[fighting game] Kinetic Damage

[BlueRadius]

you a fan of tekken? I always played tekken tag back in the ps2 days. 3 consecutive electrics with any mashima + wave dashing = epic win. But Devils twin pistons are a beast when used properly. Ive already played Tekken Tag 2 and its great. Pauls just frame is easier to pull with me for some reason.

The most anticipated fighter im waiting for is Tekken VS SF....

Also im glad Fuudo won the grand finals at EVO 2K11 lol, hes a beast

[n0mad]

Aaaah thank you BlueRadius for the vid, I so wanted to watch it but didn't find it ! Plus, Fei Long being my main since SFIV, I can't be more happy to see him make it to the top

I like Tekken too, but have a big preference for the technicity of SFIV ^^

edit : wow, Fuudo's focus cancels on normals are outstandingly well placed ...

[BlueRadius]

Aaaah thank you BlueRadius for the vid, I so wanted to watch it but didn't find it ! Plus, Fei Long being my main since SFIV, I can't be more happy to see him make it to the top

I like Tekken too, but have a big preference for the technicity of SFIV ^^

edit : wow, Fuudo's focus cancels on normals are outstandingly well placed ...

Ah yes SF's technicity is top tier.

Yea dude Fuudo is a problem, i can never use Fei Longs D,B,4 on such good timing like him. Plus he has mad recovery!

[n0mad]

I hope he will be back next year
And I seriously hope Fei will be put into SFxTekken cast

[n0mad]

Some news !

Actually working on CPU's AI, having a truckload of fun creating it

I won't go into the classical fighting games way of doing it (aka. instant reaction to any player's move, which can lead to frustrating fights on highest difficulty), but more like a human brain approach to it :

For example, CPU will analyze in real time what strikes his opponent is able to do in the next seconds, and what strikes itself will be able to place under these predictions. Yes, a real-time, dynamic strategy.
CPU will base his actions upon an Analyze/Decision pattern, depending on a bunch of realistic stats, like reactivity, strategy, agressivity, etc. Those stats will change between opponent to opponent, being tweaked or decreased according to the AI's difficulty level.
Interesting thing is that they will be dynamically updated through the fight according to the opponent's decisions.

For example, a particularly agressive AI can suddenly turn over-defensive against an opponent who manage to own the fight. Or at the opposite, a basically defensive AI can become an unstoppable ball of strikes if the opponent is analyzed to be shy or not very skilled.

Ultimately, this will let the AI create non-linear fights, and more importantly a nearly infinite combination of different single-player fight styles.

[Neptune_Imaging]

That sounds so cool...groovy... I've always hated AI that seemingly read your inputs and causes anger. Hopefully I can learn how to create different personalities for each fighter... keep this up

[n0mad]

Thanks
I admit it's not a trivial task... Plus creating an AI is the most open door to bugs and crashes imho.
But anyway, even if I'm struggling, it's a very fun part of the job

[Neptune_Imaging]

Thanks
I admit it's not a trivial task... Plus creating an AI is the most open door to bugs and crashes imho.
But anyway, even if I'm struggling, it's a very fun part of the job

I am still struggling with the level design...so many ideas... and I hate using the terrain sometimes

[n0mad]

Take your best shot

[Neptune_Imaging]

Take your best shot

No doubt... I will post some new stuff when I get home

[n0mad]

Hey fellows,

I'm coming back from an interesting topic about State Machine coding (or Actions Authorization Manager), and I thought it could be interesting to expose how I chosed to handle a fighting game.

State Machines are a pain in the ass, litterally, but they are necessary.
I tried many iterations of different models, and finally found a simple, effective one :

Basically, everything is driven by the animations instead of hard coded values.
This means that instead of creating an ecosystem of different states (like OnActionStart, OnActionEnd, OnActionRun), the framework just checks the current animation state.

If anim is not playing anymore, it asks to run the next action, which can be :

• a queued action (ex : user pushed "move forward" while jumping, but it wasn't authorized, so the action is queued)
• an automated chain action (ex : "falling" must automatically be followed by "landing")

Adding to that the possibility to override current anim by a higher priority action, of course (ex : striking over walking).

So, this architecture requires 3 things :

• a function that can tell if one action is authorized to override another or not
• a dictionary containing each chained action for each animation
• a classification of actions. I chosed to create 2 dictionaries containing each animation's info about its posture (crouch, stand, jump, down), and its type of action (hit, strike, jump, idle, etc)

The "Authorizer" function uses the current anim's posture, and type of action to determine if the requested action is authorized to override, or not.
Example : while jumping, system asks to Authorizer the action "crouch" (= player pressing down key). Instead of testing by current action's animation name, which could produce neverending condition clauses, the function will check its posture within the appropriate dictionary. Therefore, whenever the system asks for an action that is labelled "crouch" or "stand", while current action is labelled "jump", Authorizer will say no.

Now the pipeline looks like this :

1. launch action -> launch animation
2. a coroutine is checking if the current animation is finished or not. If it is finished, it launches its chain action, and then asks if queued action is able to be performed. If yes, it overrides chain action with queued action.
3. whenever a higher priority action is requested over the current one, Authorizer tells if it is possible or not.

Et voilà.

The hardest part is the Authorizer and its batches of conditions priorities, but having those different Enum infos about each anim really eases the work.

The advantages of such a method are :

• re-usability : this SM is not restricted to execution only, as the Auhorizer can be used for virtual tests, like AI.
• flexibility : may the action request come from a button input or an AI, the same routine pipeline is used. Making the inputs inherit the AI pipeline ensures more stability than creating a parallel one just for the player.
• easy tweaking : with the chain actions dictionary, I can change in one second what movement should be followed by one other
• easy SM coding : storing posture enums (stand, crouch, jump) and actionType enums (hit, strike, special, movement, etc) for each animation greatly helps in regrouping some similar authorizations into one, by testing the actiontype instead of its absolute name.
• easy AI coding : AI can now focus on pure virtual intelligence creation, rather than performing various physical checks, thanks to the Authorizer function. It can also gather a ton of info in no time thanks to Enums.
  Ex : making a list of available moves depending on the actual action, and then provide this list to the core analysis/decision routine.

Actually the AI system comes really simplified with such states handling, to a point where I can just write some pseudo code, traduce it in an absolute anim name and submit it to the Authorizer.

Hope this blog helped

[n0mad]

Hey there, I've made a compilation of all the tips and techniques I found during the whole development of Kinetic Damage. The list will be growing until the end of development.

direct link.

[n0mad]

Here it is, how much you can actually customize your fighter at creation.
There are 8 more special class costumes that are not demoed in this video.

[hippocoder]

Quite impressive. I am not sure about the heads being headless and some of the poses of the male look a bit spastic when changing trousers, it should look a bit more heroic there.

But yeah great stuff overall

I think it will sell buckets.

[n0mad]

Thanks hippo !
Ah the headless is in fact not headless, it's just that the video compression oversaturate the colors, which make the fading-from-white effect seamless with the background
Ok for the trousers pose, I'll check into that

[ratonmalo]

amazing progress! looks great!

[Alec]

Great to see how far you've come Nomad

[n0mad]

Oh guys thank you so much, that's strengthening to read such comments. Times are rough, but I'll make it till the end I promise.
On my way to finish special strikes, then a mini scenario to keep solo players away from boredom, then a simple multiplayer, plus other secret unique features, and that's it.
The pack should be ready for Q1 2012.

I just tested on iPhone 4S, and it's over 60fps. Should let me insert some fancy one-click computed effects like Pro water, bloom, etc.
Loading times are super fast. Fighting flow looks exciting too, thanks to huge efforts on conceiving coherent starting/ending poses.
I hope players will enjoy fighting with their own character as much as I am.

[n0mad]

Oh and I suscribed for my 4th year of Taekwondo lessons 2 weeks ago, adding a new year into 13 years of Martial Arts practicing.
It will help me refine how a technical fight does feel in reality, therefore refine how fights in Kinetic Damage should feel too.

Ideally, I should have suscribed for real Taekwondo local tournaments this year too, to step up that perception, but I clearly don't have enough time.

[Neptune_Imaging]

Can't wait to play