding isn't that unrealistic. consider a real race. if you happen to find yourself miles in front, perhaps you prosecute the corners a little sloppier. perhaps you get bored and make minor mistakes, or else wish to conserve your cars tires and engine by using less engine revs.