Warrior General Games

Originally posted by Hagalaz
If that's the issue, why not solve it with simple vectors? Assign "threat" values to each incoming defender and "importance" value to your goal, each of these is the length of a vector, add all 3 and the HB suddenly knows the instant best way considering all defenders at once. It doesn't consider blockers, but it's a pretty simple solution. It even considers backpeddling if the defender threat value goes beyond the "reward" or "importance" value of maintaining the same general direction.

Was used in one of the best bots of all time back in the 90s in quake 1 with great success!

If the defender is threatening enough and fast enough, and the situation dire enough (no time on the clock, for instance)....could be pretty humorous

Originally posted by Hagalaz
Originally posted by Deathblade

It's not that vision doesn't have an effect, or that it's scaled too far out (to accommodate 150 vision players)...it's that the "best" path...isn't.

Otherwise, if vision was scaled too far, or had no effect...it wouldn't actually hurt you. However, it does actually hurt you when you go up to 70+. You get dancing that you don't with low vision.

Really, I think there is something wrong with Bort's vector calculation. I mean, if there's 2 defenders coming at you from two directions...what kind of calculation would make you alternate 90 degree angles between them? The logical "angle" would be calculating their vectors and running between it, rather than alternating between running straight at defender A, and straight at defender B.

If that's the issue, why not solve it with simple vectors? Assign "threat" values to each incoming defender and "importance" value to your goal, each of these is the length of a vector, add all 3 and the HB suddenly knows the instant best way considering all defenders at once. It doesn't consider blockers, but it's a pretty simple solution. It even considers backpeddling if the defender threat value goes beyond the "reward" or "importance" value of maintaining the same general direction.

Was used in one of the best bots of all time back in the 90s in quake 1 with great success!

That sounds great if they could find a way to account for blockers and blockers ability to pick up a defender with a move the HB makes.

Originally posted by burn_209
Semi-douchebag but mostly helpful

Closer, i think

Originally posted by jrry32
That sounds great if they could find a way to account for blockers and blockers ability to pick up a defender with a move the HB makes.

Honestly...if the whole "calculate distance that can be traveled" thing is somehow done...calculations could be made on all of the blockers to see if they could get to to an intercept point between the defender and point X.

But really...most of this is probably just pipe dreams, with huge flaws in it.

Originally posted by TyrannyVaunt
http://www.youtube.com/watch?v=BYROBfMPCso


Laterals!

get on it Bort!

Originally posted by tautology
Laterals!

get on it Bort!

lol'd at running over the band

Originally posted by Deathblade
Honestly...if the whole "calculate distance that can be traveled" thing is somehow done...calculations could be made on all of the blockers to see if they could get to to an intercept point between the defender and point X.

But really...most of this is probably just pipe dreams, with huge flaws in it.

It is a whole discipline in itself, to work through all the pathing. That's why I would be okay with some interesting workarounds.

Originally posted by Deathblade
Originally posted by jrry32

That sounds great if they could find a way to account for blockers and blockers ability to pick up a defender with a move the HB makes.

Honestly...if the whole "calculate distance that can be traveled" thing is somehow done...calculations could be made on all of the blockers to see if they could get to to an intercept point between the defender and point X.

But really...most of this is probably just pipe dreams, with huge flaws in it.

Well lets dream big and see what can be done!

Might just start with the sidelines threat not being affected by what running style you're on.

Then go to trying to understand how to build a dynamic 2d map of everywhere the HB would like to go weighted by threats, blockers, and personal tactics.

Originally posted by Deathblade
Originally posted by tautology

Laterals!

get on it Bort!

lol'd at running over the band

Even funnier was when the Saints did that vs. the Jags, they scored on a miracle TD after laterals then missed the extra point to lose the game by 1 point.

Originally posted by Deathblade
Honestly...if the whole "calculate distance that can be traveled" thing is somehow done...calculations could be made on all of the blockers to see if they could get to to an intercept point between the defender and point X.

But really...most of this is probably just pipe dreams, with huge flaws in it.

I eh... I don't think this method can work with friendlies, because it's not just potential blocker movement that matters, but also, eh how to explain this, in what direction will the blocker push/hold the defender. This method can't really work with that, it would take a lot of change for it to work ^^ although I don't think it is impossible, it stops being a simple solution to turn into a research into a possible quite harder solution

EDIT: Although if a block is already established, it can work as a reduction to the threat value of a defender.

Originally posted by kurieg
Might just start with the sidelines threat not being affected by what running style you're on.

Then go to trying to understand how to build a dynamic 2d map of everywhere the HB would like to go weighted by threats, blockers, and personal tactics.

20 years since I took vector calculus...not going to be a lot of help, except to know that it is a fairly tough problem to actually make it accommodate different styles and look like football.

Originally posted by kurieg
Might just start with the sidelines threat not being affected by what running style you're on.

Then go to trying to understand how to build a dynamic 2d map of everywhere the HB would like to go weighted by threats, blockers, and personal tactics.

The sideline is what makes me actually think it works like I think it does.

The whole "run into defender" thing on the sideline. Bort said that the sideline gave off a vector equal to the defender, or something like that...which "makes sense" in that it would cause the HB to not run out of bounds. However, it doesn't make sense in the sense that the HB thinks the sideline is running at him.

The sideline has zero mobility...and when faced with a defender and the sideline, an inch from the sideline should always be your goal...since the sideline doesn't move. It has no "danger area" like a defender does, you can get as close to it as you want, and have zero fear of it getting you.

Originally posted by Hagalaz
Originally posted by Deathblade

Honestly...if the whole "calculate distance that can be traveled" thing is somehow done...calculations could be made on all of the blockers to see if they could get to to an intercept point between the defender and point X.

But really...most of this is probably just pipe dreams, with huge flaws in it.

I eh... I don't think this method can work with friendlies, because it's not just potential blocker movement that matters, but also, eh how to explain this, in what direction will the blocker push/hold the defender. This method can't really work with that, it would take a lot of change for it to work ^^ although I don't think it is impossible, it stops being a simple solution to turn into a research into a possible quite harder solution

EDIT: Although if a block is already established, it can work as a reduction to the threat value of a defender.

The play has to be set so that say for instance... It is a pitch left. The left side is naturally a reduction to the threat value for the HB. The offensive blocking interaction would use the same pathing to block left (FB leads left)... And as you stated, if the block is already engaged, the HB would view that as a lower threat value. Couldn't this work?