Warrior General Games

Originally posted by Bort
Reposting because you didn't get to my Q, steel:

Originally posted by Bort

What if I put in something that's going to give more fractional type results, like 7%, or 33%? What is the formula you're applying, and do you floor, round, ceiling?



Sorry, missed you. You really need to get a higher profile if you want to survive in GLB

This is real simple way of doing it, and it is close enough for horse shoes and hand gernades. Everything input has a counter, and every output has a ticker every time it occurs. Ignore the first roll. After the first roll, each output ticker is divided by the play counter. If the percentage is high than the set percent, than that output is excluded for that roll. It keeps doing that check every time a roll is made.

Input
play A - 53%
play B - 17%
play C - 30%

Roll 1
play A occurs

Roll 2
play B occurs

Roll 3
play B is excluded from the roll as its ticker value / input counter is .33

Roll 4
play B is still excluded at .25

Roll 5
play B is still excluded at .20

Roll 6
play B is back at .1666

If you map that out at over a number of rolls, you'll get rough percentages. The ordered list way would be more coding.

Aha! Now I getcha. Makes sense.

I might try it out just to see how I like it.

Originally posted by Bort
Aha! Now I getcha. Makes sense.

I might try it out just to see how I like it.

Any response to the idea of making players remember the way certain things happen within a play. My main idea was QBs and the target receiver, but this could go for defenders and tackling a player who faked them out earlier. Maybe they're less likely to fall for it again because they 'expect it', or a LDE remembering that last time the O came out in Strong I, they ran a screen, so he's a little less aggressive against the pass, etc.

Originally posted by Staz
Any response to the idea of making players remember the way certain things happen within a play. My main idea was QBs and the target receiver, but this could go for defenders and tackling a player who faked them out earlier. Maybe they're less likely to fall for it again because they 'expect it', or a LDE remembering that last time the O came out in Strong I, they ran a screen, so he's a little less aggressive against the pass, etc.

There is some of that already in play, actually. If a player keeps getting hit with passes, his risk value increases, and defenders will cover him closer, etc. It's not really applied in a ton of areas, though.

Originally posted by Bort
There is some of that already in play, actually. If a player keeps getting hit with passes, his risk value increases, and defenders will cover him closer, etc. It's not really applied in a ton of areas, though.

I definitely think that would help players perform in a more logical, "human-like" way by adding a sort of memory.

the best defensive AIs in this game have 1 play against multiple formations rather than multiple plays against minimal formations...
therefore, the %'s should be irrelevant.

Just to add to this thread.

In a game yesterday I had a setting with
40% Play 1
25% Play 2
35% Play 3

In 7 instances there were 6 Play 1 called and 1 Play 3 called. 0 Play 2 were called.

In another input I had
35% Play 1
25% Play 2
20% Play 3
20% Play 4

In 8 instances Play 1 was called 1 time, Play 2 was called 4 times, Play 3 once and Play 4 once.

In both cases, auto adjust is off.
The probability of these distributions is low. Anything that creates a more even distribution matching my intended play calling is a good thing.

What about using a Sampling Without Replacement strategy with, say, 100 units binned for each package?

Example
Play A 25% or 25 units
Play B 25% or 25 units
Play C 25% or 25 units
Play D 25% or 25 units

When auto-adjust is set to 0%, then it's a true random without replacement strategy. Over 100 plays, each play will be called 25% of the time.

If auto-adjust is raised above zero, then you do replace when a play succeeds (more if AA is higher). And when a play fails, you don't just subtract the counter a single unit; you subtract additional units (depending on if AA is higher).

I really like Steel's idea.

IMO anyone who has the opinion that it needs to be more random than the proposed system is either a problem gambler or isn't thinking about it hard enough. The level of randomness in the current system provides virtually no advantages.

Bump for editing in Steel's idea to the OP.

Trifecta

in light of recent updates to OP, i think +1 is in order.

Originally posted by steellithium
Sorry, missed you. You really need to get a higher profile if you want to survive in GLB

This is real simple way of doing it, and it is close enough for horse shoes and hand gernades. Everything input has a counter, and every output has a ticker every time it occurs. Ignore the first roll. After the first roll, each output ticker is divided by the play counter. If the percentage is high than the set percent, than that output is excluded for that roll. It keeps doing that check every time a roll is made.

Input
play A - 53%
play B - 17%
play C - 30%

Roll 1
play A occurs

Roll 2
play B occurs

Roll 3
play B is excluded from the roll as its ticker value / input counter is .33

Roll 4
play B is still excluded at .25

Roll 5
play B is still excluded at .20

Roll 6
play B is back at .1666

If you map that out at over a number of rolls, you'll get rough percentages. The ordered list way would be more coding.

So which play occurs on rolls 3, 4, 5 & 6? Would it be A, C, A, B or A, C, A, C? Not getting if it will start at the top of the list with each roll and go down until it finds a play that's not excluded. Or if it will choose the play with the highest percentage that is not excluded with each roll. Seems like a great idea though. Will you still be able to auto adjust?

Originally posted by mwindle

So which play occurs on rolls 3, 4, 5 & 6? Would it be A, C, A, B or A, C, A, C?

Roll 3 could either be play A or play C. What happens on 4 through 6 would depend on what happened on 3.

Originally posted by mwindle
Not getting if it will start at the top of the list with each roll and go down until it finds a play that's not excluded. Or if it will choose the play with the highest percentage that is not excluded with each roll.

Neither. A random roll will occur with each possible play having the proper weight.

Originally posted by mwindle
Will you still be able to auto adjust?

Yes.

Bump. I'd really like this to happen!